chore(deps): bump prismjs

Bumps the npm_and_yarn group with 1 update in the /frontend directory: [prismjs](https://github.com/PrismJS/prism).


Updates `prismjs` from 1.27.0 to 1.30.0
- [Release notes](https://github.com/PrismJS/prism/releases)
- [Changelog](https://github.com/PrismJS/prism/blob/v2/CHANGELOG.md)
- [Commits](https://github.com/PrismJS/prism/compare/v1.27.0...v1.30.0)

---
updated-dependencies:
- dependency-name: prismjs
  dependency-version: 1.30.0
  dependency-type: indirect
  dependency-group: npm_and_yarn
...

Signed-off-by: dependabot[bot] <support@github.com>

.github/dependabot.yml

@@ -13,7 +13,11 @@ updates:
     directory: "/frontend" # Location of package manifests
     schedule:
       interval: "daily"
+  - package-ecosystem: "npm"
+    directory: "/extensions/react-widget"
+    schedule:
+      interval: "daily"
   - package-ecosystem: "github-actions"
     directory: "/"
     schedule:
-      interval: "daily"
+      interval: "daily"

.github/styles/DocsGPT/Spelling.yml

@@ -0,0 +1,11 @@
+extends: spelling
+level: warning
+message: "Did you really mean '%s'?"
+ignore:
+  - "**/node_modules/**"
+  - "**/dist/**"
+  - "**/build/**"
+  - "**/coverage/**"
+  - "**/public/**"
+  - "**/static/**"
+vocab: DocsGPT

.github/styles/config/vocabularies/DocsGPT/accept.txt

@@ -0,0 +1,46 @@
+Ollama
+Qdrant
+Milvus
+Chatwoot
+Nextra
+VSCode
+npm
+LLMs
+APIs
+Groq
+SGLang
+LMDeploy
+OAuth
+Vite
+LLM
+JSONPath
+UIs
+configs
+uncomment
+qdrant
+vectorstore
+docsgpt
+llm
+GPUs
+kubectl
+Lightsail
+enqueues
+chatbot
+VSCode's
+Shareability
+feedbacks
+automations
+Premade
+Signup
+Repo
+repo
+env
+URl
+agentic
+llama_cpp
+parsable
+SDKs
+boolean
+bool
+hardcode
+EOL

.github/workflows/vale.yml

@@ -0,0 +1,26 @@
+name: Vale Documentation Linter
+
+on:
+  pull_request:
+    paths:
+      - 'docs/**/*.md'
+      - 'docs/**/*.mdx'
+      - '**/*.md'
+      - '.vale.ini'
+      - '.github/styles/**'
+
+jobs:
+  vale:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+
+      - name: Vale linter
+        uses: errata-ai/vale-action@v2
+        with:
+          files: docs
+          fail_on_error: false
+          version: 3.0.5
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.gitignore

@@ -3,6 +3,7 @@ __pycache__/
 *.py[cod]
 *$py.class
 
+experiments
 # C extensions
 *.so
 *.next

.vale.ini

@@ -0,0 +1,5 @@
+MinAlertLevel = warning
+StylesPath = .github/styles
+
+[*.{md,mdx}]
+BasedOnStyles = DocsGPT

CONTRIBUTING.md

@@ -147,5 +147,5 @@ Here's a step-by-step guide on how to contribute to DocsGPT:
 Thank you for considering contributing to DocsGPT! 🙏
 
 ## Questions/collaboration
-Feel free to join our [Discord](https://discord.gg/n5BX8dh8rU). We're very friendly and welcoming to new contributors, so don't hesitate to reach out.
+Feel free to join our [Discord](https://discord.gg/vN7YFfdMpj). We're very friendly and welcoming to new contributors, so don't hesitate to reach out.
 # Thank you so much for considering to contributing DocsGPT!🙏

HACKTOBERFEST.md

@@ -3,6 +3,7 @@
 Welcome, contributors! We're excited to announce that DocsGPT is participating in Hacktoberfest. Get involved by submitting meaningful pull requests.
 
 All Meaningful contributors with accepted PRs that were created for issues with the `hacktoberfest` label (set by our maintainer team: dartpain, siiddhantt, pabik, ManishMadan2882) will receive a cool T-shirt! 🤩.
+<img width="1331" height="678" alt="hacktoberfest-mocks-preview" src="https://github.com/user-attachments/assets/633f6377-38db-48f5-b519-a8b3855a9eb4" />
 
 Fill in [this form](https://forms.gle/Npaba4n9Epfyx56S8
 ) after your PR was merged please 
@@ -31,7 +32,7 @@ Non-Code Contributions:
 - Before contributing check existing [issues](https://github.com/arc53/DocsGPT/issues) or [create](https://github.com/arc53/DocsGPT/issues/new/choose) an issue and wait to get assigned.
 - Once you are finished with your contribution, please fill in this [form](https://forms.gle/Npaba4n9Epfyx56S8).
 - Refer to the [Documentation](https://docs.docsgpt.cloud/).
-- Feel free to join our [Discord](https://discord.gg/n5BX8dh8rU) server. We're here to help newcomers, so don't hesitate to jump in! Join us [here](https://discord.gg/n5BX8dh8rU).
+- Feel free to join our [Discord](https://discord.gg/vN7YFfdMpj) server. We're here to help newcomers, so don't hesitate to jump in! Join us [here](https://discord.gg/vN7YFfdMpj).
   
 Thank you very much for considering contributing to DocsGPT during Hacktoberfest! 🙏 Your contributions (not just simple typos) could earn you a stylish new t-shirt.
 

README.md

@@ -16,10 +16,10 @@
   <a href="https://github.com/arc53/DocsGPT">![link to main GitHub showing Forks number](https://img.shields.io/github/forks/arc53/docsgpt?style=social)</a>
   <a href="https://github.com/arc53/DocsGPT/blob/main/LICENSE">![link to license file](https://img.shields.io/github/license/arc53/docsgpt)</a>
   <a href="https://www.bestpractices.dev/projects/9907"><img src="https://www.bestpractices.dev/projects/9907/badge"></a>
-  <a href="https://discord.gg/n5BX8dh8rU">![link to discord](https://img.shields.io/discord/1070046503302877216)</a>
+  <a href="https://discord.gg/vN7YFfdMpj">![link to discord](https://img.shields.io/discord/1070046503302877216)</a>
   <a href="https://x.com/docsgptai">![X (formerly Twitter) URL](https://img.shields.io/twitter/follow/docsgptai)</a>
 
-<a href="https://docs.docsgpt.cloud/quickstart">⚡️ Quickstart</a> • <a href="https://app.docsgpt.cloud/">☁️ Cloud Version</a> • <a href="https://discord.gg/n5BX8dh8rU">💬 Discord</a>
+<a href="https://docs.docsgpt.cloud/quickstart">⚡️ Quickstart</a> • <a href="https://app.docsgpt.cloud/">☁️ Cloud Version</a> • <a href="https://discord.gg/vN7YFfdMpj">💬 Discord</a>
 <br>
 <a href="https://docs.docsgpt.cloud/">📖 Documentation</a> • <a href="https://github.com/arc53/DocsGPT/blob/main/CONTRIBUTING.md">👫 Contribute</a> • <a href="https://blog.docsgpt.cloud/">🗞 Blog</a>
 <br>

application/agents/agent_creator.py

@@ -1,5 +1,8 @@
 from application.agents.classic_agent import ClassicAgent
 from application.agents.react_agent import ReActAgent
+import logging
+
+logger = logging.getLogger(__name__)
 
 
 class AgentCreator:
@@ -13,4 +16,5 @@ class AgentCreator:
         agent_class = cls.agents.get(type.lower())
         if not agent_class:
             raise ValueError(f"No agent class found for type {type}")
+        
         return agent_class(*args, **kwargs)

application/agents/base.py

@@ -12,7 +12,6 @@ from application.core.settings import settings
 from application.llm.handlers.handler_creator import LLMHandlerCreator
 from application.llm.llm_creator import LLMCreator
 from application.logging import build_stack_data, log_activity, LogContext
-from application.retriever.base import BaseRetriever
 
 logger = logging.getLogger(__name__)
 
@@ -22,11 +21,12 @@ class BaseAgent(ABC):
         self,
         endpoint: str,
         llm_name: str,
-        gpt_model: str,
+        model_id: str,
         api_key: str,
         user_api_key: Optional[str] = None,
         prompt: str = "",
         chat_history: Optional[List[Dict]] = None,
+        retrieved_docs: Optional[List[Dict]] = None,
         decoded_token: Optional[Dict] = None,
         attachments: Optional[List[Dict]] = None,
         json_schema: Optional[Dict] = None,
@@ -37,7 +37,7 @@ class BaseAgent(ABC):
     ):
         self.endpoint = endpoint
         self.llm_name = llm_name
-        self.gpt_model = gpt_model
+        self.model_id = model_id
         self.api_key = api_key
         self.user_api_key = user_api_key
         self.prompt = prompt
@@ -52,7 +52,9 @@ class BaseAgent(ABC):
             api_key=api_key,
             user_api_key=user_api_key,
             decoded_token=decoded_token,
+            model_id=model_id,
         )
+        self.retrieved_docs = retrieved_docs or []
         self.llm_handler = LLMHandlerCreator.create_handler(
             llm_name if llm_name else "default"
         )
@@ -65,13 +67,13 @@ class BaseAgent(ABC):
 
     @log_activity()
     def gen(
-        self, query: str, retriever: BaseRetriever, log_context: LogContext = None
+        self, query: str, log_context: LogContext = None
     ) -> Generator[Dict, None, None]:
-        yield from self._gen_inner(query, retriever, log_context)
+        yield from self._gen_inner(query, log_context)
 
     @abstractmethod
     def _gen_inner(
-        self, query: str, retriever: BaseRetriever, log_context: LogContext
+        self, query: str, log_context: LogContext
     ) -> Generator[Dict, None, None]:
         pass
 
@@ -150,6 +152,7 @@ class BaseAgent(ABC):
         call_id = getattr(call, "id", None) or str(uuid.uuid4())
 
         # Check if parsing failed
+
         if tool_id is None or action_name is None:
             error_message = f"Error: Failed to parse LLM tool call. Tool name: {getattr(call, 'name', 'unknown')}"
             logger.error(error_message)
@@ -164,13 +167,14 @@ class BaseAgent(ABC):
             yield {"type": "tool_call", "data": {**tool_call_data, "status": "error"}}
             self.tool_calls.append(tool_call_data)
             return "Failed to parse tool call.", call_id
-
         # Check if tool_id exists in available tools
+
         if tool_id not in tools_dict:
             error_message = f"Error: Tool ID '{tool_id}' extracted from LLM call not found in available tools_dict. Available IDs: {list(tools_dict.keys())}"
             logger.error(error_message)
 
             # Return error result
+
             tool_call_data = {
                 "tool_name": "unknown",
                 "call_id": call_id,
@@ -181,7 +185,6 @@ class BaseAgent(ABC):
             yield {"type": "tool_call", "data": {**tool_call_data, "status": "error"}}
             self.tool_calls.append(tool_call_data)
             return f"Tool with ID {tool_id} not found.", call_id
-
         tool_call_data = {
             "tool_name": tools_dict[tool_id]["name"],
             "call_id": call_id,
@@ -223,6 +226,7 @@ class BaseAgent(ABC):
         tm = ToolManager(config={})
 
         # Prepare tool_config and add tool_id for memory tools
+
         if tool_data["name"] == "api_tool":
             tool_config = {
                 "url": tool_data["config"]["actions"][action_name]["url"],
@@ -234,8 +238,8 @@ class BaseAgent(ABC):
             tool_config = tool_data["config"].copy() if tool_data["config"] else {}
             # Add tool_id from MongoDB _id for tools that need instance isolation (like memory tool)
             # Use MongoDB _id if available, otherwise fall back to enumerated tool_id
-            tool_config["tool_id"] = str(tool_data.get("_id", tool_id))
 
+            tool_config["tool_id"] = str(tool_data.get("_id", tool_id))
         tool = tm.load_tool(
             tool_data["name"],
             tool_config=tool_config,
@@ -276,24 +280,14 @@ class BaseAgent(ABC):
         self,
         system_prompt: str,
         query: str,
-        retrieved_data: List[Dict],
     ) -> List[Dict]:
-        docs_with_filenames = []
-        for doc in retrieved_data:
-            filename = doc.get("filename") or doc.get("title") or doc.get("source")
-            if filename:
-                chunk_header = str(filename)
-                docs_with_filenames.append(f"{chunk_header}\n{doc['text']}")
-            else:
-                docs_with_filenames.append(doc["text"])
-        docs_together = "\n\n".join(docs_with_filenames)
-        p_chat_combine = system_prompt.replace("{summaries}", docs_together)
-        messages_combine = [{"role": "system", "content": p_chat_combine}]
+        """Build messages using pre-rendered system prompt"""
+        messages = [{"role": "system", "content": system_prompt}]
 
         for i in self.chat_history:
             if "prompt" in i and "response" in i:
-                messages_combine.append({"role": "user", "content": i["prompt"]})
-                messages_combine.append({"role": "assistant", "content": i["response"]})
+                messages.append({"role": "user", "content": i["prompt"]})
+                messages.append({"role": "assistant", "content": i["response"]})
             if "tool_calls" in i:
                 for tool_call in i["tool_calls"]:
                     call_id = tool_call.get("call_id") or str(uuid.uuid4())
@@ -313,29 +307,17 @@ class BaseAgent(ABC):
                         }
                     }
 
-                    messages_combine.append(
+                    messages.append(
                         {"role": "assistant", "content": [function_call_dict]}
                     )
-                    messages_combine.append(
+                    messages.append(
                         {"role": "tool", "content": [function_response_dict]}
                     )
-        messages_combine.append({"role": "user", "content": query})
-        return messages_combine
-
-    def _retriever_search(
-        self,
-        retriever: BaseRetriever,
-        query: str,
-        log_context: Optional[LogContext] = None,
-    ) -> List[Dict]:
-        retrieved_data = retriever.search(query)
-        if log_context:
-            data = build_stack_data(retriever, exclude_attributes=["llm"])
-            log_context.stacks.append({"component": "retriever", "data": data})
-        return retrieved_data
+        messages.append({"role": "user", "content": query})
+        return messages
 
     def _llm_gen(self, messages: List[Dict], log_context: Optional[LogContext] = None):
-        gen_kwargs = {"model": self.gpt_model, "messages": messages}
+        gen_kwargs = {"model": self.model_id, "messages": messages}
 
         if (
             hasattr(self.llm, "_supports_tools")
@@ -343,7 +325,6 @@ class BaseAgent(ABC):
             and self.tools
         ):
             gen_kwargs["tools"] = self.tools
-
         if (
             self.json_schema
             and hasattr(self.llm, "_supports_structured_output")
@@ -357,7 +338,6 @@ class BaseAgent(ABC):
                     gen_kwargs["response_format"] = structured_format
                 elif self.llm_name == "google":
                     gen_kwargs["response_schema"] = structured_format
-
         resp = self.llm.gen_stream(**gen_kwargs)
 
         if log_context:

application/agents/classic_agent.py

@@ -1,32 +1,20 @@
+import logging
 from typing import Dict, Generator
+
 from application.agents.base import BaseAgent
 from application.logging import LogContext
-from application.retriever.base import BaseRetriever
-import logging
 
 logger = logging.getLogger(__name__)
 
 
 class ClassicAgent(BaseAgent):
-    """A simplified agent with clear execution flow.
-
-    Usage:
-    1. Processes a query through retrieval
-    2. Sets up available tools
-    3. Generates responses using LLM
-    4. Handles tool interactions if needed
-    5. Returns standardized outputs
-
-    Easy to extend by overriding specific steps.
-    """
+    """A simplified agent with clear execution flow"""
 
     def _gen_inner(
-        self, query: str, retriever: BaseRetriever, log_context: LogContext
+        self, query: str, log_context: LogContext
     ) -> Generator[Dict, None, None]:
-        # Step 1: Retrieve relevant data
-        retrieved_data = self._retriever_search(retriever, query, log_context)
+        """Core generator function for ClassicAgent execution flow"""
 
-        # Step 2: Prepare tools
         tools_dict = (
             self._get_user_tools(self.user)
             if not self.user_api_key
@@ -34,20 +22,16 @@ class ClassicAgent(BaseAgent):
         )
         self._prepare_tools(tools_dict)
 
-        # Step 3: Build and process messages
-        messages = self._build_messages(self.prompt, query, retrieved_data)
+        messages = self._build_messages(self.prompt, query)
         llm_response = self._llm_gen(messages, log_context)
 
-        # Step 4: Handle the response
         yield from self._handle_response(
             llm_response, tools_dict, messages, log_context
         )
 
-        # Step 5: Return metadata
-        yield {"sources": retrieved_data}
+        yield {"sources": self.retrieved_docs}
         yield {"tool_calls": self._get_truncated_tool_calls()}
 
-        # Log tool calls for debugging
         log_context.stacks.append(
             {"component": "agent", "data": {"tool_calls": self.tool_calls.copy()}}
         )

application/agents/react_agent.py

@@ -1,284 +1,238 @@
-import os
-from typing import Dict, Generator, List, Any
 import logging
+import os
+from typing import Any, Dict, Generator, List
 
 from application.agents.base import BaseAgent
 from application.logging import build_stack_data, LogContext
-from application.retriever.base import BaseRetriever
 
 logger = logging.getLogger(__name__)
 
+MAX_ITERATIONS_REASONING = 10
+
 current_dir = os.path.dirname(
     os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
 )
 with open(
     os.path.join(current_dir, "application/prompts", "react_planning_prompt.txt"), "r"
 ) as f:
-    planning_prompt_template = f.read()
+    PLANNING_PROMPT_TEMPLATE = f.read()
 with open(
-    os.path.join(current_dir, "application/prompts", "react_final_prompt.txt"),
-    "r",
+    os.path.join(current_dir, "application/prompts", "react_final_prompt.txt"), "r"
 ) as f:
-    final_prompt_template = f.read()
-
-MAX_ITERATIONS_REASONING = 10
+    FINAL_PROMPT_TEMPLATE = f.read()
 
 
 class ReActAgent(BaseAgent):
+    """
+    Research and Action (ReAct) Agent - Advanced reasoning agent with iterative planning.
+
+    Implements a think-act-observe loop for complex problem-solving:
+    1. Creates a strategic plan based on the query
+    2. Executes tools and gathers observations
+    3. Iteratively refines approach until satisfied
+    4. Synthesizes final answer from all observations
+    """
+
     def __init__(self, *args, **kwargs):
         super().__init__(*args, **kwargs)
         self.plan: str = ""
         self.observations: List[str] = []
 
-    def _extract_content_from_llm_response(self, resp: Any) -> str:
-        """
-        Helper to extract string content from various LLM response types.
-        Handles strings, message objects (OpenAI-like), and streams.
-        Adapt stream handling for your specific LLM client if not OpenAI.
-        """
-        collected_content = []
-        if isinstance(resp, str):
-            collected_content.append(resp)
-        elif (  # OpenAI non-streaming or Anthropic non-streaming (older SDK style)
-            hasattr(resp, "message")
-            and hasattr(resp.message, "content")
-            and resp.message.content is not None
-        ):
-            collected_content.append(resp.message.content)
-        elif (  # OpenAI non-streaming (Pydantic model), Anthropic new SDK non-streaming
-            hasattr(resp, "choices")
-            and resp.choices
-            and hasattr(resp.choices[0], "message")
-            and hasattr(resp.choices[0].message, "content")
-            and resp.choices[0].message.content is not None
-        ):
-            collected_content.append(resp.choices[0].message.content)  # OpenAI
-        elif (  # Anthropic new SDK non-streaming content block
-            hasattr(resp, "content")
-            and isinstance(resp.content, list)
-            and resp.content
-            and hasattr(resp.content[0], "text")
-        ):
-            collected_content.append(resp.content[0].text)  # Anthropic
-        else:
-            # Assume resp is a stream if not a recognized object
-            chunk = None
-            try:
-                for (
-                    chunk
-                ) in (
-                    resp
-                ):  # This will fail if resp is not iterable (e.g. a non-streaming response object)
-                    content_piece = ""
-                    # OpenAI-like stream
-                    if (
-                        hasattr(chunk, "choices")
-                        and len(chunk.choices) > 0
-                        and hasattr(chunk.choices[0], "delta")
-                        and hasattr(chunk.choices[0].delta, "content")
-                        and chunk.choices[0].delta.content is not None
-                    ):
-                        content_piece = chunk.choices[0].delta.content
-                    # Anthropic-like stream (ContentBlockDelta)
-                    elif (
-                        hasattr(chunk, "type")
-                        and chunk.type == "content_block_delta"
-                        and hasattr(chunk, "delta")
-                        and hasattr(chunk.delta, "text")
-                    ):
-                        content_piece = chunk.delta.text
-                    elif isinstance(chunk, str):  # Simplest case: stream of strings
-                        content_piece = chunk
-
-                    if content_piece:
-                        collected_content.append(content_piece)
-            except (
-                TypeError
-            ):  # If resp is not iterable (e.g. a final response object that wasn't caught above)
-                logger.debug(
-                    f"Response type {type(resp)} could not be iterated as a stream. It might be a non-streaming object not handled by specific checks."
-                )
-            except Exception as e:
-                logger.error(
-                    f"Error processing potential stream chunk: {e}, chunk was: {getattr(chunk, '__dict__', chunk) if chunk is not None else 'N/A'}"
-                )
-
-        return "".join(collected_content)
-
     def _gen_inner(
-        self, query: str, retriever: BaseRetriever, log_context: LogContext
+        self, query: str, log_context: LogContext
     ) -> Generator[Dict, None, None]:
-        # Reset state for this generation call
-        self.plan = ""
-        self.observations = []
-        retrieved_data = self._retriever_search(retriever, query, log_context)
+        """Execute ReAct reasoning loop with planning, action, and observation cycles"""
 
-        if self.user_api_key:
-            tools_dict = self._get_tools(self.user_api_key)
-        else:
-            tools_dict = self._get_user_tools(self.user)
+        self._reset_state()
+
+        tools_dict = (
+            self._get_tools(self.user_api_key)
+            if self.user_api_key
+            else self._get_user_tools(self.user)
+        )
         self._prepare_tools(tools_dict)
 
-        docs_together = "\n".join([doc["text"] for doc in retrieved_data])
-        iterating_reasoning = 0
-        while iterating_reasoning < MAX_ITERATIONS_REASONING:
-            iterating_reasoning += 1
-            # 1. Create Plan
-            logger.info("ReActAgent: Creating plan...")
-            plan_stream = self._create_plan(query, docs_together, log_context)
-            current_plan_parts = []
-            yield {"thought": f"Reasoning... (iteration {iterating_reasoning})\n\n"}
-            for line_chunk in plan_stream:
-                current_plan_parts.append(line_chunk)
-                yield {"thought": line_chunk}
-            self.plan = "".join(current_plan_parts)
-            if self.plan:
-                self.observations.append(
-                    f"Plan: {self.plan} Iteration: {iterating_reasoning}"
-                )
+        for iteration in range(1, MAX_ITERATIONS_REASONING + 1):
+            yield {"thought": f"Reasoning... (iteration {iteration})\n\n"}
 
-            max_obs_len = 20000
-            obs_str = "\n".join(self.observations)
-            if len(obs_str) > max_obs_len:
-                obs_str = obs_str[:max_obs_len] + "\n...[observations truncated]"
-            execution_prompt_str = (
-                (self.prompt or "")
-                + f"\n\nFollow this plan:\n{self.plan}"
-                + f"\n\nObservations:\n{obs_str}"
-                + f"\n\nIf there is enough data to complete user query '{query}', Respond with 'SATISFIED' only. Otherwise, continue. Dont Menstion 'SATISFIED' in your response if you are not ready. "
-            )
+            yield from self._planning_phase(query, log_context)
 
-            messages = self._build_messages(execution_prompt_str, query, retrieved_data)
-
-            resp_from_llm_gen = self._llm_gen(messages, log_context)
-
-            initial_llm_thought_content = self._extract_content_from_llm_response(
-                resp_from_llm_gen
-            )
-            if initial_llm_thought_content:
-                self.observations.append(
-                    f"Initial thought/response: {initial_llm_thought_content}"
-                )
-            else:
-                logger.info(
-                    "ReActAgent: Initial LLM response (before handler) had no textual content (might be only tool calls)."
-                )
-            resp_after_handler = self._llm_handler(
-                resp_from_llm_gen, tools_dict, messages, log_context
-            )
-
-            for (
-                tool_call_info
-            ) in (
-                self.tool_calls
-            ):  # Iterate over self.tool_calls populated by _llm_handler
-                observation_string = (
-                    f"Executed Action: Tool '{tool_call_info.get('tool_name', 'N/A')}' "
-                    f"with arguments '{tool_call_info.get('arguments', '{}')}'. Result: '{str(tool_call_info.get('result', ''))[:200]}...'"
-                )
-                self.observations.append(observation_string)
-
-            content_after_handler = self._extract_content_from_llm_response(
-                resp_after_handler
-            )
-            if content_after_handler:
-                self.observations.append(
-                    f"Response after tool execution: {content_after_handler}"
-                )
-            else:
-                logger.info(
-                    "ReActAgent: LLM response after handler had no textual content."
-                )
-
-            if log_context:
-                log_context.stacks.append(
-                    {
-                        "component": "agent_tool_calls",
-                        "data": {"tool_calls": self.tool_calls.copy()},
-                    }
-                )
-
-            yield {"sources": retrieved_data}
-
-            display_tool_calls = []
-            for tc in self.tool_calls:
-                cleaned_tc = tc.copy()
-                if len(str(cleaned_tc.get("result", ""))) > 50:
-                    cleaned_tc["result"] = str(cleaned_tc["result"])[:50] + "..."
-                display_tool_calls.append(cleaned_tc)
-            if display_tool_calls:
-                yield {"tool_calls": display_tool_calls}
-
-            if "SATISFIED" in content_after_handler:
-                logger.info(
-                    "ReActAgent: LLM satisfied with the plan and data. Stopping reasoning."
+            if not self.plan:
+                logger.warning(
+                    f"ReActAgent: No plan generated in iteration {iteration}"
                 )
                 break
+            self.observations.append(f"Plan (iteration {iteration}): {self.plan}")
 
-        # 3. Create Final Answer based on all observations
-        final_answer_stream = self._create_final_answer(
-            query, self.observations, log_context
-        )
-        for answer_chunk in final_answer_stream:
-            yield {"answer": answer_chunk}
-        logger.info("ReActAgent: Finished generating final answer.")
+            satisfied = yield from self._execution_phase(query, tools_dict, log_context)
 
-    def _create_plan(
-        self, query: str, docs_data: str, log_context: LogContext = None
-    ) -> Generator[str, None, None]:
-        plan_prompt_filled = planning_prompt_template.replace("{query}", query)
-        if "{summaries}" in plan_prompt_filled:
-            summaries = docs_data if docs_data else "No documents retrieved."
-            plan_prompt_filled = plan_prompt_filled.replace("{summaries}", summaries)
-        plan_prompt_filled = plan_prompt_filled.replace("{prompt}", self.prompt or "")
-        plan_prompt_filled = plan_prompt_filled.replace(
-            "{observations}", "\n".join(self.observations)
-        )
+            if satisfied:
+                logger.info("ReActAgent: Goal satisfied, stopping reasoning loop")
+                break
+        yield from self._synthesis_phase(query, log_context)
 
-        messages = [{"role": "user", "content": plan_prompt_filled}]
+    def _reset_state(self):
+        """Reset agent state for new query"""
+        self.plan = ""
+        self.observations = []
 
-        plan_stream_from_llm = self.llm.gen_stream(
-            model=self.gpt_model,
+    def _planning_phase(
+        self, query: str, log_context: LogContext
+    ) -> Generator[Dict, None, None]:
+        """Generate strategic plan for query"""
+        logger.info("ReActAgent: Creating plan...")
+
+        plan_prompt = self._build_planning_prompt(query)
+        messages = [{"role": "user", "content": plan_prompt}]
+
+        plan_stream = self.llm.gen_stream(
+            model=self.model_id,
             messages=messages,
-            tools=getattr(self, "tools", None),  # Use self.tools
+            tools=self.tools if self.tools else None,
         )
+
         if log_context:
-            data = build_stack_data(self.llm)
-            log_context.stacks.append({"component": "planning_llm", "data": data})
-
-        for chunk in plan_stream_from_llm:
-            content_piece = self._extract_content_from_llm_response(chunk)
-            if content_piece:
-                yield content_piece
-
-    def _create_final_answer(
-        self, query: str, observations: List[str], log_context: LogContext = None
-    ) -> Generator[str, None, None]:
-        observation_string = "\n".join(observations)
-        max_obs_len = 10000
-        if len(observation_string) > max_obs_len:
-            observation_string = (
-                observation_string[:max_obs_len] + "\n...[observations truncated]"
-            )
-            logger.warning(
-                "ReActAgent: Truncated observations for final answer prompt due to length."
+            log_context.stacks.append(
+                {"component": "planning_llm", "data": build_stack_data(self.llm)}
             )
+        plan_parts = []
+        for chunk in plan_stream:
+            content = self._extract_content(chunk)
+            if content:
+                plan_parts.append(content)
+                yield {"thought": content}
+        self.plan = "".join(plan_parts)
 
-        final_answer_prompt_filled = final_prompt_template.format(
-            query=query, observations=observation_string
+    def _execution_phase(
+        self, query: str, tools_dict: Dict, log_context: LogContext
+    ) -> Generator[bool, None, None]:
+        """Execute plan with tool calls and observations"""
+        execution_prompt = self._build_execution_prompt(query)
+        messages = self._build_messages(execution_prompt, query)
+
+        llm_response = self._llm_gen(messages, log_context)
+        initial_content = self._extract_content(llm_response)
+
+        if initial_content:
+            self.observations.append(f"Initial response: {initial_content}")
+        processed_response = self._llm_handler(
+            llm_response, tools_dict, messages, log_context
         )
 
-        messages = [{"role": "user", "content": final_answer_prompt_filled}]
-
-        # Final answer should synthesize, not call tools.
-        final_answer_stream_from_llm = self.llm.gen_stream(
-            model=self.gpt_model, messages=messages, tools=None
-        )
+        for tool_call in self.tool_calls:
+            observation = (
+                f"Executed: {tool_call.get('tool_name', 'Unknown')} "
+                f"with args {tool_call.get('arguments', {})}. "
+                f"Result: {str(tool_call.get('result', ''))[:200]}"
+            )
+            self.observations.append(observation)
+        final_content = self._extract_content(processed_response)
+        if final_content:
+            self.observations.append(f"Response after tools: {final_content}")
         if log_context:
-            data = build_stack_data(self.llm)
-            log_context.stacks.append({"component": "final_answer_llm", "data": data})
+            log_context.stacks.append(
+                {
+                    "component": "agent_tool_calls",
+                    "data": {"tool_calls": self.tool_calls.copy()},
+                }
+            )
+        yield {"sources": self.retrieved_docs}
+        yield {"tool_calls": self._get_truncated_tool_calls()}
 
-        for chunk in final_answer_stream_from_llm:
-            content_piece = self._extract_content_from_llm_response(chunk)
-            if content_piece:
-                yield content_piece
+        return "SATISFIED" in (final_content or "")
+
+    def _synthesis_phase(
+        self, query: str, log_context: LogContext
+    ) -> Generator[Dict, None, None]:
+        """Synthesize final answer from all observations"""
+        logger.info("ReActAgent: Generating final answer...")
+
+        final_prompt = self._build_final_answer_prompt(query)
+        messages = [{"role": "user", "content": final_prompt}]
+
+        final_stream = self.llm.gen_stream(
+            model=self.model_id, messages=messages, tools=None
+        )
+
+        if log_context:
+            log_context.stacks.append(
+                {"component": "final_answer_llm", "data": build_stack_data(self.llm)}
+            )
+        for chunk in final_stream:
+            content = self._extract_content(chunk)
+            if content:
+                yield {"answer": content}
+
+    def _build_planning_prompt(self, query: str) -> str:
+        """Build planning phase prompt"""
+        prompt = PLANNING_PROMPT_TEMPLATE.replace("{query}", query)
+        prompt = prompt.replace("{prompt}", self.prompt or "")
+        prompt = prompt.replace("{summaries}", "")
+        prompt = prompt.replace("{observations}", "\n".join(self.observations))
+        return prompt
+
+    def _build_execution_prompt(self, query: str) -> str:
+        """Build execution phase prompt with plan and observations"""
+        observations_str = "\n".join(self.observations)
+
+        if len(observations_str) > 20000:
+            observations_str = observations_str[:20000] + "\n...[truncated]"
+        return (
+            f"{self.prompt or ''}\n\n"
+            f"Follow this plan:\n{self.plan}\n\n"
+            f"Observations:\n{observations_str}\n\n"
+            f"If sufficient data exists to answer '{query}', respond with 'SATISFIED'. "
+            f"Otherwise, continue executing the plan."
+        )
+
+    def _build_final_answer_prompt(self, query: str) -> str:
+        """Build final synthesis prompt"""
+        observations_str = "\n".join(self.observations)
+
+        if len(observations_str) > 10000:
+            observations_str = observations_str[:10000] + "\n...[truncated]"
+            logger.warning("ReActAgent: Observations truncated for final answer")
+        return FINAL_PROMPT_TEMPLATE.format(query=query, observations=observations_str)
+
+    def _extract_content(self, response: Any) -> str:
+        """Extract text content from various LLM response formats"""
+        if not response:
+            return ""
+        collected = []
+
+        if isinstance(response, str):
+            return response
+        if hasattr(response, "message") and hasattr(response.message, "content"):
+            if response.message.content:
+                return response.message.content
+        if hasattr(response, "choices") and response.choices:
+            if hasattr(response.choices[0], "message"):
+                content = response.choices[0].message.content
+                if content:
+                    return content
+        if hasattr(response, "content") and isinstance(response.content, list):
+            if response.content and hasattr(response.content[0], "text"):
+                return response.content[0].text
+        try:
+            for chunk in response:
+                content_piece = ""
+
+                if hasattr(chunk, "choices") and chunk.choices:
+                    if hasattr(chunk.choices[0], "delta"):
+                        delta_content = chunk.choices[0].delta.content
+                        if delta_content:
+                            content_piece = delta_content
+                elif hasattr(chunk, "type") and chunk.type == "content_block_delta":
+                    if hasattr(chunk, "delta") and hasattr(chunk.delta, "text"):
+                        content_piece = chunk.delta.text
+                elif isinstance(chunk, str):
+                    content_piece = chunk
+                if content_piece:
+                    collected.append(content_piece)
+        except (TypeError, AttributeError):
+            logger.debug(
+                f"Response not iterable or unexpected format: {type(response)}"
+            )
+        except Exception as e:
+            logger.error(f"Error extracting content: {e}")
+        return "".join(collected)

application/agents/tools/todo_list.py

@@ -0,0 +1,321 @@
+from datetime import datetime
+from typing import Any, Dict, List, Optional
+import uuid
+
+from .base import Tool
+from application.core.mongo_db import MongoDB
+from application.core.settings import settings
+
+
+class TodoListTool(Tool):
+    """Todo List
+
+    Manages todo items for users. Supports creating, viewing, updating, and deleting todos.
+    """
+
+    def __init__(self, tool_config: Optional[Dict[str, Any]] = None, user_id: Optional[str] = None) -> None:
+        """Initialize the tool.
+
+        Args:
+            tool_config: Optional tool configuration. Should include:
+                - tool_id: Unique identifier for this todo list tool instance (from user_tools._id)
+                           This ensures each user's tool configuration has isolated todos
+            user_id: The authenticated user's id (should come from decoded_token["sub"]).
+        """
+        self.user_id: Optional[str] = user_id
+
+        # Get tool_id from configuration (passed from user_tools._id in production)
+        # In production, tool_id is the MongoDB ObjectId string from user_tools collection
+        if tool_config and "tool_id" in tool_config:
+            self.tool_id = tool_config["tool_id"]
+        elif user_id:
+            # Fallback for backward compatibility or testing
+            self.tool_id = f"default_{user_id}"
+        else:
+            # Last resort fallback (shouldn't happen in normal use)
+            self.tool_id = str(uuid.uuid4())
+
+        db = MongoDB.get_client()[settings.MONGO_DB_NAME]
+        self.collection = db["todos"]
+
+    # -----------------------------
+    # Action implementations
+    # -----------------------------
+    def execute_action(self, action_name: str, **kwargs: Any) -> str:
+        """Execute an action by name.
+
+        Args:
+            action_name: One of list, create, get, update, complete, delete.
+            **kwargs: Parameters for the action.
+
+        Returns:
+            A human-readable string result.
+        """
+        if not self.user_id:
+            return "Error: TodoListTool requires a valid user_id."
+
+        if action_name == "list":
+            return self._list()
+
+        if action_name == "create":
+            return self._create(kwargs.get("title", ""))
+
+        if action_name == "get":
+            return self._get(kwargs.get("todo_id"))
+
+        if action_name == "update":
+            return self._update(
+                kwargs.get("todo_id"),
+                kwargs.get("title", "")
+            )
+
+        if action_name == "complete":
+            return self._complete(kwargs.get("todo_id"))
+
+        if action_name == "delete":
+            return self._delete(kwargs.get("todo_id"))
+
+        return f"Unknown action: {action_name}"
+
+    def get_actions_metadata(self) -> List[Dict[str, Any]]:
+        """Return JSON metadata describing supported actions for tool schemas."""
+        return [
+            {
+                "name": "list",
+                "description": "List all todos for the user.",
+                "parameters": {"type": "object", "properties": {}},
+            },
+            {
+                "name": "create",
+                "description": "Create a new todo item.",
+                "parameters": {
+                    "type": "object",
+                    "properties": {
+                        "title": {
+                            "type": "string",
+                            "description": "Title of the todo item."
+                        }
+                    },
+                    "required": ["title"],
+                },
+            },
+            {
+                "name": "get",
+                "description": "Get a specific todo by ID.",
+                "parameters": {
+                    "type": "object",
+                    "properties": {
+                        "todo_id": {
+                            "type": "integer",
+                            "description": "The ID of the todo to retrieve."
+                        }
+                    },
+                    "required": ["todo_id"],
+                },
+            },
+            {
+                "name": "update",
+                "description": "Update a todo's title by ID.",
+                "parameters": {
+                    "type": "object",
+                    "properties": {
+                        "todo_id": {
+                            "type": "integer",
+                            "description": "The ID of the todo to update."
+                        },
+                        "title": {
+                            "type": "string",
+                            "description": "The new title for the todo."
+                        }
+                    },
+                    "required": ["todo_id", "title"],
+                },
+            },
+            {
+                "name": "complete",
+                "description": "Mark a todo as completed.",
+                "parameters": {
+                    "type": "object",
+                    "properties": {
+                        "todo_id": {
+                            "type": "integer",
+                            "description": "The ID of the todo to mark as completed."
+                        }
+                    },
+                    "required": ["todo_id"],
+                },
+            },
+            {
+                "name": "delete",
+                "description": "Delete a specific todo by ID.",
+                "parameters": {
+                    "type": "object",
+                    "properties": {
+                        "todo_id": {
+                            "type": "integer",
+                            "description": "The ID of the todo to delete."
+                        }
+                    },
+                    "required": ["todo_id"],
+                },
+            },
+        ]
+
+    def get_config_requirements(self) -> Dict[str, Any]:
+        """Return configuration requirements."""
+        return {}
+
+    # -----------------------------
+    # Internal helpers
+    # -----------------------------
+    def _coerce_todo_id(self, value: Optional[Any]) -> Optional[int]:
+        """Convert todo identifiers to sequential integers."""
+        if value is None:
+            return None
+
+        if isinstance(value, int):
+            return value if value > 0 else None
+
+        if isinstance(value, str):
+            stripped = value.strip()
+            if stripped.isdigit():
+                numeric_value = int(stripped)
+                return numeric_value if numeric_value > 0 else None
+
+        return None
+
+    def _get_next_todo_id(self) -> int:
+        """Get the next sequential todo_id for this user and tool.
+
+        Returns a simple integer (1, 2, 3, ...) scoped to this user/tool.
+        With 5-10 todos max, scanning is negligible.
+        """
+        # Find all todos for this user/tool and get their IDs
+        todos = list(self.collection.find(
+            {"user_id": self.user_id, "tool_id": self.tool_id},
+            {"todo_id": 1}
+        ))
+
+        # Find the maximum todo_id
+        max_id = 0
+        for todo in todos:
+            todo_id = self._coerce_todo_id(todo.get("todo_id"))
+            if todo_id is not None:
+                max_id = max(max_id, todo_id)
+
+        return max_id + 1
+
+    def _list(self) -> str:
+        """List all todos for the user."""
+        cursor = self.collection.find({"user_id": self.user_id, "tool_id": self.tool_id})
+        todos = list(cursor)
+
+        if not todos:
+            return "No todos found."
+
+        result_lines = ["Todos:"]
+        for doc in todos:
+            todo_id = doc.get("todo_id")
+            title = doc.get("title", "Untitled")
+            status = doc.get("status", "open")
+
+            line = f"[{todo_id}] {title} ({status})"
+            result_lines.append(line)
+
+        return "\n".join(result_lines)
+
+    def _create(self, title: str) -> str:
+        """Create a new todo item."""
+        title = (title or "").strip()
+        if not title:
+            return "Error: Title is required."
+
+        now = datetime.now()
+        todo_id = self._get_next_todo_id()
+
+        doc = {
+            "todo_id": todo_id,
+            "user_id": self.user_id,
+            "tool_id": self.tool_id,
+            "title": title,
+            "status": "open",
+            "created_at": now,
+            "updated_at": now,
+        }
+        self.collection.insert_one(doc)
+        return f"Todo created with ID {todo_id}: {title}"
+
+    def _get(self, todo_id: Optional[Any]) -> str:
+        """Get a specific todo by ID."""
+        parsed_todo_id = self._coerce_todo_id(todo_id)
+        if parsed_todo_id is None:
+            return "Error: todo_id must be a positive integer."
+
+        doc = self.collection.find_one({
+            "user_id": self.user_id,
+            "tool_id": self.tool_id,
+            "todo_id": parsed_todo_id
+        })
+
+        if not doc:
+            return f"Error: Todo with ID {parsed_todo_id} not found."
+
+        title = doc.get("title", "Untitled")
+        status = doc.get("status", "open")
+
+        result = f"Todo [{parsed_todo_id}]:\nTitle: {title}\nStatus: {status}"
+
+        return result
+
+    def _update(self, todo_id: Optional[Any], title: str) -> str:
+        """Update a todo's title by ID."""
+        parsed_todo_id = self._coerce_todo_id(todo_id)
+        if parsed_todo_id is None:
+            return "Error: todo_id must be a positive integer."
+
+        title = (title or "").strip()
+        if not title:
+            return "Error: Title is required."
+
+        result = self.collection.update_one(
+            {"user_id": self.user_id, "tool_id": self.tool_id, "todo_id": parsed_todo_id},
+            {"$set": {"title": title, "updated_at": datetime.now()}}
+        )
+
+        if result.matched_count == 0:
+            return f"Error: Todo with ID {parsed_todo_id} not found."
+
+        return f"Todo {parsed_todo_id} updated to: {title}"
+
+    def _complete(self, todo_id: Optional[Any]) -> str:
+        """Mark a todo as completed."""
+        parsed_todo_id = self._coerce_todo_id(todo_id)
+        if parsed_todo_id is None:
+            return "Error: todo_id must be a positive integer."
+
+        result = self.collection.update_one(
+            {"user_id": self.user_id, "tool_id": self.tool_id, "todo_id": parsed_todo_id},
+            {"$set": {"status": "completed", "updated_at": datetime.now()}}
+        )
+
+        if result.matched_count == 0:
+            return f"Error: Todo with ID {parsed_todo_id} not found."
+
+        return f"Todo {parsed_todo_id} marked as completed."
+
+    def _delete(self, todo_id: Optional[Any]) -> str:
+        """Delete a specific todo by ID."""
+        parsed_todo_id = self._coerce_todo_id(todo_id)
+        if parsed_todo_id is None:
+            return "Error: todo_id must be a positive integer."
+
+        result = self.collection.delete_one({
+            "user_id": self.user_id,
+            "tool_id": self.tool_id,
+            "todo_id": parsed_todo_id
+        })
+
+        if result.deleted_count == 0:
+            return f"Error: Todo with ID {parsed_todo_id} not found."
+
+        return f"Todo {parsed_todo_id} deleted."

application/agents/tools/tool_manager.py

@@ -28,7 +28,7 @@ class ToolManager:
         module = importlib.import_module(f"application.agents.tools.{tool_name}")
         for member_name, obj in inspect.getmembers(module, inspect.isclass):
             if issubclass(obj, Tool) and obj is not Tool:
-                if tool_name in {"mcp_tool", "notes", "memory"} and user_id:
+                if tool_name in {"mcp_tool", "notes", "memory", "todo_list"} and user_id:
                     return obj(tool_config, user_id)
                 else:
                     return obj(tool_config)
@@ -36,7 +36,7 @@ class ToolManager:
     def execute_action(self, tool_name, action_name, user_id=None, **kwargs):
         if tool_name not in self.tools:
             raise ValueError(f"Tool '{tool_name}' not loaded")
-        if tool_name in {"mcp_tool", "memory"} and user_id:
+        if tool_name in {"mcp_tool", "memory", "todo_list"} and user_id:
             tool_config = self.config.get(tool_name, {})
             tool = self.load_tool(tool_name, tool_config, user_id)
             return tool.execute_action(action_name, **kwargs)

application/api/answer/routes/answer.py

@@ -54,6 +54,14 @@ class AnswerResource(Resource, BaseAnswerResource):
                 default=True,
                 description="Whether to save the conversation",
             ),
+            "model_id": fields.String(
+                required=False,
+                description="Model ID to use for this request",
+            ),
+            "passthrough": fields.Raw(
+                required=False,
+                description="Dynamic parameters to inject into prompt template",
+            ),
         },
     )
 
@@ -69,8 +77,17 @@ class AnswerResource(Resource, BaseAnswerResource):
             processor.initialize()
             if not processor.decoded_token:
                 return make_response({"error": "Unauthorized"}, 401)
-            agent = processor.create_agent()
-            retriever = processor.create_retriever()
+
+            docs_together, docs_list = processor.pre_fetch_docs(
+                data.get("question", "")
+            )
+            tools_data = processor.pre_fetch_tools()
+
+            agent = processor.create_agent(
+                docs_together=docs_together,
+                docs=docs_list,
+                tools_data=tools_data,
+            )
 
             if error := self.check_usage(processor.agent_config):
                 return error
@@ -78,13 +95,13 @@ class AnswerResource(Resource, BaseAnswerResource):
             stream = self.complete_stream(
                 question=data["question"],
                 agent=agent,
-                retriever=retriever,
                 conversation_id=processor.conversation_id,
                 user_api_key=processor.agent_config.get("user_api_key"),
                 decoded_token=processor.decoded_token,
                 isNoneDoc=data.get("isNoneDoc"),
                 index=None,
                 should_save_conversation=data.get("save_conversation", True),
+                model_id=processor.model_id,
             )
             stream_result = self.process_response_stream(stream)
 

application/api/answer/routes/base.py

@@ -3,15 +3,20 @@ import json
 import logging
 from typing import Any, Dict, Generator, List, Optional
 
-from flask import Response, make_response, jsonify
+from flask import jsonify, make_response, Response
 from flask_restx import Namespace
 
 from application.api.answer.services.conversation_service import ConversationService
+from application.core.model_utils import (
+    get_api_key_for_provider,
+    get_default_model_id,
+    get_provider_from_model_id,
+)
 
 from application.core.mongo_db import MongoDB
 from application.core.settings import settings
 from application.llm.llm_creator import LLMCreator
-from application.utils import check_required_fields, get_gpt_model
+from application.utils import check_required_fields
 
 logger = logging.getLogger(__name__)
 
@@ -27,7 +32,7 @@ class BaseAnswerResource:
         db = mongo[settings.MONGO_DB_NAME]
         self.db = db
         self.user_logs_collection = db["user_logs"]
-        self.gpt_model = get_gpt_model()
+        self.default_model_id = get_default_model_id()
         self.conversation_service = ConversationService()
 
     def validate_request(
@@ -41,9 +46,7 @@ class BaseAnswerResource:
             return missing_fields
         return None
 
-    def check_usage(
-            self, agent_config: Dict
-    ) -> Optional[Response]:
+    def check_usage(self, agent_config: Dict) -> Optional[Response]:
         """Check if there is a usage limit and if it is exceeded
 
         Args:
@@ -51,30 +54,38 @@ class BaseAnswerResource:
 
         Returns:
             None or Response if either of limits exceeded.
-        
+
         """
         api_key = agent_config.get("user_api_key")
         if not api_key:
             return None
-        
         agents_collection = self.db["agents"]
         agent = agents_collection.find_one({"key": api_key})
 
         if not agent:
             return make_response(
-                jsonify(
-                    {
-                        "success": False,
-                        "message": "Invalid API key."
-                    }
-                ),
-                401
+                jsonify({"success": False, "message": "Invalid API key."}), 401
             )
+        limited_token_mode_raw = agent.get("limited_token_mode", False)
+        limited_request_mode_raw = agent.get("limited_request_mode", False)
 
-        limited_token_mode = agent.get("limited_token_mode", False)
-        limited_request_mode = agent.get("limited_request_mode", False)
-        token_limit = int(agent.get("token_limit", settings.DEFAULT_AGENT_LIMITS["token_limit"]))
-        request_limit = int(agent.get("request_limit", settings.DEFAULT_AGENT_LIMITS["request_limit"]))
+        limited_token_mode = (
+            limited_token_mode_raw
+            if isinstance(limited_token_mode_raw, bool)
+            else limited_token_mode_raw == "True"
+        )
+        limited_request_mode = (
+            limited_request_mode_raw
+            if isinstance(limited_request_mode_raw, bool)
+            else limited_request_mode_raw == "True"
+        )
+
+        token_limit = int(
+            agent.get("token_limit", settings.DEFAULT_AGENT_LIMITS["token_limit"])
+        )
+        request_limit = int(
+            agent.get("request_limit", settings.DEFAULT_AGENT_LIMITS["request_limit"])
+        )
 
         token_usage_collection = self.db["token_usage"]
 
@@ -83,51 +94,56 @@ class BaseAnswerResource:
 
         match_query = {
             "timestamp": {"$gte": start_date, "$lte": end_date},
-            "api_key": api_key
+            "api_key": api_key,
         }
-        
+
         if limited_token_mode:
             token_pipeline = [
                 {"$match": match_query},
                 {
                     "$group": {
                         "_id": None,
-                        "total_tokens": {"$sum": {"$add": ["$prompt_tokens", "$generated_tokens"]}}
+                        "total_tokens": {
+                            "$sum": {"$add": ["$prompt_tokens", "$generated_tokens"]}
+                        },
                     }
-                }
+                },
             ]
             token_result = list(token_usage_collection.aggregate(token_pipeline))
             daily_token_usage = token_result[0]["total_tokens"] if token_result else 0
         else:
             daily_token_usage = 0
-
         if limited_request_mode:
             daily_request_usage = token_usage_collection.count_documents(match_query)
         else:
             daily_request_usage = 0
-
         if not limited_token_mode and not limited_request_mode:
             return None
-        elif limited_token_mode and token_limit > daily_token_usage:
-            return None
-        elif limited_request_mode and request_limit > daily_request_usage:
-            return None
-
-        return make_response(
-            jsonify(
-                {
-                    "success": False,
-                    "message": "Exceeding usage limit, please try again later."
-                }
-            ),
-            429, # too many requests
+        token_exceeded = (
+            limited_token_mode and token_limit > 0 and daily_token_usage >= token_limit
         )
+        request_exceeded = (
+            limited_request_mode
+            and request_limit > 0
+            and daily_request_usage >= request_limit
+        )
+
+        if token_exceeded or request_exceeded:
+            return make_response(
+                jsonify(
+                    {
+                        "success": False,
+                        "message": "Exceeding usage limit, please try again later.",
+                    }
+                ),
+                429,
+            )
+        return None
 
     def complete_stream(
         self,
         question: str,
         agent: Any,
-        retriever: Any,
         conversation_id: Optional[str],
         user_api_key: Optional[str],
         decoded_token: Dict[str, Any],
@@ -138,6 +154,7 @@ class BaseAnswerResource:
         agent_id: Optional[str] = None,
         is_shared_usage: bool = False,
         shared_token: Optional[str] = None,
+        model_id: Optional[str] = None,
     ) -> Generator[str, None, None]:
         """
         Generator function that streams the complete conversation response.
@@ -156,6 +173,8 @@ class BaseAnswerResource:
             agent_id: ID of agent used
             is_shared_usage: Flag for shared agent usage
             shared_token: Token for shared agent
+            model_id: Model ID used for the request
+            retrieved_docs: Pre-fetched documents for sources (optional)
 
         Yields:
             Server-sent event strings
@@ -166,7 +185,7 @@ class BaseAnswerResource:
             schema_info = None
             structured_chunks = []
 
-            for line in agent.gen(query=question, retriever=retriever):
+            for line in agent.gen(query=question):
                 if "answer" in line:
                     response_full += str(line["answer"])
                     if line.get("structured"):
@@ -202,7 +221,6 @@ class BaseAnswerResource:
                 elif "type" in line:
                     data = json.dumps(line)
                     yield f"data: {data}\n\n"
-
             if is_structured and structured_chunks:
                 structured_data = {
                     "type": "structured_answer",
@@ -212,15 +230,22 @@ class BaseAnswerResource:
                 }
                 data = json.dumps(structured_data)
                 yield f"data: {data}\n\n"
-
             if isNoneDoc:
                 for doc in source_log_docs:
                     doc["source"] = "None"
+            provider = (
+                get_provider_from_model_id(model_id)
+                if model_id
+                else settings.LLM_PROVIDER
+            )
+            system_api_key = get_api_key_for_provider(provider or settings.LLM_PROVIDER)
+
             llm = LLMCreator.create_llm(
-                settings.LLM_PROVIDER,
-                api_key=settings.API_KEY,
+                provider or settings.LLM_PROVIDER,
+                api_key=system_api_key,
                 user_api_key=user_api_key,
                 decoded_token=decoded_token,
+                model_id=model_id,
             )
 
             if should_save_conversation:
@@ -232,7 +257,7 @@ class BaseAnswerResource:
                     source_log_docs,
                     tool_calls,
                     llm,
-                    self.gpt_model,
+                    model_id or self.default_model_id,
                     decoded_token,
                     index=index,
                     api_key=user_api_key,
@@ -247,7 +272,6 @@ class BaseAnswerResource:
             data = json.dumps(id_data)
             yield f"data: {data}\n\n"
 
-            retriever_params = retriever.get_params()
             log_data = {
                 "action": "stream_answer",
                 "level": "info",
@@ -256,7 +280,6 @@ class BaseAnswerResource:
                 "question": question,
                 "response": response_full,
                 "sources": source_log_docs,
-                "retriever_params": retriever_params,
                 "attachments": attachment_ids,
                 "timestamp": datetime.datetime.now(datetime.timezone.utc),
             }
@@ -264,24 +287,19 @@ class BaseAnswerResource:
                 log_data["structured_output"] = True
                 if schema_info:
                     log_data["schema"] = schema_info
-  
-            # clean up text fields to be no longer than 10000 characters
+            # Clean up text fields to be no longer than 10000 characters
+
             for key, value in log_data.items():
                 if isinstance(value, str) and len(value) > 10000:
                     log_data[key] = value[:10000]
-            
             self.user_logs_collection.insert_one(log_data)
 
-            # End of stream
-
             data = json.dumps({"type": "end"})
             yield f"data: {data}\n\n"
         except GeneratorExit:
-            # Client aborted the connection
-            logger.info(
-                f"Stream aborted by client for question: {question[:50]}... "
-            )
-            # Save partial response to database before exiting
+            logger.info(f"Stream aborted by client for question: {question[:50]}... ")
+            # Save partial response
+
             if should_save_conversation and response_full:
                 try:
                     if isNoneDoc:
@@ -301,7 +319,7 @@ class BaseAnswerResource:
                         source_log_docs,
                         tool_calls,
                         llm,
-                        self.gpt_model,
+                        model_id or self.default_model_id,
                         decoded_token,
                         index=index,
                         api_key=user_api_key,
@@ -311,7 +329,9 @@ class BaseAnswerResource:
                         attachment_ids=attachment_ids,
                     )
                 except Exception as e:
-                    logger.error(f"Error saving partial response: {str(e)}", exc_info=True)
+                    logger.error(
+                        f"Error saving partial response: {str(e)}", exc_info=True
+                    )
             raise
         except Exception as e:
             logger.error(f"Error in stream: {str(e)}", exc_info=True)
@@ -356,7 +376,7 @@ class BaseAnswerResource:
                     thought = event["thought"]
                 elif event["type"] == "error":
                     logger.error(f"Error from stream: {event['error']}")
-                    return None, None, None, None, event["error"]
+                    return None, None, None, None, event["error"], None
                 elif event["type"] == "end":
                     stream_ended = True
             except (json.JSONDecodeError, KeyError) as e:
@@ -364,8 +384,7 @@ class BaseAnswerResource:
                 continue
         if not stream_ended:
             logger.error("Stream ended unexpectedly without an 'end' event.")
-            return None, None, None, None, "Stream ended unexpectedly"
-
+            return None, None, None, None, "Stream ended unexpectedly", None
         result = (
             conversation_id,
             response_full,
@@ -377,7 +396,6 @@ class BaseAnswerResource:
 
         if is_structured:
             result = result + ({"structured": True, "schema": schema_info},)
-
         return result
 
     def error_stream_generate(self, err_response):

application/api/answer/routes/stream.py

@@ -57,9 +57,17 @@ class StreamResource(Resource, BaseAnswerResource):
                 default=True,
                 description="Whether to save the conversation",
             ),
+            "model_id": fields.String(
+                required=False,
+                description="Model ID to use for this request",
+            ),
             "attachments": fields.List(
                 fields.String, required=False, description="List of attachment IDs"
             ),
+            "passthrough": fields.Raw(
+                required=False,
+                description="Dynamic parameters to inject into prompt template",
+            ),
         },
     )
 
@@ -73,17 +81,20 @@ class StreamResource(Resource, BaseAnswerResource):
         processor = StreamProcessor(data, decoded_token)
         try:
             processor.initialize()
-            agent = processor.create_agent()
-            retriever = processor.create_retriever()
+
+            docs_together, docs_list = processor.pre_fetch_docs(data["question"])
+            tools_data = processor.pre_fetch_tools()
+
+            agent = processor.create_agent(
+                docs_together=docs_together, docs=docs_list, tools_data=tools_data
+            )
 
             if error := self.check_usage(processor.agent_config):
                 return error
-
             return Response(
                 self.complete_stream(
                     question=data["question"],
                     agent=agent,
-                    retriever=retriever,
                     conversation_id=processor.conversation_id,
                     user_api_key=processor.agent_config.get("user_api_key"),
                     decoded_token=processor.decoded_token,
@@ -94,6 +105,7 @@ class StreamResource(Resource, BaseAnswerResource):
                     agent_id=data.get("agent_id"),
                     is_shared_usage=processor.is_shared_usage,
                     shared_token=processor.shared_token,
+                    model_id=processor.model_id,
                 ),
                 mimetype="text/event-stream",
             )

application/api/answer/services/conversation_service.py

@@ -52,7 +52,7 @@ class ConversationService:
         sources: List[Dict[str, Any]],
         tool_calls: List[Dict[str, Any]],
         llm: Any,
-        gpt_model: str,
+        model_id: str,
         decoded_token: Dict[str, Any],
         index: Optional[int] = None,
         api_key: Optional[str] = None,
@@ -66,7 +66,7 @@ class ConversationService:
         if not user_id:
             raise ValueError("User ID not found in token")
         current_time = datetime.now(timezone.utc)
-        
+
         # clean up in sources array such that we save max 1k characters for text part
         for source in sources:
             if "text" in source and isinstance(source["text"], str):
@@ -90,6 +90,7 @@ class ConversationService:
                         f"queries.{index}.tool_calls": tool_calls,
                         f"queries.{index}.timestamp": current_time,
                         f"queries.{index}.attachments": attachment_ids,
+                        f"queries.{index}.model_id": model_id,
                     }
                 },
             )
@@ -120,6 +121,7 @@ class ConversationService:
                             "tool_calls": tool_calls,
                             "timestamp": current_time,
                             "attachments": attachment_ids,
+                            "model_id": model_id,
                         }
                     }
                 },
@@ -133,10 +135,9 @@ class ConversationService:
 
             messages_summary = [
                 {
-                    "role": "assistant",
-                    "content": "Summarise following conversation in no more than 3 "
-                    "words, respond ONLY with the summary, use the same "
-                    "language as the user query",
+                    "role": "system",
+                    "content": "You are a helpful assistant that creates concise conversation titles. "
+                    "Summarize conversations in 3 words or less using the same language as the user.",
                 },
                 {
                     "role": "user",
@@ -147,7 +148,7 @@ class ConversationService:
             ]
 
             completion = llm.gen(
-                model=gpt_model, messages=messages_summary, max_tokens=30
+                model=model_id, messages=messages_summary, max_tokens=30
             )
 
             conversation_data = {
@@ -163,6 +164,7 @@ class ConversationService:
                         "tool_calls": tool_calls,
                         "timestamp": current_time,
                         "attachments": attachment_ids,
+                        "model_id": model_id,
                     }
                 ],
             }

application/api/answer/services/prompt_renderer.py

@@ -0,0 +1,97 @@
+import logging
+from typing import Any, Dict, Optional
+
+from application.templates.namespaces import NamespaceManager
+
+from application.templates.template_engine import TemplateEngine, TemplateRenderError
+
+logger = logging.getLogger(__name__)
+
+
+class PromptRenderer:
+    """Service for rendering prompts with dynamic context using namespaces"""
+
+    def __init__(self):
+        self.template_engine = TemplateEngine()
+        self.namespace_manager = NamespaceManager()
+
+    def render_prompt(
+        self,
+        prompt_content: str,
+        user_id: Optional[str] = None,
+        request_id: Optional[str] = None,
+        passthrough_data: Optional[Dict[str, Any]] = None,
+        docs: Optional[list] = None,
+        docs_together: Optional[str] = None,
+        tools_data: Optional[Dict[str, Any]] = None,
+        **kwargs,
+    ) -> str:
+        """
+        Render prompt with full context from all namespaces.
+
+        Args:
+            prompt_content: Raw prompt template string
+            user_id: Current user identifier
+            request_id: Unique request identifier
+            passthrough_data: Parameters from web request
+            docs: RAG retrieved documents
+            docs_together: Concatenated document content
+            tools_data: Pre-fetched tool results organized by tool name
+            **kwargs: Additional parameters for namespace builders
+
+        Returns:
+            Rendered prompt string with all variables substituted
+
+        Raises:
+            TemplateRenderError: If template rendering fails
+        """
+        if not prompt_content:
+            return ""
+
+        uses_template = self._uses_template_syntax(prompt_content)
+
+        if not uses_template:
+            return self._apply_legacy_substitutions(prompt_content, docs_together)
+
+        try:
+            context = self.namespace_manager.build_context(
+                user_id=user_id,
+                request_id=request_id,
+                passthrough_data=passthrough_data,
+                docs=docs,
+                docs_together=docs_together,
+                tools_data=tools_data,
+                **kwargs,
+            )
+
+            return self.template_engine.render(prompt_content, context)
+        except TemplateRenderError:
+            raise
+        except Exception as e:
+            error_msg = f"Prompt rendering failed: {str(e)}"
+            logger.error(error_msg)
+            raise TemplateRenderError(error_msg) from e
+
+    def _uses_template_syntax(self, prompt_content: str) -> bool:
+        """Check if prompt uses Jinja2 template syntax"""
+        return "{{" in prompt_content and "}}" in prompt_content
+
+    def _apply_legacy_substitutions(
+        self, prompt_content: str, docs_together: Optional[str] = None
+    ) -> str:
+        """
+        Apply backward-compatible substitutions for old prompt format.
+
+        Handles legacy {summaries} and {query} placeholders during transition period.
+        """
+        if docs_together:
+            prompt_content = prompt_content.replace("{summaries}", docs_together)
+        return prompt_content
+
+    def validate_template(self, prompt_content: str) -> bool:
+        """Validate prompt template syntax"""
+        return self.template_engine.validate_template(prompt_content)
+
+    def extract_variables(self, prompt_content: str) -> set[str]:
+        """Extract all variable names from prompt template"""
+        return self.template_engine.extract_variables(prompt_content)

application/api/answer/services/stream_processor.py

@@ -3,7 +3,7 @@ import json
 import logging
 import os
 from pathlib import Path
-from typing import Any, Dict, Optional
+from typing import Any, Dict, Optional, Set
 
 from bson.dbref import DBRef
 
@@ -11,10 +11,20 @@ from bson.objectid import ObjectId
 
 from application.agents.agent_creator import AgentCreator
 from application.api.answer.services.conversation_service import ConversationService
+from application.api.answer.services.prompt_renderer import PromptRenderer
+from application.core.model_utils import (
+    get_api_key_for_provider,
+    get_default_model_id,
+    get_provider_from_model_id,
+    validate_model_id,
+)
 from application.core.mongo_db import MongoDB
 from application.core.settings import settings
 from application.retriever.retriever_creator import RetrieverCreator
-from application.utils import get_gpt_model, limit_chat_history
+from application.utils import (
+    calculate_doc_token_budget,
+    limit_chat_history,
+)
 
 logger = logging.getLogger(__name__)
 
@@ -73,15 +83,20 @@ class StreamProcessor:
         self.all_sources = []
         self.attachments = []
         self.history = []
+        self.retrieved_docs = []
         self.agent_config = {}
         self.retriever_config = {}
         self.is_shared_usage = False
         self.shared_token = None
-        self.gpt_model = get_gpt_model()
+        self.model_id: Optional[str] = None
         self.conversation_service = ConversationService()
+        self.prompt_renderer = PromptRenderer()
+        self._prompt_content: Optional[str] = None
+        self._required_tool_actions: Optional[Dict[str, Set[Optional[str]]]] = None
 
     def initialize(self):
         """Initialize all required components for processing"""
+        self._validate_and_set_model()
         self._configure_agent()
         self._configure_source()
         self._configure_retriever()
@@ -103,7 +118,7 @@ class StreamProcessor:
             ]
         else:
             self.history = limit_chat_history(
-                json.loads(self.data.get("history", "[]")), gpt_model=self.gpt_model
+                json.loads(self.data.get("history", "[]")), model_id=self.model_id
             )
 
     def _process_attachments(self):
@@ -134,6 +149,25 @@ class StreamProcessor:
                 )
         return attachments
 
+    def _validate_and_set_model(self):
+        """Validate and set model_id from request"""
+        from application.core.model_settings import ModelRegistry
+
+        requested_model = self.data.get("model_id")
+
+        if requested_model:
+            if not validate_model_id(requested_model):
+                registry = ModelRegistry.get_instance()
+                available_models = [m.id for m in registry.get_enabled_models()]
+                raise ValueError(
+                    f"Invalid model_id '{requested_model}'. "
+                    f"Available models: {', '.join(available_models[:5])}"
+                    + (f" and {len(available_models) - 5} more" if len(available_models) > 5 else "")
+                )
+            self.model_id = requested_model
+        else:
+            self.model_id = get_default_model_id()
+
     def _get_agent_key(self, agent_id: Optional[str], user_id: Optional[str]) -> tuple:
         """Get API key for agent with access control"""
         if not agent_id:
@@ -311,43 +345,330 @@ class StreamProcessor:
             )
 
     def _configure_retriever(self):
-        """Configure the retriever based on request data"""
+        history_token_limit = int(self.data.get("token_limit", 2000))
+        doc_token_limit = calculate_doc_token_budget(
+            model_id=self.model_id, history_token_limit=history_token_limit
+        )
+
         self.retriever_config = {
             "retriever_name": self.data.get("retriever", "classic"),
             "chunks": int(self.data.get("chunks", 2)),
-            "token_limit": self.data.get("token_limit", settings.DEFAULT_MAX_HISTORY),
+            "doc_token_limit": doc_token_limit,
+            "history_token_limit": history_token_limit,
         }
 
         api_key = self.data.get("api_key") or self.agent_key
         if not api_key and "isNoneDoc" in self.data and self.data["isNoneDoc"]:
             self.retriever_config["chunks"] = 0
 
-    def create_agent(self):
-        """Create and return the configured agent"""
-        return AgentCreator.create_agent(
-            self.agent_config["agent_type"],
-            endpoint="stream",
-            llm_name=settings.LLM_PROVIDER,
-            gpt_model=self.gpt_model,
-            api_key=settings.API_KEY,
-            user_api_key=self.agent_config["user_api_key"],
-            prompt=get_prompt(self.agent_config["prompt_id"], self.prompts_collection),
-            chat_history=self.history,
-            decoded_token=self.decoded_token,
-            attachments=self.attachments,
-            json_schema=self.agent_config.get("json_schema"),
-        )
-
     def create_retriever(self):
-        """Create and return the configured retriever"""
         return RetrieverCreator.create_retriever(
             self.retriever_config["retriever_name"],
             source=self.source,
             chat_history=self.history,
             prompt=get_prompt(self.agent_config["prompt_id"], self.prompts_collection),
             chunks=self.retriever_config["chunks"],
-            token_limit=self.retriever_config["token_limit"],
-            gpt_model=self.gpt_model,
+            doc_token_limit=self.retriever_config.get("doc_token_limit", 50000),
+            model_id=self.model_id,
             user_api_key=self.agent_config["user_api_key"],
             decoded_token=self.decoded_token,
         )
+
+    def pre_fetch_docs(self, question: str) -> tuple[Optional[str], Optional[list]]:
+        """Pre-fetch documents for template rendering before agent creation"""
+        if self.data.get("isNoneDoc", False):
+            logger.info("Pre-fetch skipped: isNoneDoc=True")
+            return None, None
+        try:
+            retriever = self.create_retriever()
+            logger.info(
+                f"Pre-fetching docs with chunks={retriever.chunks}, doc_token_limit={retriever.doc_token_limit}"
+            )
+            docs = retriever.search(question)
+            logger.info(f"Pre-fetch retrieved {len(docs) if docs else 0} documents")
+
+            if not docs:
+                logger.info("Pre-fetch: No documents returned from search")
+                return None, None
+            self.retrieved_docs = docs
+
+            docs_with_filenames = []
+            for doc in docs:
+                filename = doc.get("filename") or doc.get("title") or doc.get("source")
+                if filename:
+                    chunk_header = str(filename)
+                    docs_with_filenames.append(f"{chunk_header}\n{doc['text']}")
+                else:
+                    docs_with_filenames.append(doc["text"])
+            docs_together = "\n\n".join(docs_with_filenames)
+
+            logger.info(f"Pre-fetch docs_together size: {len(docs_together)} chars")
+
+            return docs_together, docs
+        except Exception as e:
+            logger.error(f"Failed to pre-fetch docs: {str(e)}", exc_info=True)
+            return None, None
+
+    def pre_fetch_tools(self) -> Optional[Dict[str, Any]]:
+        """Pre-fetch tool data for template rendering before agent creation
+
+        Can be controlled via:
+        1. Global setting: ENABLE_TOOL_PREFETCH in .env
+        2. Per-request: disable_tool_prefetch in request data
+        """
+        if not settings.ENABLE_TOOL_PREFETCH:
+            logger.info(
+                "Tool pre-fetching disabled globally via ENABLE_TOOL_PREFETCH setting"
+            )
+            return None
+
+        if self.data.get("disable_tool_prefetch", False):
+            logger.info("Tool pre-fetching disabled for this request")
+            return None
+
+        required_tool_actions = self._get_required_tool_actions()
+        filtering_enabled = required_tool_actions is not None
+
+        try:
+            user_tools_collection = self.db["user_tools"]
+            user_id = self.initial_user_id or "local"
+
+            user_tools = list(
+                user_tools_collection.find({"user": user_id, "status": True})
+            )
+
+            if not user_tools:
+                return None
+
+            tools_data = {}
+
+            for tool_doc in user_tools:
+                tool_name = tool_doc.get("name")
+                tool_id = str(tool_doc.get("_id"))
+
+                if filtering_enabled:
+                    required_actions_by_name = required_tool_actions.get(
+                        tool_name, set()
+                    )
+                    required_actions_by_id = required_tool_actions.get(tool_id, set())
+
+                    required_actions = required_actions_by_name | required_actions_by_id
+
+                    if not required_actions:
+                        continue
+                else:
+                    required_actions = None
+
+                tool_data = self._fetch_tool_data(tool_doc, required_actions)
+                if tool_data:
+                    tools_data[tool_name] = tool_data
+                    tools_data[tool_id] = tool_data
+
+            return tools_data if tools_data else None
+        except Exception as e:
+            logger.warning(f"Failed to pre-fetch tools: {type(e).__name__}")
+            return None
+
+    def _fetch_tool_data(
+        self,
+        tool_doc: Dict[str, Any],
+        required_actions: Optional[Set[Optional[str]]],
+    ) -> Optional[Dict[str, Any]]:
+        """Fetch and execute tool actions with saved parameters"""
+        try:
+            from application.agents.tools.tool_manager import ToolManager
+
+            tool_name = tool_doc.get("name")
+            tool_config = tool_doc.get("config", {}).copy()
+            tool_config["tool_id"] = str(tool_doc["_id"])
+
+            tool_manager = ToolManager(config={tool_name: tool_config})
+            user_id = self.initial_user_id or "local"
+            tool = tool_manager.load_tool(tool_name, tool_config, user_id=user_id)
+
+            if not tool:
+                logger.debug(f"Tool '{tool_name}' failed to load")
+                return None
+
+            tool_actions = tool.get_actions_metadata()
+            if not tool_actions:
+                logger.debug(f"Tool '{tool_name}' has no actions")
+                return None
+
+            saved_actions = tool_doc.get("actions", [])
+
+            include_all_actions = required_actions is None or (
+                required_actions and None in required_actions
+            )
+            allowed_actions: Set[str] = (
+                {action for action in required_actions if isinstance(action, str)}
+                if required_actions
+                else set()
+            )
+
+            action_results = {}
+            for action_meta in tool_actions:
+                action_name = action_meta.get("name")
+                if action_name is None:
+                    continue
+                if (
+                    not include_all_actions
+                    and allowed_actions
+                    and action_name not in allowed_actions
+                ):
+                    continue
+
+                try:
+                    saved_action = None
+                    for sa in saved_actions:
+                        if sa.get("name") == action_name:
+                            saved_action = sa
+                            break
+
+                    action_params = action_meta.get("parameters", {})
+                    properties = action_params.get("properties", {})
+
+                    kwargs = {}
+                    for param_name, param_spec in properties.items():
+                        if saved_action:
+                            saved_props = saved_action.get("parameters", {}).get(
+                                "properties", {}
+                            )
+                            if param_name in saved_props:
+                                param_value = saved_props[param_name].get("value")
+                                if param_value is not None:
+                                    kwargs[param_name] = param_value
+                                    continue
+
+                        if param_name in tool_config:
+                            kwargs[param_name] = tool_config[param_name]
+                        elif "default" in param_spec:
+                            kwargs[param_name] = param_spec["default"]
+
+                    result = tool.execute_action(action_name, **kwargs)
+                    action_results[action_name] = result
+                except Exception as e:
+                    logger.debug(
+                        f"Action '{action_name}' execution failed: {type(e).__name__}"
+                    )
+                    continue
+
+            return action_results if action_results else None
+
+        except Exception as e:
+            logger.debug(f"Tool pre-fetch failed for '{tool_name}': {type(e).__name__}")
+            return None
+
+    def _get_prompt_content(self) -> Optional[str]:
+        """Retrieve and cache the raw prompt content for the current agent configuration."""
+        if self._prompt_content is not None:
+            return self._prompt_content
+        prompt_id = (
+            self.agent_config.get("prompt_id")
+            if isinstance(self.agent_config, dict)
+            else None
+        )
+        if not prompt_id:
+            return None
+        try:
+            self._prompt_content = get_prompt(prompt_id, self.prompts_collection)
+        except ValueError as e:
+            logger.debug(f"Invalid prompt ID '{prompt_id}': {str(e)}")
+            self._prompt_content = None
+        except Exception as e:
+            logger.debug(f"Failed to fetch prompt '{prompt_id}': {type(e).__name__}")
+            self._prompt_content = None
+        return self._prompt_content
+
+    def _get_required_tool_actions(self) -> Optional[Dict[str, Set[Optional[str]]]]:
+        """Determine which tool actions are referenced in the prompt template"""
+        if self._required_tool_actions is not None:
+            return self._required_tool_actions
+
+        prompt_content = self._get_prompt_content()
+        if prompt_content is None:
+            return None
+
+        if "{{" not in prompt_content or "}}" not in prompt_content:
+            self._required_tool_actions = {}
+            return self._required_tool_actions
+
+        try:
+            from application.templates.template_engine import TemplateEngine
+
+            template_engine = TemplateEngine()
+            usages = template_engine.extract_tool_usages(prompt_content)
+            self._required_tool_actions = usages
+            return self._required_tool_actions
+        except Exception as e:
+            logger.debug(f"Failed to extract tool usages: {type(e).__name__}")
+            self._required_tool_actions = {}
+            return self._required_tool_actions
+
+    def _fetch_memory_tool_data(
+        self, tool_doc: Dict[str, Any]
+    ) -> Optional[Dict[str, Any]]:
+        """Fetch memory tool data for pre-injection into prompt"""
+        try:
+            tool_config = tool_doc.get("config", {}).copy()
+            tool_config["tool_id"] = str(tool_doc["_id"])
+
+            from application.agents.tools.memory import MemoryTool
+
+            memory_tool = MemoryTool(tool_config, self.initial_user_id)
+
+            root_view = memory_tool.execute_action("view", path="/")
+
+            if "Error:" in root_view or not root_view.strip():
+                return None
+
+            return {"root": root_view, "available": True}
+        except Exception as e:
+            logger.warning(f"Failed to fetch memory tool data: {str(e)}")
+            return None
+
+    def create_agent(
+        self,
+        docs_together: Optional[str] = None,
+        docs: Optional[list] = None,
+        tools_data: Optional[Dict[str, Any]] = None,
+    ):
+        """Create and return the configured agent with rendered prompt"""
+        raw_prompt = self._get_prompt_content()
+        if raw_prompt is None:
+            raw_prompt = get_prompt(
+                self.agent_config["prompt_id"], self.prompts_collection
+            )
+            self._prompt_content = raw_prompt
+
+        rendered_prompt = self.prompt_renderer.render_prompt(
+            prompt_content=raw_prompt,
+            user_id=self.initial_user_id,
+            request_id=self.data.get("request_id"),
+            passthrough_data=self.data.get("passthrough"),
+            docs=docs,
+            docs_together=docs_together,
+            tools_data=tools_data,
+        )
+
+        provider = (
+            get_provider_from_model_id(self.model_id)
+            if self.model_id
+            else settings.LLM_PROVIDER
+        )
+        system_api_key = get_api_key_for_provider(provider or settings.LLM_PROVIDER)
+
+        return AgentCreator.create_agent(
+            self.agent_config["agent_type"],
+            endpoint="stream",
+            llm_name=provider or settings.LLM_PROVIDER,
+            model_id=self.model_id,
+            api_key=system_api_key,
+            user_api_key=self.agent_config["user_api_key"],
+            prompt=rendered_prompt,
+            chat_history=self.history,
+            retrieved_docs=self.retrieved_docs,
+            decoded_token=self.decoded_token,
+            attachments=self.attachments,
+            json_schema=self.agent_config.get("json_schema"),
+        )

application/api/user/agents/routes.py

@@ -10,7 +10,6 @@ from flask import current_app, jsonify, make_response, request
 from flask_restx import fields, Namespace, Resource
 
 from application.api import api
-from application.core.settings import settings
 from application.api.user.base import (
     agents_collection,
     db,
@@ -20,6 +19,7 @@ from application.api.user.base import (
     storage,
     users_collection,
 )
+from application.core.settings import settings
 from application.utils import (
     check_required_fields,
     generate_image_url,
@@ -76,9 +76,13 @@ class GetAgent(Resource):
                 "status": agent.get("status", ""),
                 "json_schema": agent.get("json_schema"),
                 "limited_token_mode": agent.get("limited_token_mode", False),
-                "token_limit": agent.get("token_limit", settings.DEFAULT_AGENT_LIMITS["token_limit"]),
+                "token_limit": agent.get(
+                    "token_limit", settings.DEFAULT_AGENT_LIMITS["token_limit"]
+                ),
                 "limited_request_mode": agent.get("limited_request_mode", False),
-                "request_limit": agent.get("request_limit", settings.DEFAULT_AGENT_LIMITS["request_limit"]),
+                "request_limit": agent.get(
+                    "request_limit", settings.DEFAULT_AGENT_LIMITS["request_limit"]
+                ),
                 "created_at": agent.get("createdAt", ""),
                 "updated_at": agent.get("updatedAt", ""),
                 "last_used_at": agent.get("lastUsedAt", ""),
@@ -91,6 +95,8 @@ class GetAgent(Resource):
                 "shared": agent.get("shared_publicly", False),
                 "shared_metadata": agent.get("shared_metadata", {}),
                 "shared_token": agent.get("shared_token", ""),
+                "models": agent.get("models", []),
+                "default_model_id": agent.get("default_model_id", ""),
             }
             return make_response(jsonify(data), 200)
         except Exception as e:
@@ -149,9 +155,13 @@ class GetAgents(Resource):
                     "status": agent.get("status", ""),
                     "json_schema": agent.get("json_schema"),
                     "limited_token_mode": agent.get("limited_token_mode", False),
-                    "token_limit": agent.get("token_limit", settings.DEFAULT_AGENT_LIMITS["token_limit"]),
+                    "token_limit": agent.get(
+                        "token_limit", settings.DEFAULT_AGENT_LIMITS["token_limit"]
+                    ),
                     "limited_request_mode": agent.get("limited_request_mode", False),
-                    "request_limit": agent.get("request_limit", settings.DEFAULT_AGENT_LIMITS["request_limit"]),
+                    "request_limit": agent.get(
+                        "request_limit", settings.DEFAULT_AGENT_LIMITS["request_limit"]
+                    ),
                     "created_at": agent.get("createdAt", ""),
                     "updated_at": agent.get("updatedAt", ""),
                     "last_used_at": agent.get("lastUsedAt", ""),
@@ -164,6 +174,8 @@ class GetAgents(Resource):
                     "shared": agent.get("shared_publicly", False),
                     "shared_metadata": agent.get("shared_metadata", {}),
                     "shared_token": agent.get("shared_token", ""),
+                    "models": agent.get("models", []),
+                    "default_model_id": agent.get("default_model_id", ""),
                 }
                 for agent in agents
                 if "source" in agent or "retriever" in agent
@@ -209,21 +221,27 @@ class CreateAgent(Resource):
                 description="JSON schema for enforcing structured output format",
             ),
             "limited_token_mode": fields.Boolean(
-                required=False,
-                description="Whether the agent is in limited token mode"
+                required=False, description="Whether the agent is in limited token mode"
             ),
             "token_limit": fields.Integer(
-                required=False,
-                description="Token limit for the agent in limited mode"
+                required=False, description="Token limit for the agent in limited mode"
             ),
             "limited_request_mode": fields.Boolean(
                 required=False,
-                description="Whether the agent is in limited request mode"
+                description="Whether the agent is in limited request mode",
             ),
             "request_limit": fields.Integer(
                 required=False,
-                description="Request limit for the agent in limited mode"
-            )
+                description="Request limit for the agent in limited mode",
+            ),
+            "models": fields.List(
+                fields.String,
+                required=False,
+                description="List of available model IDs for this agent",
+            ),
+            "default_model_id": fields.String(
+                required=False, description="Default model ID for this agent"
+            ),
         },
     )
 
@@ -252,6 +270,11 @@ class CreateAgent(Resource):
                     data["json_schema"] = json.loads(data["json_schema"])
                 except json.JSONDecodeError:
                     data["json_schema"] = None
+            if "models" in data:
+                try:
+                    data["models"] = json.loads(data["models"])
+                except json.JSONDecodeError:
+                    data["models"] = []
         print(f"Received data: {data}")
 
         # Validate JSON schema if provided
@@ -369,14 +392,32 @@ class CreateAgent(Resource):
                 "agent_type": data.get("agent_type", ""),
                 "status": data.get("status"),
                 "json_schema": data.get("json_schema"),
-                "limited_token_mode": data.get("limited_token_mode", False),
-                "token_limit": data.get("token_limit", settings.DEFAULT_AGENT_LIMITS["token_limit"]),
-                "limited_request_mode": data.get("limited_request_mode", False),
-                "request_limit": data.get("request_limit", settings.DEFAULT_AGENT_LIMITS["request_limit"]),
+                "limited_token_mode": (
+                    data.get("limited_token_mode") == "True"
+                    if isinstance(data.get("limited_token_mode"), str)
+                    else bool(data.get("limited_token_mode", False))
+                ),
+                "token_limit": int(
+                    data.get(
+                        "token_limit", settings.DEFAULT_AGENT_LIMITS["token_limit"]
+                    )
+                ),
+                "limited_request_mode": (
+                    data.get("limited_request_mode") == "True"
+                    if isinstance(data.get("limited_request_mode"), str)
+                    else bool(data.get("limited_request_mode", False))
+                ),
+                "request_limit": int(
+                    data.get(
+                        "request_limit", settings.DEFAULT_AGENT_LIMITS["request_limit"]
+                    )
+                ),
                 "createdAt": datetime.datetime.now(datetime.timezone.utc),
                 "updatedAt": datetime.datetime.now(datetime.timezone.utc),
                 "lastUsedAt": None,
                 "key": key,
+                "models": data.get("models", []),
+                "default_model_id": data.get("default_model_id", ""),
             }
             if new_agent["chunks"] == "":
                 new_agent["chunks"] = "2"
@@ -429,21 +470,27 @@ class UpdateAgent(Resource):
                 description="JSON schema for enforcing structured output format",
             ),
             "limited_token_mode": fields.Boolean(
-                required=False,
-                description="Whether the agent is in limited token mode"
+                required=False, description="Whether the agent is in limited token mode"
             ),
             "token_limit": fields.Integer(
-                required=False,
-                description="Token limit for the agent in limited mode"
+                required=False, description="Token limit for the agent in limited mode"
             ),
             "limited_request_mode": fields.Boolean(
                 require=False,
-                description="Whether the agent is in limited request mode"
+                description="Whether the agent is in limited request mode",
             ),
             "request_limit": fields.Integer(
                 required=False,
-                description="Request limit for the agent in limited mode"
-            )
+                description="Request limit for the agent in limited mode",
+            ),
+            "models": fields.List(
+                fields.String,
+                required=False,
+                description="List of available model IDs for this agent",
+            ),
+            "default_model_id": fields.String(
+                required=False, description="Default model ID for this agent"
+            ),
         },
     )
 
@@ -467,7 +514,7 @@ class UpdateAgent(Resource):
                 data = request.get_json()
             else:
                 data = request.form.to_dict()
-                json_fields = ["tools", "sources", "json_schema"]
+                json_fields = ["tools", "sources", "json_schema", "models"]
                 for field in json_fields:
                     if field in data and data[field]:
                         try:
@@ -534,7 +581,9 @@ class UpdateAgent(Resource):
             "limited_token_mode",
             "token_limit",
             "limited_request_mode",
-            "request_limit"
+            "request_limit",
+            "models",
+            "default_model_id",
         ]
 
         for field in allowed_fields:
@@ -652,8 +701,15 @@ class UpdateAgent(Resource):
                 else:
                     update_fields[field] = None
             elif field == "limited_token_mode":
-                is_mode_enabled = data.get("limited_token_mode", False)
-                if is_mode_enabled and data.get("token_limit") is None:
+                raw_value = data.get("limited_token_mode", False)
+                bool_value = (
+                    raw_value == "True"
+                    if isinstance(raw_value, str)
+                    else bool(raw_value)
+                )
+                update_fields[field] = bool_value
+
+                if bool_value and data.get("token_limit") is None:
                     return make_response(
                         jsonify(
                             {
@@ -664,8 +720,15 @@ class UpdateAgent(Resource):
                         400,
                     )
             elif field == "limited_request_mode":
-                is_mode_enabled = data.get("limited_request_mode", False)
-                if is_mode_enabled and data.get("request_limit") is None:
+                raw_value = data.get("limited_request_mode", False)
+                bool_value = (
+                    raw_value == "True"
+                    if isinstance(raw_value, str)
+                    else bool(raw_value)
+                )
+                update_fields[field] = bool_value
+
+                if bool_value and data.get("request_limit") is None:
                     return make_response(
                         jsonify(
                             {
@@ -677,7 +740,11 @@ class UpdateAgent(Resource):
                     )
             elif field == "token_limit":
                 token_limit = data.get("token_limit")
-                if token_limit is not None and not data.get("limited_token_mode"):
+                # Convert to int and store
+                update_fields[field] = int(token_limit) if token_limit else 0
+
+                # Validate consistency with mode
+                if update_fields[field] > 0 and not data.get("limited_token_mode"):
                     return make_response(
                         jsonify(
                             {
@@ -689,7 +756,9 @@ class UpdateAgent(Resource):
                     )
             elif field == "request_limit":
                 request_limit = data.get("request_limit")
-                if request_limit is not None and not data.get("limited_request_mode"):
+                update_fields[field] = int(request_limit) if request_limit else 0
+
+                if update_fields[field] > 0 and not data.get("limited_request_mode"):
                     return make_response(
                         jsonify(
                             {

application/api/user/attachments/routes.py

@@ -10,7 +10,7 @@ from application.api import api
 from application.api.user.base import agents_collection, storage
 from application.api.user.tasks import store_attachment
 from application.core.settings import settings
-from application.tts.google_tts import GoogleTTS
+from application.tts.tts_creator import TTSCreator
 from application.utils import safe_filename
 
 
@@ -25,7 +25,7 @@ class StoreAttachment(Resource):
         api.model(
             "AttachmentModel",
             {
-                "file": fields.Raw(required=True, description="File to upload"),
+                "file": fields.Raw(required=True, description="File(s) to upload"),
                 "api_key": fields.String(
                     required=False, description="API key (optional)"
                 ),
@@ -33,18 +33,24 @@ class StoreAttachment(Resource):
         )
     )
     @api.doc(
-        description="Stores a single attachment without vectorization or training. Supports user or API key authentication."
+        description="Stores one or multiple attachments without vectorization or training. Supports user or API key authentication."
     )
     def post(self):
         decoded_token = getattr(request, "decoded_token", None)
         api_key = request.form.get("api_key") or request.args.get("api_key")
-        file = request.files.get("file")
-
-        if not file or file.filename == "":
+        
+        files = request.files.getlist("file")
+        if not files:
+            single_file = request.files.get("file")
+            if single_file:
+                files = [single_file]
+        
+        if not files or all(f.filename == "" for f in files):
             return make_response(
-                jsonify({"status": "error", "message": "Missing file"}),
+                jsonify({"status": "error", "message": "Missing file(s)"}),
                 400,
             )
+        
         user = None
         if decoded_token:
             user = safe_filename(decoded_token.get("sub"))
@@ -59,32 +65,74 @@ class StoreAttachment(Resource):
             return make_response(
                 jsonify({"success": False, "message": "Authentication required"}), 401
             )
+        
         try:
-            attachment_id = ObjectId()
-            original_filename = safe_filename(os.path.basename(file.filename))
-            relative_path = f"{settings.UPLOAD_FOLDER}/{user}/attachments/{str(attachment_id)}/{original_filename}"
+            tasks = []
+            errors = []
+            original_file_count = len(files)
+            
+            for idx, file in enumerate(files):
+                try:
+                    attachment_id = ObjectId()
+                    original_filename = safe_filename(os.path.basename(file.filename))
+                    relative_path = f"{settings.UPLOAD_FOLDER}/{user}/attachments/{str(attachment_id)}/{original_filename}"
 
-            metadata = storage.save_file(file, relative_path)
-
-            file_info = {
-                "filename": original_filename,
-                "attachment_id": str(attachment_id),
-                "path": relative_path,
-                "metadata": metadata,
-            }
-
-            task = store_attachment.delay(file_info, user)
-
-            return make_response(
-                jsonify(
-                    {
-                        "success": True,
-                        "task_id": task.id,
-                        "message": "File uploaded successfully. Processing started.",
+                    metadata = storage.save_file(file, relative_path)
+                    file_info = {
+                        "filename": original_filename,
+                        "attachment_id": str(attachment_id),
+                        "path": relative_path,
+                        "metadata": metadata,
                     }
-                ),
-                200,
-            )
+
+                    task = store_attachment.delay(file_info, user)
+                    tasks.append({
+                        "task_id": task.id,
+                        "filename": original_filename,
+                        "attachment_id": str(attachment_id),
+                    })
+                except Exception as file_err:
+                    current_app.logger.error(f"Error processing file {idx} ({file.filename}): {file_err}", exc_info=True)
+                    errors.append({
+                        "filename": file.filename,
+                        "error": str(file_err)
+                    })
+            
+            if not tasks:
+                error_msg = "No valid files to upload"
+                if errors:
+                    error_msg += f". Errors: {errors}"
+                return make_response(
+                    jsonify({"status": "error", "message": error_msg, "errors": errors}),
+                    400,
+                )
+            
+            if original_file_count == 1 and len(tasks) == 1:
+                current_app.logger.info("Returning single task_id response")
+                return make_response(
+                    jsonify(
+                        {
+                            "success": True,
+                            "task_id": tasks[0]["task_id"],
+                            "message": "File uploaded successfully. Processing started.",
+                        }
+                    ),
+                    200,
+                )
+            else:
+                response_data = {
+                    "success": True,
+                    "tasks": tasks,
+                    "message": f"{len(tasks)} file(s) uploaded successfully. Processing started.",
+                }
+                if errors:
+                    response_data["errors"] = errors
+                    response_data["message"] += f" {len(errors)} file(s) failed."
+                
+                return make_response(
+                    jsonify(response_data),
+                    200,
+                )
         except Exception as err:
             current_app.logger.error(f"Error storing attachment: {err}", exc_info=True)
             return make_response(jsonify({"success": False, "error": str(err)}), 400)
@@ -133,7 +181,7 @@ class TextToSpeech(Resource):
         data = request.get_json()
         text = data["text"]
         try:
-            tts_instance = GoogleTTS()
+            tts_instance = TTSCreator.create_tts(settings.TTS_PROVIDER)
             audio_base64, detected_language = tts_instance.text_to_speech(text)
             return make_response(
                 jsonify(

application/api/user/models/__init__.py

@@ -0,0 +1,3 @@
+from .routes import models_ns
+
+__all__ = ["models_ns"]

application/api/user/models/routes.py

@@ -0,0 +1,25 @@
+from flask import current_app, jsonify, make_response
+from flask_restx import Namespace, Resource
+
+from application.core.model_settings import ModelRegistry
+
+models_ns = Namespace("models", description="Available models", path="/api")
+
+
+@models_ns.route("/models")
+class ModelsListResource(Resource):
+    def get(self):
+        """Get list of available models with their capabilities."""
+        try:
+            registry = ModelRegistry.get_instance()
+            models = registry.get_enabled_models()
+
+            response = {
+                "models": [model.to_dict() for model in models],
+                "default_model_id": registry.default_model_id,
+                "count": len(models),
+            }
+        except Exception as err:
+            current_app.logger.error(f"Error fetching models: {err}", exc_info=True)
+            return make_response(jsonify({"success": False}), 500)
+        return make_response(jsonify(response), 200)

application/api/user/routes.py

@@ -10,6 +10,7 @@ from .agents import agents_ns, agents_sharing_ns, agents_webhooks_ns
 from .analytics import analytics_ns
 from .attachments import attachments_ns
 from .conversations import conversations_ns
+from .models import models_ns
 from .prompts import prompts_ns
 from .sharing import sharing_ns
 from .sources import sources_chunks_ns, sources_ns, sources_upload_ns
@@ -27,6 +28,9 @@ api.add_namespace(attachments_ns)
 # Conversations
 api.add_namespace(conversations_ns)
 
+# Models
+api.add_namespace(models_ns)
+
 # Agents (main, sharing, webhooks)
 api.add_namespace(agents_ns)
 api.add_namespace(agents_sharing_ns)

application/api/user/sharing/routes.py

@@ -13,7 +13,6 @@ from application.api.user.base import (
     agents_collection,
     attachments_collection,
     conversations_collection,
-    db,
     shared_conversations_collections,
 )
 from application.utils import check_required_fields
@@ -97,9 +96,7 @@ class ShareConversation(Resource):
                     api_uuid = pre_existing_api_document["key"]
                     pre_existing = shared_conversations_collections.find_one(
                         {
-                            "conversation_id": DBRef(
-                                "conversations", ObjectId(conversation_id)
-                            ),
+                            "conversation_id": ObjectId(conversation_id),
                             "isPromptable": is_promptable,
                             "first_n_queries": current_n_queries,
                             "user": user,
@@ -120,10 +117,7 @@ class ShareConversation(Resource):
                         shared_conversations_collections.insert_one(
                             {
                                 "uuid": explicit_binary,
-                                "conversation_id": {
-                                    "$ref": "conversations",
-                                    "$id": ObjectId(conversation_id),
-                                },
+                                "conversation_id": ObjectId(conversation_id),
                                 "isPromptable": is_promptable,
                                 "first_n_queries": current_n_queries,
                                 "user": user,
@@ -154,10 +148,7 @@ class ShareConversation(Resource):
                     shared_conversations_collections.insert_one(
                         {
                             "uuid": explicit_binary,
-                            "conversation_id": {
-                                "$ref": "conversations",
-                                "$id": ObjectId(conversation_id),
-                            },
+                            "conversation_id": ObjectId(conversation_id),
                             "isPromptable": is_promptable,
                             "first_n_queries": current_n_queries,
                             "user": user,
@@ -175,9 +166,7 @@ class ShareConversation(Resource):
                     )
             pre_existing = shared_conversations_collections.find_one(
                 {
-                    "conversation_id": DBRef(
-                        "conversations", ObjectId(conversation_id)
-                    ),
+                    "conversation_id": ObjectId(conversation_id),
                     "isPromptable": is_promptable,
                     "first_n_queries": current_n_queries,
                     "user": user,
@@ -197,10 +186,7 @@ class ShareConversation(Resource):
                 shared_conversations_collections.insert_one(
                     {
                         "uuid": explicit_binary,
-                        "conversation_id": {
-                            "$ref": "conversations",
-                            "$id": ObjectId(conversation_id),
-                        },
+                        "conversation_id": ObjectId(conversation_id),
                         "isPromptable": is_promptable,
                         "first_n_queries": current_n_queries,
                         "user": user,
@@ -233,10 +219,12 @@ class GetPubliclySharedConversations(Resource):
             if (
                 shared
                 and "conversation_id" in shared
-                and isinstance(shared["conversation_id"], DBRef)
             ):
-                conversation_ref = shared["conversation_id"]
-                conversation = db.dereference(conversation_ref)
+                # conversation_id is now stored as an ObjectId, not a DBRef
+                conversation_id = shared["conversation_id"]
+                conversation = conversations_collection.find_one(
+                    {"_id": conversation_id}
+                )
                 if conversation is None:
                     return make_response(
                         jsonify(

application/api/user/tools/routes.py

@@ -56,9 +56,10 @@ class GetTools(Resource):
             tools = user_tools_collection.find({"user": user})
             user_tools = []
             for tool in tools:
-                tool["id"] = str(tool["_id"])
-                tool.pop("_id")
-                user_tools.append(tool)
+                tool_copy = {**tool}
+                tool_copy["id"] = str(tool["_id"])
+                tool_copy.pop("_id", None)
+                user_tools.append(tool_copy)
         except Exception as err:
             current_app.logger.error(f"Error getting user tools: {err}", exc_info=True)
             return make_response(jsonify({"success": False}), 400)

application/core/model_configs.py

@@ -0,0 +1,223 @@
+"""
+Model configurations for all supported LLM providers.
+"""
+
+from application.core.model_settings import (
+    AvailableModel,
+    ModelCapabilities,
+    ModelProvider,
+)
+
+OPENAI_ATTACHMENTS = [
+    "application/pdf",
+    "image/png",
+    "image/jpeg",
+    "image/jpg",
+    "image/webp",
+    "image/gif",
+]
+
+GOOGLE_ATTACHMENTS = [
+    "application/pdf",
+    "image/png",
+    "image/jpeg",
+    "image/jpg",
+    "image/webp",
+    "image/gif",
+]
+
+
+OPENAI_MODELS = [
+    AvailableModel(
+        id="gpt-4o",
+        provider=ModelProvider.OPENAI,
+        display_name="GPT-4 Omni",
+        description="Latest and most capable model",
+        capabilities=ModelCapabilities(
+            supports_tools=True,
+            supports_structured_output=True,
+            supported_attachment_types=OPENAI_ATTACHMENTS,
+            context_window=128000,
+        ),
+    ),
+    AvailableModel(
+        id="gpt-4o-mini",
+        provider=ModelProvider.OPENAI,
+        display_name="GPT-4 Omni Mini",
+        description="Fast and efficient",
+        capabilities=ModelCapabilities(
+            supports_tools=True,
+            supports_structured_output=True,
+            supported_attachment_types=OPENAI_ATTACHMENTS,
+            context_window=128000,
+        ),
+    ),
+    AvailableModel(
+        id="gpt-4-turbo",
+        provider=ModelProvider.OPENAI,
+        display_name="GPT-4 Turbo",
+        description="Fast GPT-4 with 128k context",
+        capabilities=ModelCapabilities(
+            supports_tools=True,
+            supports_structured_output=True,
+            supported_attachment_types=OPENAI_ATTACHMENTS,
+            context_window=128000,
+        ),
+    ),
+    AvailableModel(
+        id="gpt-4",
+        provider=ModelProvider.OPENAI,
+        display_name="GPT-4",
+        description="Most capable model",
+        capabilities=ModelCapabilities(
+            supports_tools=True,
+            supports_structured_output=True,
+            supported_attachment_types=OPENAI_ATTACHMENTS,
+            context_window=8192,
+        ),
+    ),
+    AvailableModel(
+        id="gpt-3.5-turbo",
+        provider=ModelProvider.OPENAI,
+        display_name="GPT-3.5 Turbo",
+        description="Fast and cost-effective",
+        capabilities=ModelCapabilities(
+            supports_tools=True,
+            context_window=4096,
+        ),
+    ),
+]
+
+
+ANTHROPIC_MODELS = [
+    AvailableModel(
+        id="claude-3-5-sonnet-20241022",
+        provider=ModelProvider.ANTHROPIC,
+        display_name="Claude 3.5 Sonnet (Latest)",
+        description="Latest Claude 3.5 Sonnet with enhanced capabilities",
+        capabilities=ModelCapabilities(
+            supports_tools=True,
+            context_window=200000,
+        ),
+    ),
+    AvailableModel(
+        id="claude-3-5-sonnet",
+        provider=ModelProvider.ANTHROPIC,
+        display_name="Claude 3.5 Sonnet",
+        description="Balanced performance and capability",
+        capabilities=ModelCapabilities(
+            supports_tools=True,
+            context_window=200000,
+        ),
+    ),
+    AvailableModel(
+        id="claude-3-opus",
+        provider=ModelProvider.ANTHROPIC,
+        display_name="Claude 3 Opus",
+        description="Most capable Claude model",
+        capabilities=ModelCapabilities(
+            supports_tools=True,
+            context_window=200000,
+        ),
+    ),
+    AvailableModel(
+        id="claude-3-haiku",
+        provider=ModelProvider.ANTHROPIC,
+        display_name="Claude 3 Haiku",
+        description="Fastest Claude model",
+        capabilities=ModelCapabilities(
+            supports_tools=True,
+            context_window=200000,
+        ),
+    ),
+]
+
+
+GOOGLE_MODELS = [
+    AvailableModel(
+        id="gemini-flash-latest",
+        provider=ModelProvider.GOOGLE,
+        display_name="Gemini Flash (Latest)",
+        description="Latest experimental Gemini model",
+        capabilities=ModelCapabilities(
+            supports_tools=True,
+            supports_structured_output=True,
+            supported_attachment_types=GOOGLE_ATTACHMENTS,
+            context_window=int(1e6),
+        ),
+    ),
+    AvailableModel(
+        id="gemini-flash-lite-latest",
+        provider=ModelProvider.GOOGLE,
+        display_name="Gemini Flash Lite (Latest)",
+        description="Fast with huge context window",
+        capabilities=ModelCapabilities(
+            supports_tools=True,
+            supports_structured_output=True,
+            supported_attachment_types=GOOGLE_ATTACHMENTS,
+            context_window=int(1e6),
+        ),
+    ),
+    AvailableModel(
+        id="gemini-2.5-pro",
+        provider=ModelProvider.GOOGLE,
+        display_name="Gemini 2.5 Pro",
+        description="Most capable Gemini model",
+        capabilities=ModelCapabilities(
+            supports_tools=True,
+            supports_structured_output=True,
+            supported_attachment_types=GOOGLE_ATTACHMENTS,
+            context_window=2000000,
+        ),
+    ),
+]
+
+
+GROQ_MODELS = [
+    AvailableModel(
+        id="llama-3.3-70b-versatile",
+        provider=ModelProvider.GROQ,
+        display_name="Llama 3.3 70B",
+        description="Latest Llama model with high-speed inference",
+        capabilities=ModelCapabilities(
+            supports_tools=True,
+            context_window=128000,
+        ),
+    ),
+    AvailableModel(
+        id="llama-3.1-8b-instant",
+        provider=ModelProvider.GROQ,
+        display_name="Llama 3.1 8B",
+        description="Ultra-fast inference",
+        capabilities=ModelCapabilities(
+            supports_tools=True,
+            context_window=128000,
+        ),
+    ),
+    AvailableModel(
+        id="mixtral-8x7b-32768",
+        provider=ModelProvider.GROQ,
+        display_name="Mixtral 8x7B",
+        description="High-speed inference with tools",
+        capabilities=ModelCapabilities(
+            supports_tools=True,
+            context_window=32768,
+        ),
+    ),
+]
+
+
+AZURE_OPENAI_MODELS = [
+    AvailableModel(
+        id="azure-gpt-4",
+        provider=ModelProvider.AZURE_OPENAI,
+        display_name="Azure OpenAI GPT-4",
+        description="Azure-hosted GPT model",
+        capabilities=ModelCapabilities(
+            supports_tools=True,
+            supports_structured_output=True,
+            supported_attachment_types=OPENAI_ATTACHMENTS,
+            context_window=8192,
+        ),
+    ),
+]

application/core/model_settings.py

@@ -0,0 +1,236 @@
+import logging
+from dataclasses import dataclass, field
+from enum import Enum
+from typing import Dict, List, Optional
+
+logger = logging.getLogger(__name__)
+
+
+class ModelProvider(str, Enum):
+    OPENAI = "openai"
+    AZURE_OPENAI = "azure_openai"
+    ANTHROPIC = "anthropic"
+    GROQ = "groq"
+    GOOGLE = "google"
+    HUGGINGFACE = "huggingface"
+    LLAMA_CPP = "llama.cpp"
+    DOCSGPT = "docsgpt"
+    PREMAI = "premai"
+    SAGEMAKER = "sagemaker"
+    NOVITA = "novita"
+
+
+@dataclass
+class ModelCapabilities:
+    supports_tools: bool = False
+    supports_structured_output: bool = False
+    supports_streaming: bool = True
+    supported_attachment_types: List[str] = field(default_factory=list)
+    context_window: int = 128000
+    input_cost_per_token: Optional[float] = None
+    output_cost_per_token: Optional[float] = None
+
+
+@dataclass
+class AvailableModel:
+    id: str
+    provider: ModelProvider
+    display_name: str
+    description: str = ""
+    capabilities: ModelCapabilities = field(default_factory=ModelCapabilities)
+    enabled: bool = True
+    base_url: Optional[str] = None
+
+    def to_dict(self) -> Dict:
+        result = {
+            "id": self.id,
+            "provider": self.provider.value,
+            "display_name": self.display_name,
+            "description": self.description,
+            "supported_attachment_types": self.capabilities.supported_attachment_types,
+            "supports_tools": self.capabilities.supports_tools,
+            "supports_structured_output": self.capabilities.supports_structured_output,
+            "supports_streaming": self.capabilities.supports_streaming,
+            "context_window": self.capabilities.context_window,
+            "enabled": self.enabled,
+        }
+        if self.base_url:
+            result["base_url"] = self.base_url
+        return result
+
+
+class ModelRegistry:
+    _instance = None
+    _initialized = False
+
+    def __new__(cls):
+        if cls._instance is None:
+            cls._instance = super().__new__(cls)
+        return cls._instance
+
+    def __init__(self):
+        if not ModelRegistry._initialized:
+            self.models: Dict[str, AvailableModel] = {}
+            self.default_model_id: Optional[str] = None
+            self._load_models()
+            ModelRegistry._initialized = True
+
+    @classmethod
+    def get_instance(cls) -> "ModelRegistry":
+        return cls()
+
+    def _load_models(self):
+        from application.core.settings import settings
+
+        self.models.clear()
+
+        self._add_docsgpt_models(settings)
+        if settings.OPENAI_API_KEY or (
+            settings.LLM_PROVIDER == "openai" and settings.API_KEY
+        ):
+            self._add_openai_models(settings)
+        if settings.OPENAI_API_BASE or (
+            settings.LLM_PROVIDER == "azure_openai" and settings.API_KEY
+        ):
+            self._add_azure_openai_models(settings)
+        if settings.ANTHROPIC_API_KEY or (
+            settings.LLM_PROVIDER == "anthropic" and settings.API_KEY
+        ):
+            self._add_anthropic_models(settings)
+        if settings.GOOGLE_API_KEY or (
+            settings.LLM_PROVIDER == "google" and settings.API_KEY
+        ):
+            self._add_google_models(settings)
+        if settings.GROQ_API_KEY or (
+            settings.LLM_PROVIDER == "groq" and settings.API_KEY
+        ):
+            self._add_groq_models(settings)
+        if settings.HUGGINGFACE_API_KEY or (
+            settings.LLM_PROVIDER == "huggingface" and settings.API_KEY
+        ):
+            self._add_huggingface_models(settings)
+        # Default model selection
+
+        if settings.LLM_NAME and settings.LLM_NAME in self.models:
+            self.default_model_id = settings.LLM_NAME
+        elif settings.LLM_PROVIDER and settings.API_KEY:
+            for model_id, model in self.models.items():
+                if model.provider.value == settings.LLM_PROVIDER:
+                    self.default_model_id = model_id
+                    break
+        else:
+            self.default_model_id = next(iter(self.models.keys()))
+        logger.info(
+            f"ModelRegistry loaded {len(self.models)} models, default: {self.default_model_id}"
+        )
+
+    def _add_openai_models(self, settings):
+        from application.core.model_configs import OPENAI_MODELS
+
+        if settings.OPENAI_API_KEY:
+            for model in OPENAI_MODELS:
+                self.models[model.id] = model
+            return
+        if settings.LLM_PROVIDER == "openai" and settings.LLM_NAME:
+            for model in OPENAI_MODELS:
+                if model.id == settings.LLM_NAME:
+                    self.models[model.id] = model
+                    return
+        for model in OPENAI_MODELS:
+            self.models[model.id] = model
+
+    def _add_azure_openai_models(self, settings):
+        from application.core.model_configs import AZURE_OPENAI_MODELS
+
+        if settings.LLM_PROVIDER == "azure_openai" and settings.LLM_NAME:
+            for model in AZURE_OPENAI_MODELS:
+                if model.id == settings.LLM_NAME:
+                    self.models[model.id] = model
+                    return
+        for model in AZURE_OPENAI_MODELS:
+            self.models[model.id] = model
+
+    def _add_anthropic_models(self, settings):
+        from application.core.model_configs import ANTHROPIC_MODELS
+
+        if settings.ANTHROPIC_API_KEY:
+            for model in ANTHROPIC_MODELS:
+                self.models[model.id] = model
+            return
+        if settings.LLM_PROVIDER == "anthropic" and settings.LLM_NAME:
+            for model in ANTHROPIC_MODELS:
+                if model.id == settings.LLM_NAME:
+                    self.models[model.id] = model
+                    return
+        for model in ANTHROPIC_MODELS:
+            self.models[model.id] = model
+
+    def _add_google_models(self, settings):
+        from application.core.model_configs import GOOGLE_MODELS
+
+        if settings.GOOGLE_API_KEY:
+            for model in GOOGLE_MODELS:
+                self.models[model.id] = model
+            return
+        if settings.LLM_PROVIDER == "google" and settings.LLM_NAME:
+            for model in GOOGLE_MODELS:
+                if model.id == settings.LLM_NAME:
+                    self.models[model.id] = model
+                    return
+        for model in GOOGLE_MODELS:
+            self.models[model.id] = model
+
+    def _add_groq_models(self, settings):
+        from application.core.model_configs import GROQ_MODELS
+
+        if settings.GROQ_API_KEY:
+            for model in GROQ_MODELS:
+                self.models[model.id] = model
+            return
+        if settings.LLM_PROVIDER == "groq" and settings.LLM_NAME:
+            for model in GROQ_MODELS:
+                if model.id == settings.LLM_NAME:
+                    self.models[model.id] = model
+                    return
+        for model in GROQ_MODELS:
+            self.models[model.id] = model
+
+    def _add_docsgpt_models(self, settings):
+        model_id = "docsgpt-local"
+        model = AvailableModel(
+            id=model_id,
+            provider=ModelProvider.DOCSGPT,
+            display_name="DocsGPT Model",
+            description="Local model",
+            capabilities=ModelCapabilities(
+                supports_tools=False,
+                supported_attachment_types=[],
+            ),
+        )
+        self.models[model_id] = model
+
+    def _add_huggingface_models(self, settings):
+        model_id = "huggingface-local"
+        model = AvailableModel(
+            id=model_id,
+            provider=ModelProvider.HUGGINGFACE,
+            display_name="Hugging Face Model",
+            description="Local Hugging Face model",
+            capabilities=ModelCapabilities(
+                supports_tools=False,
+                supported_attachment_types=[],
+            ),
+        )
+        self.models[model_id] = model
+
+    def get_model(self, model_id: str) -> Optional[AvailableModel]:
+        return self.models.get(model_id)
+
+    def get_all_models(self) -> List[AvailableModel]:
+        return list(self.models.values())
+
+    def get_enabled_models(self) -> List[AvailableModel]:
+        return [m for m in self.models.values() if m.enabled]
+
+    def model_exists(self, model_id: str) -> bool:
+        return model_id in self.models

application/core/model_utils.py

@@ -0,0 +1,91 @@
+from typing import Any, Dict, Optional
+
+from application.core.model_settings import ModelRegistry
+
+
+def get_api_key_for_provider(provider: str) -> Optional[str]:
+    """Get the appropriate API key for a provider"""
+    from application.core.settings import settings
+
+    provider_key_map = {
+        "openai": settings.OPENAI_API_KEY,
+        "anthropic": settings.ANTHROPIC_API_KEY,
+        "google": settings.GOOGLE_API_KEY,
+        "groq": settings.GROQ_API_KEY,
+        "huggingface": settings.HUGGINGFACE_API_KEY,
+        "azure_openai": settings.API_KEY,
+        "docsgpt": None,
+        "llama.cpp": None,
+    }
+
+    provider_key = provider_key_map.get(provider)
+    if provider_key:
+        return provider_key
+    return settings.API_KEY
+
+
+def get_all_available_models() -> Dict[str, Dict[str, Any]]:
+    """Get all available models with metadata for API response"""
+    registry = ModelRegistry.get_instance()
+    return {model.id: model.to_dict() for model in registry.get_enabled_models()}
+
+
+def validate_model_id(model_id: str) -> bool:
+    """Check if a model ID exists in registry"""
+    registry = ModelRegistry.get_instance()
+    return registry.model_exists(model_id)
+
+
+def get_model_capabilities(model_id: str) -> Optional[Dict[str, Any]]:
+    """Get capabilities for a specific model"""
+    registry = ModelRegistry.get_instance()
+    model = registry.get_model(model_id)
+    if model:
+        return {
+            "supported_attachment_types": model.capabilities.supported_attachment_types,
+            "supports_tools": model.capabilities.supports_tools,
+            "supports_structured_output": model.capabilities.supports_structured_output,
+            "context_window": model.capabilities.context_window,
+        }
+    return None
+
+
+def get_default_model_id() -> str:
+    """Get the system default model ID"""
+    registry = ModelRegistry.get_instance()
+    return registry.default_model_id
+
+
+def get_provider_from_model_id(model_id: str) -> Optional[str]:
+    """Get the provider name for a given model_id"""
+    registry = ModelRegistry.get_instance()
+    model = registry.get_model(model_id)
+    if model:
+        return model.provider.value
+    return None
+
+
+def get_token_limit(model_id: str) -> int:
+    """
+    Get context window (token limit) for a model.
+    Returns model's context_window or default 128000 if model not found.
+    """
+    from application.core.settings import settings
+
+    registry = ModelRegistry.get_instance()
+    model = registry.get_model(model_id)
+    if model:
+        return model.capabilities.context_window
+    return settings.DEFAULT_LLM_TOKEN_LIMIT
+
+
+def get_base_url_for_model(model_id: str) -> Optional[str]:
+    """
+    Get the custom base_url for a specific model if configured.
+    Returns None if no custom base_url is set.
+    """
+    registry = ModelRegistry.get_instance()
+    model = registry.get_model(model_id)
+    if model:
+        return model.base_url
+    return None

application/core/settings.py

@@ -22,11 +22,11 @@ class Settings(BaseSettings):
     MONGO_DB_NAME: str = "docsgpt"
     LLM_PATH: str = os.path.join(current_dir, "models/docsgpt-7b-f16.gguf")
     DEFAULT_MAX_HISTORY: int = 150
-    LLM_TOKEN_LIMITS: dict = {
-        "gpt-4o-mini": 128000,
-        "gpt-3.5-turbo": 4096,
-        "claude-2": 1e5,
-        "gemini-2.5-flash": 1e6,
+    DEFAULT_LLM_TOKEN_LIMIT: int = 128000  # Fallback when model not found in registry
+    RESERVED_TOKENS: dict = {
+        "system_prompt": 500,
+        "current_query": 500,
+        "safety_buffer": 1000,
     }
     DEFAULT_AGENT_LIMITS: dict = {
         "token_limit": 50000,
@@ -56,14 +56,22 @@ class Settings(BaseSettings):
     )
 
     # GitHub source
-    GITHUB_ACCESS_TOKEN: Optional[str] = None # PAT token with read repo access
+    GITHUB_ACCESS_TOKEN: Optional[str] = None  # PAT token with read repo access
 
     # LLM Cache
     CACHE_REDIS_URL: str = "redis://localhost:6379/2"
 
     API_URL: str = "http://localhost:7091"  # backend url for celery worker
 
-    API_KEY: Optional[str] = None  # LLM api key
+    API_KEY: Optional[str] = None  # LLM api key (used by LLM_PROVIDER)
+
+    # Provider-specific API keys (for multi-model support)
+    OPENAI_API_KEY: Optional[str] = None
+    ANTHROPIC_API_KEY: Optional[str] = None
+    GOOGLE_API_KEY: Optional[str] = None
+    GROQ_API_KEY: Optional[str] = None
+    HUGGINGFACE_API_KEY: Optional[str] = None
+
     EMBEDDINGS_KEY: Optional[str] = (
         None  # api key for embeddings (if using openai, just copy API_KEY)
     )
@@ -130,7 +138,12 @@ class Settings(BaseSettings):
     # Encryption settings
     ENCRYPTION_SECRET_KEY: str = "default-docsgpt-encryption-key"
 
+    TTS_PROVIDER: str = "google_tts"  # google_tts or elevenlabs
     ELEVENLABS_API_KEY: Optional[str] = None
 
+    # Tool pre-fetch settings
+    ENABLE_TOOL_PREFETCH: bool = True
+
+
 path = Path(__file__).parent.parent.absolute()
 settings = Settings(_env_file=path.joinpath(".env"), _env_file_encoding="utf-8")

application/llm/anthropic.py

@@ -1,30 +1,41 @@
-from application.llm.base import BaseLLM
+from anthropic import AI_PROMPT, Anthropic, HUMAN_PROMPT
+
 from application.core.settings import settings
+from application.llm.base import BaseLLM
 
 
 class AnthropicLLM(BaseLLM):
 
-    def __init__(self, api_key=None, user_api_key=None, *args, **kwargs):
-        from anthropic import Anthropic, HUMAN_PROMPT, AI_PROMPT
+    def __init__(self, api_key=None, user_api_key=None, base_url=None, *args, **kwargs):
 
         super().__init__(*args, **kwargs)
-        self.api_key = (
-            api_key or settings.ANTHROPIC_API_KEY
-        )  # If not provided, use a default from settings
+        self.api_key = api_key or settings.ANTHROPIC_API_KEY or settings.API_KEY
         self.user_api_key = user_api_key
-        self.anthropic = Anthropic(api_key=self.api_key)
+
+        # Use custom base_url if provided
+        if base_url:
+            self.anthropic = Anthropic(api_key=self.api_key, base_url=base_url)
+        else:
+            self.anthropic = Anthropic(api_key=self.api_key)
+
         self.HUMAN_PROMPT = HUMAN_PROMPT
         self.AI_PROMPT = AI_PROMPT
 
     def _raw_gen(
-        self, baseself, model, messages, stream=False, tools=None, max_tokens=300, **kwargs
+        self,
+        baseself,
+        model,
+        messages,
+        stream=False,
+        tools=None,
+        max_tokens=300,
+        **kwargs,
     ):
         context = messages[0]["content"]
         user_question = messages[-1]["content"]
         prompt = f"### Context \n {context} \n ### Question \n {user_question}"
         if stream:
             return self.gen_stream(model, prompt, stream, max_tokens, **kwargs)
-
         completion = self.anthropic.completions.create(
             model=model,
             max_tokens_to_sample=max_tokens,
@@ -34,7 +45,14 @@ class AnthropicLLM(BaseLLM):
         return completion.completion
 
     def _raw_gen_stream(
-        self, baseself, model, messages, stream=True, tools=None, max_tokens=300, **kwargs
+        self,
+        baseself,
+        model,
+        messages,
+        stream=True,
+        tools=None,
+        max_tokens=300,
+        **kwargs,
     ):
         context = messages[0]["content"]
         user_question = messages[-1]["content"]
@@ -50,5 +68,5 @@ class AnthropicLLM(BaseLLM):
             for completion in stream_response:
                 yield completion.completion
         finally:
-            if hasattr(stream_response, 'close'):
+            if hasattr(stream_response, "close"):
                 stream_response.close()

application/llm/base.py

@@ -13,30 +13,32 @@ class BaseLLM(ABC):
     def __init__(
         self,
         decoded_token=None,
+        model_id=None,
+        base_url=None,
     ):
         self.decoded_token = decoded_token
+        self.model_id = model_id
+        self.base_url = base_url
         self.token_usage = {"prompt_tokens": 0, "generated_tokens": 0}
-        self.fallback_provider = settings.FALLBACK_LLM_PROVIDER
-        self.fallback_model_name = settings.FALLBACK_LLM_NAME
-        self.fallback_llm_api_key = settings.FALLBACK_LLM_API_KEY
         self._fallback_llm = None
+        self._fallback_sequence_index = 0
 
     @property
     def fallback_llm(self):
-        """Lazy-loaded fallback LLM instance."""
-        if (
-            self._fallback_llm is None
-            and self.fallback_provider
-            and self.fallback_model_name
-        ):
+        """Lazy-loaded fallback LLM from FALLBACK_* settings."""
+        if self._fallback_llm is None and settings.FALLBACK_LLM_PROVIDER:
             try:
                 from application.llm.llm_creator import LLMCreator
 
                 self._fallback_llm = LLMCreator.create_llm(
-                    self.fallback_provider,
-                    self.fallback_llm_api_key,
-                    None,
-                    self.decoded_token,
+                    settings.FALLBACK_LLM_PROVIDER,
+                    api_key=settings.FALLBACK_LLM_API_KEY or settings.API_KEY,
+                    user_api_key=None,
+                    decoded_token=self.decoded_token,
+                    model_id=settings.FALLBACK_LLM_NAME,
+                )
+                logger.info(
+                    f"Fallback LLM initialized: {settings.FALLBACK_LLM_PROVIDER}/{settings.FALLBACK_LLM_NAME}"
                 )
             except Exception as e:
                 logger.error(
@@ -44,11 +46,17 @@ class BaseLLM(ABC):
                 )
         return self._fallback_llm
 
+    @staticmethod
+    def _remove_null_values(args_dict):
+        if not isinstance(args_dict, dict):
+            return args_dict
+        return {k: v for k, v in args_dict.items() if v is not None}
+
     def _execute_with_fallback(
         self, method_name: str, decorators: list, *args, **kwargs
     ):
         """
-        Unified method execution with fallback support.
+        Execute method with fallback support.
 
         Args:
             method_name: Name of the raw method ('_raw_gen' or '_raw_gen_stream')
@@ -67,10 +75,10 @@ class BaseLLM(ABC):
             return decorated_method()
         except Exception as e:
             if not self.fallback_llm:
-                logger.error(f"Primary LLM failed and no fallback available: {str(e)}")
+                logger.error(f"Primary LLM failed and no fallback configured: {str(e)}")
                 raise
             logger.warning(
-                f"Falling back to {self.fallback_provider}/{self.fallback_model_name}. Error: {str(e)}"
+                f"Primary LLM failed. Falling back to {settings.FALLBACK_LLM_PROVIDER}/{settings.FALLBACK_LLM_NAME}. Error: {str(e)}"
             )
 
             fallback_method = getattr(

application/llm/docsgpt_provider.py

@@ -1,5 +1,7 @@
 import json
 
+from openai import OpenAI
+
 from application.core.settings import settings
 from application.llm.base import BaseLLM
 
@@ -7,12 +9,11 @@ from application.llm.base import BaseLLM
 class DocsGPTAPILLM(BaseLLM):
 
     def __init__(self, api_key=None, user_api_key=None, *args, **kwargs):
-        from openai import OpenAI
 
         super().__init__(*args, **kwargs)
-        self.client = OpenAI(api_key="sk-docsgpt-public", base_url="https://oai.arc53.com")
+        self.api_key = "sk-docsgpt-public"
+        self.client = OpenAI(api_key=self.api_key, base_url="https://oai.arc53.com")
         self.user_api_key = user_api_key
-        self.api_key = api_key
 
     def _clean_messages_openai(self, messages):
         cleaned_messages = []
@@ -22,7 +23,6 @@ class DocsGPTAPILLM(BaseLLM):
 
             if role == "model":
                 role = "assistant"
-
             if role and content is not None:
                 if isinstance(content, str):
                     cleaned_messages.append({"role": role, "content": content})
@@ -33,14 +33,15 @@ class DocsGPTAPILLM(BaseLLM):
                                 {"role": role, "content": item["text"]}
                             )
                         elif "function_call" in item:
+                            cleaned_args = self._remove_null_values(
+                                item["function_call"]["args"]
+                            )
                             tool_call = {
                                 "id": item["function_call"]["call_id"],
                                 "type": "function",
                                 "function": {
                                     "name": item["function_call"]["name"],
-                                    "arguments": json.dumps(
-                                        item["function_call"]["args"]
-                                    ),
+                                    "arguments": json.dumps(cleaned_args),
                                 },
                             }
                             cleaned_messages.append(
@@ -68,7 +69,6 @@ class DocsGPTAPILLM(BaseLLM):
                             )
                 else:
                     raise ValueError(f"Unexpected content type: {type(content)}")
-
         return cleaned_messages
 
     def _raw_gen(
@@ -120,7 +120,6 @@ class DocsGPTAPILLM(BaseLLM):
             response = self.client.chat.completions.create(
                 model="docsgpt", messages=messages, stream=stream, **kwargs
             )
-
         try:
             for line in response:
                 if (
@@ -132,8 +131,8 @@ class DocsGPTAPILLM(BaseLLM):
                 elif len(line.choices) > 0:
                     yield line.choices[0]
         finally:
-            if hasattr(response, 'close'):
+            if hasattr(response, "close"):
                 response.close()
 
     def _supports_tools(self):
-        return True
+        return True

application/llm/google_ai.py

@@ -13,8 +13,9 @@ from application.storage.storage_creator import StorageCreator
 class GoogleLLM(BaseLLM):
     def __init__(self, api_key=None, user_api_key=None, *args, **kwargs):
         super().__init__(*args, **kwargs)
-        self.api_key = api_key
+        self.api_key = api_key or settings.GOOGLE_API_KEY or settings.API_KEY
         self.user_api_key = user_api_key
+        
         self.client = genai.Client(api_key=self.api_key)
         self.storage = StorageCreator.get_storage()
 
@@ -47,21 +48,19 @@ class GoogleLLM(BaseLLM):
         """
         if not attachments:
             return messages
-
         prepared_messages = messages.copy()
 
         # Find the user message to attach files to the last one
+
         user_message_index = None
         for i in range(len(prepared_messages) - 1, -1, -1):
             if prepared_messages[i].get("role") == "user":
                 user_message_index = i
                 break
-
         if user_message_index is None:
             user_message = {"role": "user", "content": []}
             prepared_messages.append(user_message)
             user_message_index = len(prepared_messages) - 1
-
         if isinstance(prepared_messages[user_message_index].get("content"), str):
             text_content = prepared_messages[user_message_index]["content"]
             prepared_messages[user_message_index]["content"] = [
@@ -69,7 +68,6 @@ class GoogleLLM(BaseLLM):
             ]
         elif not isinstance(prepared_messages[user_message_index].get("content"), list):
             prepared_messages[user_message_index]["content"] = []
-
         files = []
         for attachment in attachments:
             mime_type = attachment.get("mime_type")
@@ -92,11 +90,9 @@ class GoogleLLM(BaseLLM):
                                 "text": f"[File could not be processed: {attachment.get('path', 'unknown')}]",
                             }
                         )
-
         if files:
             logging.info(f"GoogleLLM: Adding {len(files)} files to message")
             prepared_messages[user_message_index]["content"].append({"files": files})
-
         return prepared_messages
 
     def _upload_file_to_google(self, attachment):
@@ -111,14 +107,11 @@ class GoogleLLM(BaseLLM):
         """
         if "google_file_uri" in attachment:
             return attachment["google_file_uri"]
-
         file_path = attachment.get("path")
         if not file_path:
             raise ValueError("No file path provided in attachment")
-
         if not self.storage.file_exists(file_path):
             raise FileNotFoundError(f"File not found: {file_path}")
-
         try:
             file_uri = self.storage.process_file(
                 file_path,
@@ -136,7 +129,6 @@ class GoogleLLM(BaseLLM):
                 attachments_collection.update_one(
                     {"_id": attachment["_id"]}, {"$set": {"google_file_uri": file_uri}}
                 )
-
             return file_uri
         except Exception as e:
             logging.error(f"Error uploading file to Google AI: {e}", exc_info=True)
@@ -153,7 +145,6 @@ class GoogleLLM(BaseLLM):
                 role = "model"
             elif role == "tool":
                 role = "model"
-
             parts = []
             if role and content is not None:
                 if isinstance(content, str):
@@ -163,10 +154,15 @@ class GoogleLLM(BaseLLM):
                         if "text" in item:
                             parts.append(types.Part.from_text(text=item["text"]))
                         elif "function_call" in item:
+                            # Remove null values from args to avoid API errors
+
+                            cleaned_args = self._remove_null_values(
+                                item["function_call"]["args"]
+                            )
                             parts.append(
                                 types.Part.from_function_call(
                                     name=item["function_call"]["name"],
-                                    args=item["function_call"]["args"],
+                                    args=cleaned_args,
                                 )
                             )
                         elif "function_response" in item:
@@ -190,10 +186,8 @@ class GoogleLLM(BaseLLM):
                             )
                 else:
                     raise ValueError(f"Unexpected content type: {type(content)}")
-
                 if parts:
                     cleaned_messages.append(types.Content(role=role, parts=parts))
-
         return cleaned_messages
 
     def _clean_schema(self, schema_obj):
@@ -229,8 +223,8 @@ class GoogleLLM(BaseLLM):
                 cleaned[key] = [self._clean_schema(item) for item in value]
             else:
                 cleaned[key] = value
-
         # Validate that required properties actually exist in properties
+
         if "required" in cleaned and "properties" in cleaned:
             valid_required = []
             properties_keys = set(cleaned["properties"].keys())
@@ -243,7 +237,6 @@ class GoogleLLM(BaseLLM):
                 cleaned.pop("required", None)
         elif "required" in cleaned and "properties" not in cleaned:
             cleaned.pop("required", None)
-
         return cleaned
 
     def _clean_tools_format(self, tools_list):
@@ -259,7 +252,6 @@ class GoogleLLM(BaseLLM):
                     cleaned_properties = {}
                     for k, v in properties.items():
                         cleaned_properties[k] = self._clean_schema(v)
-
                     genai_function = dict(
                         name=function["name"],
                         description=function["description"],
@@ -278,10 +270,8 @@ class GoogleLLM(BaseLLM):
                         name=function["name"],
                         description=function["description"],
                     )
-
                 genai_tool = types.Tool(function_declarations=[genai_function])
                 genai_tools.append(genai_tool)
-
         return genai_tools
 
     def _raw_gen(
@@ -303,16 +293,14 @@ class GoogleLLM(BaseLLM):
         if messages[0].role == "system":
             config.system_instruction = messages[0].parts[0].text
             messages = messages[1:]
-
         if tools:
             cleaned_tools = self._clean_tools_format(tools)
             config.tools = cleaned_tools
-
         # Add response schema for structured output if provided
+
         if response_schema:
             config.response_schema = response_schema
             config.response_mime_type = "application/json"
-
         response = client.models.generate_content(
             model=model,
             contents=messages,
@@ -343,17 +331,16 @@ class GoogleLLM(BaseLLM):
         if messages[0].role == "system":
             config.system_instruction = messages[0].parts[0].text
             messages = messages[1:]
-
         if tools:
             cleaned_tools = self._clean_tools_format(tools)
             config.tools = cleaned_tools
-
         # Add response schema for structured output if provided
+
         if response_schema:
             config.response_schema = response_schema
             config.response_mime_type = "application/json"
-
         # Check if we have both tools and file attachments
+
         has_attachments = False
         for message in messages:
             for part in message.parts:
@@ -362,7 +349,6 @@ class GoogleLLM(BaseLLM):
                     break
             if has_attachments:
                 break
-
         logging.info(
             f"GoogleLLM: Starting stream generation. Model: {model}, Messages: {json.dumps(messages, default=str)}, Has attachments: {has_attachments}"
         )
@@ -386,7 +372,7 @@ class GoogleLLM(BaseLLM):
                 elif hasattr(chunk, "text"):
                     yield chunk.text
         finally:
-            if hasattr(response, 'close'):
+            if hasattr(response, "close"):
                 response.close()
 
     def _supports_tools(self):
@@ -401,7 +387,6 @@ class GoogleLLM(BaseLLM):
         """Convert JSON schema to Google AI structured output format."""
         if not json_schema:
             return None
-
         type_map = {
             "object": "OBJECT",
             "array": "ARRAY",
@@ -414,12 +399,10 @@ class GoogleLLM(BaseLLM):
         def convert(schema):
             if not isinstance(schema, dict):
                 return schema
-
             result = {}
             schema_type = schema.get("type")
             if schema_type:
                 result["type"] = type_map.get(schema_type.lower(), schema_type.upper())
-
             for key in [
                 "description",
                 "nullable",
@@ -431,7 +414,6 @@ class GoogleLLM(BaseLLM):
             ]:
                 if key in schema:
                     result[key] = schema[key]
-
             if "format" in schema:
                 format_value = schema["format"]
                 if schema_type == "string":
@@ -441,21 +423,17 @@ class GoogleLLM(BaseLLM):
                         result["format"] = format_value
                 else:
                     result["format"] = format_value
-
             if "properties" in schema:
                 result["properties"] = {
                     k: convert(v) for k, v in schema["properties"].items()
                 }
                 if "propertyOrdering" not in result and result.get("type") == "OBJECT":
                     result["propertyOrdering"] = list(result["properties"].keys())
-
             if "items" in schema:
                 result["items"] = convert(schema["items"])
-
             for field in ["anyOf", "oneOf", "allOf"]:
                 if field in schema:
                     result[field] = [convert(s) for s in schema[field]]
-
             return result
 
         try:

application/llm/groq.py

@@ -1,13 +1,18 @@
-from application.llm.base import BaseLLM
 from openai import OpenAI
 
+from application.core.settings import settings
+from application.llm.base import BaseLLM
+
 
 class GroqLLM(BaseLLM):
     def __init__(self, api_key=None, user_api_key=None, *args, **kwargs):
+
         super().__init__(*args, **kwargs)
-        self.client = OpenAI(api_key=api_key, base_url="https://api.groq.com/openai/v1")
-        self.api_key = api_key
+        self.api_key = api_key or settings.GROQ_API_KEY or settings.API_KEY
         self.user_api_key = user_api_key
+        self.client = OpenAI(
+            api_key=self.api_key, base_url="https://api.groq.com/openai/v1"
+        )
 
     def _raw_gen(self, baseself, model, messages, stream=False, tools=None, **kwargs):
         if tools:

application/llm/handlers/base.py

@@ -282,7 +282,7 @@ class LLMHandler(ABC):
                     messages = e.value
                     break
             response = agent.llm.gen(
-                model=agent.gpt_model, messages=messages, tools=agent.tools
+                model=agent.model_id, messages=messages, tools=agent.tools
             )
             parsed = self.parse_response(response)
             self.llm_calls.append(build_stack_data(agent.llm))
@@ -337,7 +337,7 @@ class LLMHandler(ABC):
                 tool_calls = {}
 
                 response = agent.llm.gen_stream(
-                    model=agent.gpt_model, messages=messages, tools=agent.tools
+                    model=agent.model_id, messages=messages, tools=agent.tools
                 )
                 self.llm_calls.append(build_stack_data(agent.llm))
 

application/llm/llm_creator.py

@@ -1,13 +1,17 @@
-from application.llm.groq import GroqLLM
-from application.llm.openai import OpenAILLM, AzureOpenAILLM
-from application.llm.sagemaker import SagemakerAPILLM
-from application.llm.huggingface import HuggingFaceLLM
-from application.llm.llama_cpp import LlamaCpp
+import logging
+
 from application.llm.anthropic import AnthropicLLM
 from application.llm.docsgpt_provider import DocsGPTAPILLM
-from application.llm.premai import PremAILLM
 from application.llm.google_ai import GoogleLLM
+from application.llm.groq import GroqLLM
+from application.llm.huggingface import HuggingFaceLLM
+from application.llm.llama_cpp import LlamaCpp
 from application.llm.novita import NovitaLLM
+from application.llm.openai import AzureOpenAILLM, OpenAILLM
+from application.llm.premai import PremAILLM
+from application.llm.sagemaker import SagemakerAPILLM
+
+logger = logging.getLogger(__name__)
 
 
 class LLMCreator:
@@ -26,10 +30,26 @@ class LLMCreator:
     }
 
     @classmethod
-    def create_llm(cls, type, api_key, user_api_key, decoded_token, *args, **kwargs):
+    def create_llm(
+        cls, type, api_key, user_api_key, decoded_token, model_id=None, *args, **kwargs
+    ):
+        from application.core.model_utils import get_base_url_for_model
+
         llm_class = cls.llms.get(type.lower())
         if not llm_class:
             raise ValueError(f"No LLM class found for type {type}")
+
+        # Extract base_url from model configuration if model_id is provided
+        base_url = None
+        if model_id:
+            base_url = get_base_url_for_model(model_id)
+
         return llm_class(
-            api_key, user_api_key, decoded_token=decoded_token, *args, **kwargs
+            api_key,
+            user_api_key,
+            decoded_token=decoded_token,
+            model_id=model_id,
+            base_url=base_url,
+            *args,
+            **kwargs,
         )

application/llm/openai.py

@@ -2,6 +2,8 @@ import base64
 import json
 import logging
 
+from openai import OpenAI
+
 from application.core.settings import settings
 from application.llm.base import BaseLLM
 from application.storage.storage_creator import StorageCreator
@@ -9,20 +11,25 @@ from application.storage.storage_creator import StorageCreator
 
 class OpenAILLM(BaseLLM):
 
-    def __init__(self, api_key=None, user_api_key=None, *args, **kwargs):
-        from openai import OpenAI
+    def __init__(self, api_key=None, user_api_key=None, base_url=None, *args, **kwargs):
 
         super().__init__(*args, **kwargs)
-        if (
+        self.api_key = api_key or settings.OPENAI_API_KEY or settings.API_KEY
+        self.user_api_key = user_api_key
+
+        # Priority: 1) Parameter base_url, 2) Settings OPENAI_BASE_URL, 3) Default
+        effective_base_url = None
+        if base_url and isinstance(base_url, str) and base_url.strip():
+            effective_base_url = base_url
+        elif (
             isinstance(settings.OPENAI_BASE_URL, str)
             and settings.OPENAI_BASE_URL.strip()
         ):
-            self.client = OpenAI(api_key=api_key, base_url=settings.OPENAI_BASE_URL)
+            effective_base_url = settings.OPENAI_BASE_URL
         else:
-            DEFAULT_OPENAI_API_BASE = "https://api.openai.com/v1"
-            self.client = OpenAI(api_key=api_key, base_url=DEFAULT_OPENAI_API_BASE)
-        self.api_key = api_key
-        self.user_api_key = user_api_key
+            effective_base_url = "https://api.openai.com/v1"
+
+        self.client = OpenAI(api_key=self.api_key, base_url=effective_base_url)
         self.storage = StorageCreator.get_storage()
 
     def _clean_messages_openai(self, messages):
@@ -33,7 +40,6 @@ class OpenAILLM(BaseLLM):
 
             if role == "model":
                 role = "assistant"
-
             if role and content is not None:
                 if isinstance(content, str):
                     cleaned_messages.append({"role": role, "content": content})
@@ -44,14 +50,15 @@ class OpenAILLM(BaseLLM):
                                 {"role": role, "content": item["text"]}
                             )
                         elif "function_call" in item:
+                            cleaned_args = self._remove_null_values(
+                                item["function_call"]["args"]
+                            )
                             tool_call = {
                                 "id": item["function_call"]["call_id"],
                                 "type": "function",
                                 "function": {
                                     "name": item["function_call"]["name"],
-                                    "arguments": json.dumps(
-                                        item["function_call"]["args"]
-                                    ),
+                                    "arguments": json.dumps(cleaned_args),
                                 },
                             }
                             cleaned_messages.append(
@@ -106,7 +113,6 @@ class OpenAILLM(BaseLLM):
                             )
                 else:
                     raise ValueError(f"Unexpected content type: {type(content)}")
-
         return cleaned_messages
 
     def _raw_gen(
@@ -131,10 +137,8 @@ class OpenAILLM(BaseLLM):
 
         if tools:
             request_params["tools"] = tools
-
         if response_format:
             request_params["response_format"] = response_format
-
         response = self.client.chat.completions.create(**request_params)
 
         if tools:
@@ -164,10 +168,8 @@ class OpenAILLM(BaseLLM):
 
         if tools:
             request_params["tools"] = tools
-
         if response_format:
             request_params["response_format"] = response_format
-
         response = self.client.chat.completions.create(**request_params)
 
         try:
@@ -181,7 +183,7 @@ class OpenAILLM(BaseLLM):
                 elif len(line.choices) > 0:
                     yield line.choices[0]
         finally:
-            if hasattr(response, 'close'):
+            if hasattr(response, "close"):
                 response.close()
 
     def _supports_tools(self):
@@ -193,7 +195,6 @@ class OpenAILLM(BaseLLM):
     def prepare_structured_output_format(self, json_schema):
         if not json_schema:
             return None
-
         try:
 
             def add_additional_properties_false(schema_obj):
@@ -203,11 +204,11 @@ class OpenAILLM(BaseLLM):
                     if schema_copy.get("type") == "object":
                         schema_copy["additionalProperties"] = False
                         # Ensure 'required' includes all properties for OpenAI strict mode
+
                         if "properties" in schema_copy:
                             schema_copy["required"] = list(
                                 schema_copy["properties"].keys()
                             )
-
                     for key, value in schema_copy.items():
                         if key == "properties" and isinstance(value, dict):
                             schema_copy[key] = {
@@ -223,7 +224,6 @@ class OpenAILLM(BaseLLM):
                                 add_additional_properties_false(sub_schema)
                                 for sub_schema in value
                             ]
-
                     return schema_copy
                 return schema_obj
 
@@ -242,7 +242,6 @@ class OpenAILLM(BaseLLM):
             }
 
             return result
-
         except Exception as e:
             logging.error(f"Error preparing structured output format: {e}")
             return None
@@ -276,21 +275,19 @@ class OpenAILLM(BaseLLM):
         """
         if not attachments:
             return messages
-
         prepared_messages = messages.copy()
 
         # Find the user message to attach file_id to the last one
+
         user_message_index = None
         for i in range(len(prepared_messages) - 1, -1, -1):
             if prepared_messages[i].get("role") == "user":
                 user_message_index = i
                 break
-
         if user_message_index is None:
             user_message = {"role": "user", "content": []}
             prepared_messages.append(user_message)
             user_message_index = len(prepared_messages) - 1
-
         if isinstance(prepared_messages[user_message_index].get("content"), str):
             text_content = prepared_messages[user_message_index]["content"]
             prepared_messages[user_message_index]["content"] = [
@@ -298,7 +295,6 @@ class OpenAILLM(BaseLLM):
             ]
         elif not isinstance(prepared_messages[user_message_index].get("content"), list):
             prepared_messages[user_message_index]["content"] = []
-
         for attachment in attachments:
             mime_type = attachment.get("mime_type")
 
@@ -325,6 +321,7 @@ class OpenAILLM(BaseLLM):
                             }
                         )
             # Handle PDFs using the file API
+
             elif mime_type == "application/pdf":
                 try:
                     file_id = self._upload_file_to_openai(attachment)
@@ -340,7 +337,6 @@ class OpenAILLM(BaseLLM):
                                 "text": f"File content:\n\n{attachment['content']}",
                             }
                         )
-
         return prepared_messages
 
     def _get_base64_image(self, attachment):
@@ -356,7 +352,6 @@ class OpenAILLM(BaseLLM):
         file_path = attachment.get("path")
         if not file_path:
             raise ValueError("No file path provided in attachment")
-
         try:
             with self.storage.get_file(file_path) as image_file:
                 return base64.b64encode(image_file.read()).decode("utf-8")
@@ -380,12 +375,10 @@ class OpenAILLM(BaseLLM):
 
         if "openai_file_id" in attachment:
             return attachment["openai_file_id"]
-
         file_path = attachment.get("path")
 
         if not self.storage.file_exists(file_path):
             raise FileNotFoundError(f"File not found: {file_path}")
-
         try:
             file_id = self.storage.process_file(
                 file_path,
@@ -403,7 +396,6 @@ class OpenAILLM(BaseLLM):
                 attachments_collection.update_one(
                     {"_id": attachment["_id"]}, {"$set": {"openai_file_id": file_id}}
                 )
-
             return file_id
         except Exception as e:
             logging.error(f"Error uploading file to OpenAI: {e}", exc_info=True)

application/parser/embedding_pipeline.py

@@ -1,5 +1,6 @@
 import os
 import logging
+from typing import List, Any
 from retry import retry
 from tqdm import tqdm
 from application.core.settings import settings
@@ -22,13 +23,16 @@ def sanitize_content(content: str) -> str:
 
 
 @retry(tries=10, delay=60)
-def add_text_to_store_with_retry(store, doc, source_id):
-    """
-    Add a document's text and metadata to the vector store with retry logic.
+def add_text_to_store_with_retry(store: Any, doc: Any, source_id: str) -> None:
+    """Add a document's text and metadata to the vector store with retry logic.
+    
     Args:
         store: The vector store object.
         doc: The document to be added.
         source_id: Unique identifier for the source.
+        
+    Raises:
+        Exception: If document addition fails after all retry attempts.
     """
     try:
         # Sanitize content to remove NUL characters that cause ingestion failures
@@ -41,18 +45,21 @@ def add_text_to_store_with_retry(store, doc, source_id):
         raise
 
 
-def embed_and_store_documents(docs, folder_name, source_id, task_status):
-    """
-    Embeds documents and stores them in a vector store.
+def embed_and_store_documents(docs: List[Any], folder_name: str, source_id: str, task_status: Any) -> None:
+    """Embeds documents and stores them in a vector store.
 
     Args:
-        docs (list): List of documents to be embedded and stored.
-        folder_name (str): Directory to save the vector store.
-        source_id (str): Unique identifier for the source.
+        docs: List of documents to be embedded and stored.
+        folder_name: Directory to save the vector store.
+        source_id: Unique identifier for the source.
         task_status: Task state manager for progress updates.
 
     Returns:
         None
+        
+    Raises:
+        OSError: If unable to create folder or save vector store.
+        Exception: If vector store creation or document embedding fails.
     """
     # Ensure the folder exists
     if not os.path.exists(folder_name):
@@ -95,10 +102,21 @@ def embed_and_store_documents(docs, folder_name, source_id, task_status):
         except Exception as e:
             logging.error(f"Error embedding document {idx}: {e}", exc_info=True)
             logging.info(f"Saving progress at document {idx} out of {total_docs}")
-            store.save_local(folder_name)
+            try:
+                store.save_local(folder_name)
+                logging.info("Progress saved successfully")
+            except Exception as save_error:
+                logging.error(f"CRITICAL: Failed to save progress: {save_error}", exc_info=True)
+                # Continue without breaking to attempt final save
             break
 
     # Save the vector store
     if settings.VECTOR_STORE == "faiss":
-        store.save_local(folder_name)
-    logging.info("Vector store saved successfully.")
+        try:
+            store.save_local(folder_name)
+            logging.info("Vector store saved successfully.")
+        except Exception as e:
+            logging.error(f"CRITICAL: Failed to save final vector store: {e}", exc_info=True)
+            raise OSError(f"Unable to save vector store to {folder_name}: {e}") from e
+    else:
+        logging.info("Vector store saved successfully.")

application/requirements.txt

@@ -10,6 +10,7 @@ ebooklib==0.18
 escodegen==1.0.11
 esprima==4.0.1
 esutils==1.0.1
+elevenlabs==2.17.0
 Flask==3.1.1
 faiss-cpu==1.9.0.post1
 fastmcp==2.11.0

application/retriever/base.py

@@ -8,7 +8,3 @@ class BaseRetriever(ABC):
     @abstractmethod
     def search(self, *args, **kwargs):
         pass
-
-    @abstractmethod
-    def get_params(self):
-        pass

application/retriever/classic_rag.py

@@ -4,7 +4,7 @@ import os
 from application.core.settings import settings
 from application.llm.llm_creator import LLMCreator
 from application.retriever.base import BaseRetriever
-
+from application.utils import num_tokens_from_string
 from application.vectorstore.vector_creator import VectorCreator
 
 
@@ -15,14 +15,13 @@ class ClassicRAG(BaseRetriever):
         chat_history=None,
         prompt="",
         chunks=2,
-        token_limit=150,
-        gpt_model="docsgpt",
+        doc_token_limit=50000,
+        model_id="docsgpt-local",
         user_api_key=None,
         llm_name=settings.LLM_PROVIDER,
         api_key=settings.API_KEY,
         decoded_token=None,
     ):
-        """Initialize ClassicRAG retriever with vectorstore sources and LLM configuration"""
         self.original_question = source.get("question", "")
         self.chat_history = chat_history if chat_history is not None else []
         self.prompt = prompt
@@ -41,17 +40,8 @@ class ClassicRAG(BaseRetriever):
             f"ClassicRAG initialized with chunks={self.chunks}, user_api_key={user_identifier}, "
             f"sources={'active_docs' in source and source['active_docs'] is not None}"
         )
-        self.gpt_model = gpt_model
-        self.token_limit = (
-            token_limit
-            if token_limit
-            < settings.LLM_TOKEN_LIMITS.get(
-                self.gpt_model, settings.DEFAULT_MAX_HISTORY
-            )
-            else settings.LLM_TOKEN_LIMITS.get(
-                self.gpt_model, settings.DEFAULT_MAX_HISTORY
-            )
-        )
+        self.model_id = model_id
+        self.doc_token_limit = doc_token_limit
         self.user_api_key = user_api_key
         self.llm_name = llm_name
         self.api_key = api_key
@@ -110,7 +100,7 @@ class ClassicRAG(BaseRetriever):
         ]
 
         try:
-            rephrased_query = self.llm.gen(model=self.gpt_model, messages=messages)
+            rephrased_query = self.llm.gen(model=self.model_id, messages=messages)
             print(f"Rephrased query: {rephrased_query}")
             return rephrased_query if rephrased_query else self.original_question
         except Exception as e:
@@ -118,21 +108,17 @@ class ClassicRAG(BaseRetriever):
             return self.original_question
 
     def _get_data(self):
-        """Retrieve relevant documents from configured vectorstores"""
         if self.chunks == 0 or not self.vectorstores:
             logging.info(
                 f"ClassicRAG._get_data: Skipping retrieval - chunks={self.chunks}, "
                 f"vectorstores_count={len(self.vectorstores) if self.vectorstores else 0}"
             )
             return []
+
         all_docs = []
         chunks_per_source = max(1, self.chunks // len(self.vectorstores))
-
-        logging.info(
-            f"ClassicRAG._get_data: Starting retrieval with chunks={self.chunks}, "
-            f"vectorstores={self.vectorstores}, chunks_per_source={chunks_per_source}, "
-            f"query='{self.question[:50]}...'"
-        )
+        token_budget = max(int(self.doc_token_limit * 0.9), 100)
+        cumulative_tokens = 0
 
         for vectorstore_id in self.vectorstores:
             if vectorstore_id:
@@ -140,15 +126,21 @@ class ClassicRAG(BaseRetriever):
                     docsearch = VectorCreator.create_vectorstore(
                         settings.VECTOR_STORE, vectorstore_id, settings.EMBEDDINGS_KEY
                     )
-                    docs_temp = docsearch.search(self.question, k=chunks_per_source)
+                    docs_temp = docsearch.search(
+                        self.question, k=max(chunks_per_source * 2, 20)
+                    )
 
                     for doc in docs_temp:
+                        if cumulative_tokens >= token_budget:
+                            break
+
                         if hasattr(doc, "page_content") and hasattr(doc, "metadata"):
                             page_content = doc.page_content
                             metadata = doc.metadata
                         else:
                             page_content = doc.get("text", doc.get("page_content", ""))
                             metadata = doc.get("metadata", {})
+
                         title = metadata.get(
                             "title", metadata.get("post_title", page_content)
                         )
@@ -168,23 +160,35 @@ class ClassicRAG(BaseRetriever):
                         if not filename:
                             filename = title
                         source_path = metadata.get("source") or vectorstore_id
-                        all_docs.append(
-                            {
-                                "title": title,
-                                "text": page_content,
-                                "source": source_path,
-                                "filename": filename,
-                            }
-                        )
+
+                        doc_text_with_header = f"{filename}\n{page_content}"
+                        doc_tokens = num_tokens_from_string(doc_text_with_header)
+
+                        if cumulative_tokens + doc_tokens < token_budget:
+                            all_docs.append(
+                                {
+                                    "title": title,
+                                    "text": page_content,
+                                    "source": source_path,
+                                    "filename": filename,
+                                }
+                            )
+                            cumulative_tokens += doc_tokens
+
+                    if cumulative_tokens >= token_budget:
+                        break
+
                 except Exception as e:
                     logging.error(
                         f"Error searching vectorstore {vectorstore_id}: {e}",
                         exc_info=True,
                     )
                     continue
+
         logging.info(
             f"ClassicRAG._get_data: Retrieval complete - retrieved {len(all_docs)} documents "
-            f"(requested chunks={self.chunks}, chunks_per_source={chunks_per_source})"
+            f"(requested chunks={self.chunks}, chunks_per_source={chunks_per_source}, "
+            f"cumulative_tokens={cumulative_tokens}/{token_budget})"
         )
         return all_docs
 
@@ -194,15 +198,3 @@ class ClassicRAG(BaseRetriever):
             self.original_question = query
             self.question = self._rephrase_query()
         return self._get_data()
-
-    def get_params(self):
-        """Return current retriever configuration parameters"""
-        return {
-            "question": self.original_question,
-            "rephrased_question": self.question,
-            "sources": self.vectorstores,
-            "chunks": self.chunks,
-            "token_limit": self.token_limit,
-            "gpt_model": self.gpt_model,
-            "user_api_key": self.user_api_key,
-        }

application/templates/namespaces.py

@@ -0,0 +1,190 @@
+import logging
+import uuid
+from abc import ABC, abstractmethod
+from datetime import datetime, timezone
+from typing import Any, Dict, Optional
+
+logger = logging.getLogger(__name__)
+
+
+class NamespaceBuilder(ABC):
+    """Base class for building template context namespaces"""
+
+    @abstractmethod
+    def build(self, **kwargs) -> Dict[str, Any]:
+        """Build namespace context dictionary"""
+        pass
+
+    @property
+    @abstractmethod
+    def namespace_name(self) -> str:
+        """Name of this namespace for template access"""
+        pass
+
+
+class SystemNamespace(NamespaceBuilder):
+    """System metadata namespace: {{ system.* }}"""
+
+    @property
+    def namespace_name(self) -> str:
+        return "system"
+
+    def build(
+        self, request_id: Optional[str] = None, user_id: Optional[str] = None, **kwargs
+    ) -> Dict[str, Any]:
+        """
+        Build system context with metadata.
+
+        Args:
+            request_id: Unique request identifier
+            user_id: Current user identifier
+
+        Returns:
+            Dictionary with system variables
+        """
+        now = datetime.now(timezone.utc)
+
+        return {
+            "date": now.strftime("%Y-%m-%d"),
+            "time": now.strftime("%H:%M:%S"),
+            "timestamp": now.isoformat(),
+            "request_id": request_id or str(uuid.uuid4()),
+            "user_id": user_id,
+        }
+
+
+class PassthroughNamespace(NamespaceBuilder):
+    """Request parameters namespace: {{ passthrough.* }}"""
+
+    @property
+    def namespace_name(self) -> str:
+        return "passthrough"
+
+    def build(
+        self, passthrough_data: Optional[Dict[str, Any]] = None, **kwargs
+    ) -> Dict[str, Any]:
+        """
+        Build passthrough context from request parameters.
+
+        Args:
+            passthrough_data: Dictionary of parameters from web request
+
+        Returns:
+            Dictionary with passthrough variables
+        """
+        if not passthrough_data:
+            return {}
+        safe_data = {}
+        for key, value in passthrough_data.items():
+            if isinstance(value, (str, int, float, bool, type(None))):
+                safe_data[key] = value
+            else:
+                logger.warning(
+                    f"Skipping non-serializable passthrough value for key '{key}': {type(value)}"
+                )
+        return safe_data
+
+
+class SourceNamespace(NamespaceBuilder):
+    """RAG source documents namespace: {{ source.* }}"""
+
+    @property
+    def namespace_name(self) -> str:
+        return "source"
+
+    def build(
+        self, docs: Optional[list] = None, docs_together: Optional[str] = None, **kwargs
+    ) -> Dict[str, Any]:
+        """
+        Build source context from RAG retrieval results.
+
+        Args:
+            docs: List of retrieved documents
+            docs_together: Concatenated document content (for backward compatibility)
+
+        Returns:
+            Dictionary with source variables
+        """
+        context = {}
+
+        if docs:
+            context["documents"] = docs
+            context["count"] = len(docs)
+        if docs_together:
+            context["docs_together"] = docs_together  # Add docs_together for custom templates
+            context["content"] = docs_together
+            context["summaries"] = docs_together
+        return context
+
+
+class ToolsNamespace(NamespaceBuilder):
+    """Pre-executed tools namespace: {{ tools.* }}"""
+
+    @property
+    def namespace_name(self) -> str:
+        return "tools"
+
+    def build(
+        self, tools_data: Optional[Dict[str, Any]] = None, **kwargs
+    ) -> Dict[str, Any]:
+        """
+        Build tools context with pre-executed tool results.
+
+        Args:
+            tools_data: Dictionary of pre-fetched tool results organized by tool name
+                       e.g., {"memory": {"notes": "content", "tasks": "list"}}
+
+        Returns:
+            Dictionary with tool results organized by tool name
+        """
+        if not tools_data:
+            return {}
+
+        safe_data = {}
+        for tool_name, tool_result in tools_data.items():
+            if isinstance(tool_result, (str, dict, list, int, float, bool, type(None))):
+                safe_data[tool_name] = tool_result
+            else:
+                logger.warning(
+                    f"Skipping non-serializable tool result for '{tool_name}': {type(tool_result)}"
+                )
+        return safe_data
+
+
+class NamespaceManager:
+    """Manages all namespace builders and context assembly"""
+
+    def __init__(self):
+        self._builders = {
+            "system": SystemNamespace(),
+            "passthrough": PassthroughNamespace(),
+            "source": SourceNamespace(),
+            "tools": ToolsNamespace(),
+        }
+
+    def build_context(self, **kwargs) -> Dict[str, Any]:
+        """
+        Build complete template context from all namespaces.
+
+        Args:
+            **kwargs: Parameters to pass to namespace builders
+
+        Returns:
+            Complete context dictionary for template rendering
+        """
+        context = {}
+
+        for namespace_name, builder in self._builders.items():
+            try:
+                namespace_context = builder.build(**kwargs)
+                # Always include namespace, even if empty, to prevent undefined errors
+                context[namespace_name] = namespace_context if namespace_context else {}
+            except Exception as e:
+                logger.error(f"Failed to build {namespace_name} namespace: {str(e)}")
+                # Include empty namespace on error to prevent template failures
+                context[namespace_name] = {}
+        return context
+
+    def get_builder(self, namespace_name: str) -> Optional[NamespaceBuilder]:
+        """Get specific namespace builder"""
+        return self._builders.get(namespace_name)

application/templates/template_engine.py

@@ -0,0 +1,161 @@
+import logging
+from typing import Any, Dict, List, Optional, Set
+
+from jinja2 import (
+    ChainableUndefined,
+    Environment,
+    nodes,
+    select_autoescape,
+    TemplateSyntaxError,
+)
+from jinja2.exceptions import UndefinedError
+
+logger = logging.getLogger(__name__)
+
+
+class TemplateRenderError(Exception):
+    """Raised when template rendering fails"""
+
+    pass
+
+
+class TemplateEngine:
+    """Jinja2-based template engine for dynamic prompt rendering"""
+
+    def __init__(self):
+        self._env = Environment(
+            undefined=ChainableUndefined,
+            trim_blocks=True,
+            lstrip_blocks=True,
+            autoescape=select_autoescape(default_for_string=True, default=True),
+        )
+
+    def render(self, template_content: str, context: Dict[str, Any]) -> str:
+        """
+        Render template with provided context.
+
+        Args:
+            template_content: Raw template string with Jinja2 syntax
+            context: Dictionary of variables to inject into template
+
+        Returns:
+            Rendered template string
+
+        Raises:
+            TemplateRenderError: If template syntax is invalid or variables undefined
+        """
+        if not template_content:
+            return ""
+        try:
+            template = self._env.from_string(template_content)
+            return template.render(**context)
+        except TemplateSyntaxError as e:
+            error_msg = f"Template syntax error at line {e.lineno}: {e.message}"
+            logger.error(error_msg)
+            raise TemplateRenderError(error_msg) from e
+        except UndefinedError as e:
+            error_msg = f"Undefined variable in template: {e.message}"
+            logger.error(error_msg)
+            raise TemplateRenderError(error_msg) from e
+        except Exception as e:
+            error_msg = f"Template rendering failed: {str(e)}"
+            logger.error(error_msg)
+            raise TemplateRenderError(error_msg) from e
+
+    def validate_template(self, template_content: str) -> bool:
+        """
+        Validate template syntax without rendering.
+
+        Args:
+            template_content: Template string to validate
+
+        Returns:
+            True if template is syntactically valid
+        """
+        if not template_content:
+            return True
+        try:
+            self._env.from_string(template_content)
+            return True
+        except TemplateSyntaxError as e:
+            logger.debug(f"Template syntax invalid at line {e.lineno}: {e.message}")
+            return False
+        except Exception as e:
+            logger.debug(f"Template validation error: {type(e).__name__}: {str(e)}")
+            return False
+
+    def extract_variables(self, template_content: str) -> Set[str]:
+        """
+        Extract all variable names from template.
+
+        Args:
+            template_content: Template string to analyze
+
+        Returns:
+            Set of variable names found in template
+        """
+        if not template_content:
+            return set()
+        try:
+            ast = self._env.parse(template_content)
+            return set(self._env.get_template_module(ast).make_module().keys())
+        except TemplateSyntaxError as e:
+            logger.debug(f"Cannot extract variables - syntax error at line {e.lineno}")
+            return set()
+        except Exception as e:
+            logger.debug(f"Cannot extract variables: {type(e).__name__}")
+            return set()
+
+    def extract_tool_usages(
+        self, template_content: str
+    ) -> Dict[str, Set[Optional[str]]]:
+        """Extract tool and action references from a template"""
+        if not template_content:
+            return {}
+        try:
+            ast = self._env.parse(template_content)
+        except TemplateSyntaxError as e:
+            logger.debug(f"extract_tool_usages - syntax error at line {e.lineno}")
+            return {}
+        except Exception as e:
+            logger.debug(f"extract_tool_usages - parse error: {type(e).__name__}")
+            return {}
+
+        usages: Dict[str, Set[Optional[str]]] = {}
+
+        def record(path: List[str]) -> None:
+            if not path:
+                return
+            tool_name = path[0]
+            action_name = path[1] if len(path) > 1 else None
+            if not tool_name:
+                return
+            tool_entry = usages.setdefault(tool_name, set())
+            tool_entry.add(action_name)
+
+        for node in ast.find_all(nodes.Getattr):
+            path = []
+            current = node
+            while isinstance(current, nodes.Getattr):
+                path.append(current.attr)
+                current = current.node
+            if isinstance(current, nodes.Name) and current.name == "tools":
+                path.reverse()
+                record(path)
+
+        for node in ast.find_all(nodes.Getitem):
+            path = []
+            current = node
+            while isinstance(current, nodes.Getitem):
+                key = current.arg
+                if isinstance(key, nodes.Const) and isinstance(key.value, str):
+                    path.append(key.value)
+                else:
+                    path = []
+                    break
+                current = current.node
+            if path and isinstance(current, nodes.Name) and current.name == "tools":
+                path.reverse()
+                record(path)
+
+        return usages

application/tts/elevenlabs.py

@@ -15,10 +15,11 @@ class ElevenlabsTTS(BaseTTS):
 
     def text_to_speech(self, text):
         lang = "en"
-        audio = self.client.generate(
+        audio = self.client.text_to_speech.convert(
+            voice_id="nPczCjzI2devNBz1zQrb",             
+            model_id="eleven_multilingual_v2",
             text=text,
-            model="eleven_multilingual_v2",
-            voice="Brian",
+            output_format="mp3_44100_128"
         )
         audio_data = BytesIO()
         for chunk in audio:

application/tts/tts_creator.py

@@ -0,0 +1,18 @@
+from application.tts.google_tts import GoogleTTS
+from application.tts.elevenlabs import ElevenlabsTTS
+from application.tts.base import BaseTTS
+
+
+
+class TTSCreator:
+    tts_providers = {
+        "google_tts": GoogleTTS,
+        "elevenlabs": ElevenlabsTTS,
+    }
+
+    @classmethod
+    def create_tts(cls, tts_type, *args, **kwargs)-> BaseTTS:
+        tts_class = cls.tts_providers.get(tts_type.lower())
+        if not tts_class:
+            raise ValueError(f"No tts class found for type {tts_type}")
+        return tts_class(*args, **kwargs)

application/utils.py

@@ -7,6 +7,8 @@ import tiktoken
 from flask import jsonify, make_response
 from werkzeug.utils import secure_filename
 
+from application.core.model_utils import get_token_limit
+
 from application.core.settings import settings
 
 
@@ -21,7 +23,7 @@ def get_encoding():
 
 
 def get_gpt_model() -> str:
-    """Get the appropriate GPT model based on provider"""
+    """Get GPT model based on provider"""
     model_map = {
         "openai": "gpt-4o-mini",
         "anthropic": "claude-2",
@@ -32,16 +34,7 @@ def get_gpt_model() -> str:
 
 
 def safe_filename(filename):
-    """
-    Creates a safe filename that preserves the original extension.
-    Uses secure_filename, but ensures a proper filename is returned even with non-Latin characters.
-
-    Args:
-        filename (str): The original filename
-
-    Returns:
-        str: A safe filename that can be used for storage
-    """
+    """Create safe filename, preserving extension. Handles non-Latin characters."""
     if not filename:
         return str(uuid.uuid4())
     _, extension = os.path.splitext(filename)
@@ -83,8 +76,23 @@ def count_tokens_docs(docs):
     return tokens
 
 
+def calculate_doc_token_budget(
+    model_id: str = "gpt-4o", history_token_limit: int = 2000
+) -> int:
+    total_context = get_token_limit(model_id)
+    reserved = sum(settings.RESERVED_TOKENS.values())
+    doc_budget = total_context - history_token_limit - reserved
+    return max(doc_budget, 1000)
+
+
+def get_missing_fields(data, required_fields):
+    """Check for missing required fields. Returns list of missing field names."""
+    return [field for field in required_fields if field not in data]
+
+
 def check_required_fields(data, required_fields):
-    missing_fields = [field for field in required_fields if field not in data]
+    """Validate required fields. Returns Flask 400 response if validation fails, None otherwise."""
+    missing_fields = get_missing_fields(data, required_fields)
     if missing_fields:
         return make_response(
             jsonify(
@@ -98,7 +106,8 @@ def check_required_fields(data, required_fields):
     return None
 
 
-def validate_required_fields(data, required_fields):
+def get_field_validation_errors(data, required_fields):
+    """Check for missing and empty fields. Returns dict with 'missing_fields' and 'empty_fields', or None."""
     missing_fields = []
     empty_fields = []
 
@@ -107,12 +116,24 @@ def validate_required_fields(data, required_fields):
             missing_fields.append(field)
         elif not data[field]:
             empty_fields.append(field)
-    errors = []
-    if missing_fields:
-        errors.append(f"Missing required fields: {', '.join(missing_fields)}")
-    if empty_fields:
-        errors.append(f"Empty values in required fields: {', '.join(empty_fields)}")
-    if errors:
+    if missing_fields or empty_fields:
+        return {"missing_fields": missing_fields, "empty_fields": empty_fields}
+    return None
+
+
+def validate_required_fields(data, required_fields):
+    """Validate required fields (must exist and be non-empty). Returns Flask 400 response if validation fails, None otherwise."""
+    errors_dict = get_field_validation_errors(data, required_fields)
+    if errors_dict:
+        errors = []
+        if errors_dict["missing_fields"]:
+            errors.append(
+                f"Missing required fields: {', '.join(errors_dict['missing_fields'])}"
+            )
+        if errors_dict["empty_fields"]:
+            errors.append(
+                f"Empty values in required fields: {', '.join(errors_dict['empty_fields'])}"
+            )
         return make_response(
             jsonify({"success": False, "message": " | ".join(errors)}), 400
         )
@@ -123,19 +144,13 @@ def get_hash(data):
     return hashlib.md5(data.encode(), usedforsecurity=False).hexdigest()
 
 
-def limit_chat_history(history, max_token_limit=None, gpt_model="docsgpt"):
-    """
-    Limits chat history based on token count.
-    Returns a list of messages that fit within the token limit.
-    """
-    from application.core.settings import settings
-
+def limit_chat_history(history, max_token_limit=None, model_id="docsgpt-local"):
+    """Limit chat history to fit within token limit."""
+    model_token_limit = get_token_limit(model_id)
     max_token_limit = (
         max_token_limit
-        if max_token_limit
-        and max_token_limit
-        < settings.LLM_TOKEN_LIMITS.get(gpt_model, settings.DEFAULT_MAX_HISTORY)
-        else settings.LLM_TOKEN_LIMITS.get(gpt_model, settings.DEFAULT_MAX_HISTORY)
+        if max_token_limit and max_token_limit < model_token_limit
+        else model_token_limit
     )
 
     if not history:
@@ -161,7 +176,7 @@ def limit_chat_history(history, max_token_limit=None, gpt_model="docsgpt"):
 
 
 def validate_function_name(function_name):
-    """Validates if a function name matches the allowed pattern."""
+    """Validate function name matches allowed pattern (alphanumeric, underscore, hyphen)."""
     if not re.match(r"^[a-zA-Z0-9_-]+$", function_name):
         return False
     return True
@@ -180,3 +195,51 @@ def generate_image_url(image_path):
     else:
         base_url = getattr(settings, "API_URL", "http://localhost:7091")
         return f"{base_url}/api/images/{image_path}"
+
+
+def clean_text_for_tts(text: str) -> str:
+    """
+    clean text for Text-to-Speech processing.
+    """
+    # Handle code blocks and links
+
+    text = re.sub(r"```mermaid[\s\S]*?```", " flowchart, ", text)  ## ```mermaid...```
+    text = re.sub(r"```[\s\S]*?```", " code block, ", text)  ## ```code```
+    text = re.sub(r"\[([^\]]+)\]\([^\)]+\)", r"\1", text)  ## [text](url)
+    text = re.sub(r"!\[([^\]]*)\]\([^\)]+\)", "", text)  ## ![alt](url)
+
+    # Remove markdown formatting
+
+    text = re.sub(r"`([^`]+)`", r"\1", text)  ## `code`
+    text = re.sub(r"\{([^}]*)\}", r" \1 ", text)  ## {text}
+    text = re.sub(r"[{}]", " ", text)  ## unmatched {}
+    text = re.sub(r"\[([^\]]+)\]", r" \1 ", text)  ## [text]
+    text = re.sub(r"[\[\]]", " ", text)  ## unmatched []
+    text = re.sub(r"(\*\*|__)(.*?)\1", r"\2", text)  ## **bold** __bold__
+    text = re.sub(r"(\*|_)(.*?)\1", r"\2", text)  ## *italic* _italic_
+    text = re.sub(r"^#{1,6}\s+", "", text, flags=re.MULTILINE)  ## # headers
+    text = re.sub(r"^>\s+", "", text, flags=re.MULTILINE)  ## > blockquotes
+    text = re.sub(r"^[\s]*[-\*\+]\s+", "", text, flags=re.MULTILINE)  ## - * + lists
+    text = re.sub(r"^[\s]*\d+\.\s+", "", text, flags=re.MULTILINE)  ## 1. numbered lists
+    text = re.sub(
+        r"^[\*\-_]{3,}\s*$", "", text, flags=re.MULTILINE
+    )  ## --- *** ___ rules
+    text = re.sub(r"<[^>]*>", "", text)  ## <html> tags
+
+    # Remove non-ASCII (emojis, special Unicode)
+
+    text = re.sub(r"[^\x20-\x7E\n\r\t]", "", text)
+
+    # Replace special sequences
+
+    text = re.sub(r"-->", ", ", text)  ## -->
+    text = re.sub(r"<--", ", ", text)  ## <--
+    text = re.sub(r"=>", ", ", text)  ## =>
+    text = re.sub(r"::", " ", text)  ## ::
+
+    # Normalize whitespace
+
+    text = re.sub(r"\s+", " ", text)
+    text = text.strip()
+
+    return text

application/worker.py

@@ -165,7 +165,7 @@ def run_agent_logic(agent_config, input_data):
             agent_type,
             endpoint="webhook",
             llm_name=settings.LLM_PROVIDER,
-            gpt_model=settings.LLM_NAME,
+            model_id=settings.LLM_NAME,
             api_key=settings.API_KEY,
             user_api_key=user_api_key,
             prompt=prompt,
@@ -180,7 +180,7 @@ def run_agent_logic(agent_config, input_data):
             prompt=prompt,
             chunks=chunks,
             token_limit=settings.DEFAULT_MAX_HISTORY,
-            gpt_model=settings.LLM_NAME,
+            model_id=settings.LLM_NAME,
             user_api_key=user_api_key,
             decoded_token=decoded_token,
         )

docs/pages/Guides/Customising-prompts.mdx

@@ -1,49 +1,453 @@
 ---
-title: Customizing Prompts 
-description: This guide will explain how to change prompts in DocsGPT and why it might be benefitial. Additionaly this article expains additional variables that can be used in prompts.
+title: Customizing Prompts
+description: This guide explains how to customize prompts in DocsGPT using the new template-based system with dynamic variable injection.
 ---
 
 import Image from 'next/image'
 
-# Customizing the Main Prompt
+# Customizing Prompts in DocsGPT
 
-Customizing the main prompt for DocsGPT gives you the ability to tailor the AI's responses to your specific requirements. By modifying the prompt text, you can achieve more accurate and relevant answers. Here's how you can do it:
+Customizing prompts for DocsGPT gives you powerful control over the AI's behavior and responses. With the new template-based system, you can inject dynamic context through organized namespaces, making prompts flexible and maintainable without hardcoding values.
+
+## Quick Start
 
 1. Navigate to `SideBar -> Settings`.
-
-
-
-
-2.In Settings select the `Active Prompt` now you will be able to see various prompts style.x
-
-
-
-
-3.Click on the `edit icon`  on the prompt of your choice and you will be able to see the current prompt for it,you can now customise  the prompt as per your choice.
+2. In Settings, select the `Active Prompt` to see various prompt styles.
+3. Click on the `edit icon` on your chosen prompt to customize it.
 
 ### Video Demo
 <Image src="/prompts.gif" alt="prompts" width={800} height={500} />
 
+---
 
+## Template-Based Prompt System
 
-## Example Prompt Modification
+DocsGPT now uses **Jinja2 templating** with four organized namespaces for dynamic variable injection:
+
+### Available Namespaces
+
+#### 1. **`system`** - System Metadata
+Access system-level information:
+
+```jinja
+{{ system.date }}         # Current date (YYYY-MM-DD)
+{{ system.time }}         # Current time (HH:MM:SS)
+{{ system.timestamp }}    # ISO 8601 timestamp
+{{ system.request_id }}   # Unique request identifier
+{{ system.user_id }}      # Current user ID
+```
+
+#### 2. **`source`** - Retrieved Documents
+Access RAG (Retrieval-Augmented Generation) document context:
+
+```jinja
+{{ source.content }}      # Concatenated document content
+{{ source.summaries }}    # Alias for content (backward compatible)
+{{ source.documents }}    # List of document objects
+{{ source.count }}        # Number of retrieved documents
+```
+
+#### 3. **`passthrough`** - Request Parameters
+Access custom parameters passed in the API request:
+
+```jinja
+{{ passthrough.company }}     # Custom field from request
+{{ passthrough.user_name }}   # User-provided data
+{{ passthrough.context }}     # Any custom parameter
+```
+
+To use passthrough data, send it in your API request:
+```json
+{
+  "question": "What is the pricing?",
+  "passthrough": {
+    "company": "Acme Corp",
+    "user_name": "Alice",
+    "plan_type": "enterprise"
+  }
+}
+```
+
+#### 4. **`tools`** - Pre-fetched Tool Data
+Access results from tools that run before the agent (like memory tool):
+
+```jinja
+{{ tools.memory.root }}       # Memory tool directory listing
+{{ tools.memory.available }}  # Boolean: is memory available
+```
+
+---
+
+## Example Prompts
+
+### Basic Prompt with Documents
+```jinja
+You are a helpful AI assistant for DocsGPT.
+
+Current date: {{ system.date }}
+
+Use the following documents to answer the question:
+
+{{ source.content }}
+
+Provide accurate, helpful answers with code examples when relevant.
+```
+
+### Advanced Prompt with All Namespaces
+```jinja
+You are an AI assistant for {{ passthrough.company }}.
+
+**System Info:**
+- Date: {{ system.date }}
+- Request ID: {{ system.request_id }}
+
+**User Context:**
+- User: {{ passthrough.user_name }}
+- Role: {{ passthrough.role }}
+
+**Available Documents ({{ source.count }}):**
+{{ source.content }}
+
+**Memory Context:**
+{% if tools.memory.available %}
+{{ tools.memory.root }}
+{% else %}
+No saved context available.
+{% endif %}
+
+Please provide detailed, accurate answers based on the documents above.
+```
+
+### Conditional Logic Example
+```jinja
+You are a DocsGPT assistant.
+
+{% if source.count > 0 %}
+I found {{ source.count }} relevant document(s):
+
+{{ source.content }}
+
+Base your answer on these documents.
+{% else %}
+No documents were found. Please answer based on your general knowledge.
+{% endif %}
+```
+
+---
+
+## Migration Guide
+
+### Legacy Format (Still Supported)
+The old `{summaries}` format continues to work for backward compatibility:
 
-**Original Prompt:**
 ```markdown
-You are a DocsGPT, friendly and helpful AI assistant by Arc53 that provides help with documents. You give thorough answers with code examples if possible.
-Use the following pieces of context to help answer the users question. If it's not relevant to the question, provide friendly responses.
-You have access to chat history, and can use it to help answer the question.
-When using code examples, use the following format:
+You are a helpful assistant.
 
-(code)
+Documents:
 {summaries}
 ```
 
-Note that `{summaries}` allows model to see and respond to your upploaded documents. If you don't want this functionality you can safely remove it from the customized prompt.
+This will automatically substitute `{summaries}` with document content.
 
-Feel free to customize the prompt to align it with your specific use case or the kind of responses you want from the AI. For example, you can focus on specific document types, industries, or topics to get more targeted results.
+### New Template Format (Recommended)
+Migrate to the new template syntax for more flexibility:
+
+```jinja
+You are a helpful assistant.
+
+Documents:
+{{ source.content }}
+```
+
+**Migration mapping:**
+- `{summaries}` → `{{ source.content }}` or `{{ source.summaries }}`
+
+---
+
+## Best Practices
+
+### 1. **Use Descriptive Context**
+```jinja
+**Retrieved Documents:**
+{{ source.content }}
+
+**User Query Context:**
+- Company: {{ passthrough.company }}
+- Department: {{ passthrough.department }}
+```
+
+### 2. **Handle Missing Data Gracefully**
+```jinja
+{% if passthrough.user_name %}
+Hello {{ passthrough.user_name }}!
+{% endif %}
+```
+
+### 3. **Leverage Memory for Continuity**
+```jinja
+{% if tools.memory.available %}
+**Previous Context:**
+{{ tools.memory.root }}
+{% endif %}
+
+**Current Question:**
+Please consider the above context when answering.
+```
+
+### 4. **Add Clear Instructions**
+```jinja
+You are a technical support assistant.
+
+**Guidelines:**
+1. Always reference the documents below
+2. Provide step-by-step instructions
+3. Include code examples when relevant
+
+**Reference Documents:**
+{{ source.content }}
+```
+
+---
+
+## Advanced Features
+
+### Looping Over Documents
+```jinja
+{% for doc in source.documents %}
+**Source {{ loop.index }}:** {{ doc.filename }}
+{{ doc.text }}
+
+{% endfor %}
+```
+
+### Date-Based Behavior
+```jinja
+{% if system.date > "2025-01-01" %}
+Note: This is information from 2025 or later.
+{% endif %}
+```
+
+### Custom Formatting
+```jinja
+**Request Information**
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+• Request ID: {{ system.request_id }}
+• User: {{ passthrough.user_name | default("Guest") }}
+• Time: {{ system.time }}
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+```
+
+---
+
+## Tool Pre-Fetching
+
+### Memory Tool Configuration
+Enable memory tool pre-fetching to inject saved context into prompts:
+
+```python
+# In your tool configuration
+{
+  "name": "memory",
+  "config": {
+    "pre_fetch_enabled": true  # Default: true
+  }
+}
+```
+
+Control pre-fetching globally:
+```bash
+# .env file
+ENABLE_TOOL_PREFETCH=true
+```
+
+Or per-request:
+```json
+{
+  "question": "What are the requirements?",
+  "disable_tool_prefetch": false
+}
+```
+
+---
+
+## Debugging Prompts
+
+### View Rendered Prompts in Logs
+Set log level to `INFO` to see the final rendered prompt sent to the LLM:
+
+```bash
+export LOG_LEVEL=INFO
+```
+
+You'll see output like:
+```
+INFO - Rendered system prompt for agent (length: 1234 chars):
+================================================================================
+You are a helpful assistant for Acme Corp.
+
+Current date: 2025-10-30
+Request ID: req_abc123
+
+Documents:
+Technical documentation about...
+================================================================================
+```
+
+### Template Validation
+Test your template syntax before saving:
+```python
+from application.api.answer.services.prompt_renderer import PromptRenderer
+
+renderer = PromptRenderer()
+is_valid = renderer.validate_template("Your prompt with {{ variables }}")
+```
+
+---
+
+## Common Use Cases
+
+### 1. Customer Support Bot
+```jinja
+You are a customer support assistant for {{ passthrough.company }}.
+
+**Customer:** {{ passthrough.customer_name }}
+**Ticket ID:** {{ system.request_id }}
+**Date:** {{ system.date }}
+
+**Knowledge Base:**
+{{ source.content }}
+
+**Previous Interactions:**
+{{ tools.memory.root }}
+
+Please provide helpful, friendly support based on the knowledge base above.
+```
+
+### 2. Technical Documentation Assistant
+```jinja
+You are a technical documentation expert.
+
+**Available Documentation ({{ source.count }} documents):**
+{{ source.content }}
+
+**Requirements:**
+- Provide code examples in {{ passthrough.language }}
+- Focus on {{ passthrough.framework }} best practices
+- Include relevant links when possible
+```
+
+### 3. Internal Knowledge Base
+```jinja
+You are an internal AI assistant for {{ passthrough.department }}.
+
+**Employee:** {{ passthrough.employee_name }}
+**Access Level:** {{ passthrough.access_level }}
+
+**Relevant Documents:**
+{{ source.content }}
+
+Provide detailed answers appropriate for {{ passthrough.access_level }} access level.
+```
+
+---
+
+## Template Syntax Reference
+
+### Variables
+```jinja
+{{ variable_name }}              # Output variable
+{{ namespace.field }}            # Access nested field
+{{ variable | default("N/A") }}  # Default value
+```
+
+### Conditionals
+```jinja
+{% if condition %}
+  Content
+{% elif other_condition %}
+  Other content
+{% else %}
+  Default content
+{% endif %}
+```
+
+### Loops
+```jinja
+{% for item in list %}
+  {{ item.field }}
+{% endfor %}
+```
+
+### Comments
+```jinja
+{# This is a comment and won't appear in output #}
+```
+
+---
+
+## Security Considerations
+
+1. **Input Sanitization**: Passthrough data is automatically sanitized to prevent injection attacks
+2. **Type Filtering**: Only primitive types (string, int, float, bool, None) are allowed in passthrough
+3. **Autoescaping**: Jinja2 autoescaping is enabled by default
+4. **Size Limits**: Consider the token budget when including large documents
+
+---
+
+## Troubleshooting
+
+### Problem: Variables Not Rendering
+**Solution:** Ensure you're using the correct namespace:
+```jinja
+❌ {{ company }}
+✅ {{ passthrough.company }}
+```
+
+### Problem: Empty Output for Tool Data
+**Solution:** Check that tool pre-fetching is enabled and the tool is configured correctly.
+
+### Problem: Syntax Errors
+**Solution:** Validate template syntax. Common issues:
+```jinja
+❌ {{ variable }     # Missing closing brace
+❌ {% if x %         # Missing closing %}
+✅ {{ variable }}
+✅ {% if x %}...{% endif %}
+```
+
+### Problem: Legacy Prompts Not Working
+**Solution:** The system auto-detects template syntax. If your prompt uses `{summaries}`, it will work in legacy mode. To use new features, add `{{ }}` syntax.
+
+---
+
+## API Reference
+
+### Render Prompt via API
+```python
+from application.api.answer.services.prompt_renderer import PromptRenderer
+
+renderer = PromptRenderer()
+rendered = renderer.render_prompt(
+    prompt_content="Your template with {{ passthrough.name }}",
+    user_id="user_123",
+    request_id="req_456",
+    passthrough_data={"name": "Alice"},
+    docs_together="Document content here",
+    tools_data={"memory": {"root": "Files: notes.txt"}}
+)
+```
+
+---
 
 ## Conclusion
 
-Customizing the main prompt for DocsGPT allows you to tailor the AI's responses to your unique requirements. Whether you need in-depth explanations, code examples, or specific insights, you can achieve it by modifying the main prompt. Remember to experiment and fine-tune your prompts to get the best results.
+The new template-based prompt system provides powerful flexibility while maintaining backward compatibility. By leveraging namespaces, you can create dynamic, context-aware prompts that adapt to your specific use case.
 
+**Key Benefits:**
+- ✅ Dynamic variable injection
+- ✅ Organized namespaces
+- ✅ Backward compatible
+- ✅ Security built-in
+- ✅ Easy to debug
+
+Start with simple templates and gradually add complexity as needed. Happy prompting! 🚀

docs/pages/quickstart.mdx

@@ -57,7 +57,7 @@ The easiest way to launch DocsGPT is using the provided `setup.sh` script. This
 
    * **4) Connect Cloud API Provider:** This option lets you connect DocsGPT to a commercial Cloud API provider such as OpenAI, Google (Vertex AI/Gemini), Anthropic (Claude), Groq, HuggingFace Inference API, or Azure OpenAI. You will need an API key from your chosen provider. Select this if you prefer to use a powerful cloud-based LLM.
 
-   * **5) Modify DocsGPT's source code and rebuild the Docker images locally. Instead of pulling prebuilt images from Docker Hub or using the hosted/public API, you build the entire backend and frontend from source, customizing how DocsGPT works internally, or run it in an environment without internet access.
+   * **5) Modify DocsGPT's source code and rebuild the Docker images locally.** Instead of pulling prebuilt images from Docker Hub or using the hosted/public API, you build the entire backend and frontend from source, customizing how DocsGPT works internally, or run it in an environment without internet access.
 
    After selecting an option and providing any required information (like API keys or model names), the script will configure your `.env` file and start DocsGPT using Docker Compose.
 
@@ -119,4 +119,4 @@ If you prefer a more manual approach, you can follow our [Docker Deployment docu
 
 For more advanced customization of DocsGPT settings, such as configuring vector stores, embedding models, and other parameters, please refer to the [DocsGPT Settings documentation](/Deploying/DocsGPT-Settings). This guide explains how to modify the `.env` file or `settings.py` for deeper configuration.
 
-Enjoy using DocsGPT!
+Enjoy using DocsGPT!

frontend/.env.development

@@ -3,4 +3,4 @@ VITE_BASE_URL=http://localhost:5173
 VITE_API_HOST=http://127.0.0.1:7091
 VITE_API_STREAMING=true
 VITE_NOTIFICATION_TEXT="What's new in 0.14.0 — Changelog"
-VITE_NOTIFICATION_LINK="#"
+VITE_NOTIFICATION_LINK="https://blog.docsgpt.cloud/docsgpt-0-14-agents-automate-integrate-and-innovate/"

frontend/.eslintignore

@@ -1,17 +0,0 @@
-node_modules/
-dist/
-prettier.config.cjs
-.eslintrc.cjs
-env.d.ts
-public/
-assets/
-vite-env.d.ts
-.prettierignore
-package-lock.json
-package.json
-postcss.config.cjs
-prettier.config.cjs
-tailwind.config.cjs
-tsconfig.json
-tsconfig.node.json
-vite.config.ts

frontend/.eslintrc.cjs

@@ -1,45 +0,0 @@
-module.exports = {
-  env: {
-    browser: true,
-    es2021: true,
-    node: true,
-  },
-  extends: [
-    'eslint:recommended',
-    'plugin:@typescript-eslint/recommended',
-    'plugin:react/recommended',
-    'plugin:prettier/recommended',
-  ],
-  overrides: [],
-  parser: '@typescript-eslint/parser',
-  parserOptions: {
-    ecmaVersion: 'latest',
-    sourceType: 'module',
-  },
-  plugins: ['react', 'unused-imports'],
-  rules: {
-    'react/prop-types': 'off',
-    'unused-imports/no-unused-imports': 'error',
-    'react/react-in-jsx-scope': 'off',
-    'prettier/prettier': [
-      'error',
-      {
-        endOfLine: 'auto',
-      },
-    ],
-  },
-  settings: {
-    'import/parsers': {
-      '@typescript-eslint/parser': ['.ts', '.tsx'],
-    },
-    react: {
-      version: 'detect',
-    },
-    'import/resolver': {
-      node: {
-        paths: ['src'],
-        extensions: ['.js', '.jsx', '.ts', '.tsx'],
-      },
-    },
-  },
-};

frontend/eslint.config.js

@@ -0,0 +1,78 @@
+import js from '@eslint/js'
+import tsParser from '@typescript-eslint/parser'
+import tsPlugin from '@typescript-eslint/eslint-plugin'
+import react from 'eslint-plugin-react'
+import unusedImports from 'eslint-plugin-unused-imports'
+import prettier from 'eslint-plugin-prettier'
+import globals from 'globals'
+
+export default [
+  {
+    ignores: [
+      'node_modules/',
+      'dist/',
+      'prettier.config.cjs',
+      '.eslintrc.cjs',
+      'env.d.ts',
+      'public/',
+      'assets/',
+      'vite-env.d.ts',
+      '.prettierignore',
+      'package-lock.json',
+      'package.json',
+      'postcss.config.cjs',
+      'tailwind.config.cjs',
+      'tsconfig.json',
+      'tsconfig.node.json',
+      'vite.config.ts',
+    ],
+  },
+  {
+    files: ['**/*.{js,jsx,ts,tsx}'],
+    languageOptions: {
+      ecmaVersion: 'latest',
+      sourceType: 'module',
+      parser: tsParser,
+      parserOptions: {
+        ecmaFeatures: {
+          jsx: true,
+        },
+      },
+      globals: {
+        ...globals.browser,
+        ...globals.es2021,
+        ...globals.node,
+      },
+    },
+    plugins: {
+      '@typescript-eslint': tsPlugin,
+      react,
+      'unused-imports': unusedImports,
+      prettier,
+    },
+    rules: {
+      ...js.configs.recommended.rules,
+      ...tsPlugin.configs.recommended.rules,
+      ...react.configs.recommended.rules,
+      ...prettier.configs.recommended.rules,
+      'react/prop-types': 'off',
+      'unused-imports/no-unused-imports': 'error',
+      'react/react-in-jsx-scope': 'off',
+      'no-undef': 'off',
+      '@typescript-eslint/no-explicit-any': 'warn',
+      '@typescript-eslint/no-unused-vars': 'warn',
+      '@typescript-eslint/no-unused-expressions': 'warn',
+      'prettier/prettier': [
+        'error',
+        {
+          endOfLine: 'auto',
+        },
+      ],
+    },
+    settings: {
+      react: {
+        version: 'detect',
+      },
+    },
+  },
+]

frontend/package.json

@@ -19,25 +19,25 @@
     ]
   },
   "dependencies": {
-    "@reduxjs/toolkit": "^2.8.2",
+    "@reduxjs/toolkit": "^2.10.1",
     "chart.js": "^4.4.4",
     "clsx": "^2.1.1",
     "copy-to-clipboard": "^3.3.3",
     "i18next": "^25.5.3",
-    "i18next-browser-languagedetector": "^8.0.2",
+    "i18next-browser-languagedetector": "^8.2.0",
     "lodash": "^4.17.21",
-    "mermaid": "^11.12.0",
+    "mermaid": "^11.12.1",
     "prop-types": "^15.8.1",
     "react": "^19.1.0",
     "react-chartjs-2": "^5.3.0",
-    "react-dom": "^19.0.0",
+    "react-dom": "^19.1.1",
     "react-dropzone": "^14.3.8",
     "react-google-drive-picker": "^1.2.2",
-    "react-i18next": "^15.4.0",
+    "react-i18next": "^16.2.4",
     "react-markdown": "^9.0.1",
     "react-redux": "^9.2.0",
     "react-router-dom": "^7.6.1",
-    "react-syntax-highlighter": "^15.6.1",
+    "react-syntax-highlighter": "^16.1.0",
     "rehype-katex": "^7.0.1",
     "remark-gfm": "^4.0.0",
     "remark-math": "^6.0.0",
@@ -46,30 +46,28 @@
   "devDependencies": {
     "@tailwindcss/postcss": "^4.1.10",
     "@types/lodash": "^4.17.20",
-    "@types/mermaid": "^9.1.0",
     "@types/react": "^19.1.8",
-    "@types/react-dom": "^19.0.0",
+    "@types/react-dom": "^19.1.7",
     "@types/react-syntax-highlighter": "^15.5.13",
-    "@typescript-eslint/eslint-plugin": "^5.51.0",
-    "@typescript-eslint/parser": "^5.62.0",
+    "@typescript-eslint/eslint-plugin": "^8.46.3",
+    "@typescript-eslint/parser": "^8.46.3",
     "@vitejs/plugin-react": "^4.3.4",
-    "eslint": "^8.57.1",
+    "eslint": "^9.39.1",
     "eslint-config-prettier": "^10.1.5",
-    "eslint-config-standard-with-typescript": "^34.0.0",
     "eslint-plugin-import": "^2.31.0",
-    "eslint-plugin-n": "^15.7.0",
-    "eslint-plugin-prettier": "^5.2.1",
+    "eslint-plugin-n": "^17.23.1",
+    "eslint-plugin-prettier": "^5.5.4",
     "eslint-plugin-promise": "^6.6.0",
     "eslint-plugin-react": "^7.37.5",
     "eslint-plugin-unused-imports": "^4.1.4",
-    "husky": "^8.0.0",
+    "husky": "^9.1.7",
     "lint-staged": "^15.3.0",
     "postcss": "^8.4.49",
     "prettier": "^3.5.3",
-    "prettier-plugin-tailwindcss": "^0.6.13",
+    "prettier-plugin-tailwindcss": "^0.7.1",
     "tailwindcss": "^4.1.11",
     "typescript": "^5.8.3",
-    "vite": "^6.3.5",
+    "vite": "^7.2.0",
     "vite-plugin-svgr": "^4.3.0"
   }
 }

frontend/public/toolIcons/tool_todo_list.svg

@@ -0,0 +1 @@
+<svg xmlns="http://www.w3.org/2000/svg" height="24px" viewBox="0 -960 960 960" width="24px" fill="#e3e3e3"><path d="M240-80q-33 0-56.5-23.5T160-160v-640q0-33 23.5-56.5T240-880h480q33 0 56.5 23.5T800-800v640q0 33-23.5 56.5T720-80H240Zm0-80h480v-640H240v640Zm88-104 56-56-56-56-56 56 56 56Zm0-160 56-56-56-56-56 56 56 56Zm0-160 56-56-56-56-56 56 56 56Zm120 280h232v-80H448v80Zm0-160h232v-80H448v80Zm0-160h232v-80H448v80ZM240-160v-640 640Z"/></svg>

frontend/src/Hero.tsx

@@ -1,6 +1,8 @@
-import DocsGPT3 from './assets/cute_docsgpt3.svg';
 import { useTranslation } from 'react-i18next';
 
+import DocsGPT3 from './assets/cute_docsgpt3.svg';
+import DropdownModel from './components/DropdownModel';
+
 export default function Hero({
   handleQuestion,
 }: {
@@ -26,6 +28,10 @@ export default function Hero({
           <span className="text-4xl font-semibold">DocsGPT</span>
           <img className="mb-1 inline w-14" src={DocsGPT3} alt="docsgpt" />
         </div>
+        {/* Model Selector */}
+        <div className="relative w-72">
+          <DropdownModel />
+        </div>
       </div>
 
       {/* Demo Buttons Section */}
@@ -38,7 +44,7 @@ export default function Hero({
                 <button
                   key={key}
                   onClick={() => handleQuestion({ question: demo.query })}
-                  className={`border-dark-gray text-just-black hover:bg-cultured dark:border-dim-gray dark:text-chinese-white dark:hover:bg-charleston-green w-full rounded-[66px] border bg-transparent px-6 py-[14px] text-left transition-colors ${key >= 2 ? 'hidden md:block' : ''} // Show only 2 buttons on mobile`}
+                  className={`border-dark-gray text-just-black hover:bg-cultured dark:border-dim-gray dark:text-chinese-white dark:hover:bg-charleston-green w-full rounded-[66px] border bg-transparent px-6 py-[14px] text-left transition-colors ${key >= 2 ? 'hidden md:block' : ''}`}
                 >
                   <p className="text-black-1000 dark:text-bright-gray mb-2 font-semibold">
                     {demo.header}

frontend/src/Navigation.tsx

@@ -9,7 +9,8 @@ import userService from './api/services/userService';
 import Add from './assets/add.svg';
 import DocsGPT3 from './assets/cute_docsgpt3.svg';
 import Discord from './assets/discord.svg';
-import Expand from './assets/expand.svg';
+import PanelLeftClose from './assets/panel-left-close.svg';
+import PanelLeftOpen from './assets/panel-left-open.svg';
 import Github from './assets/git_nav.svg';
 import Hamburger from './assets/hamburger.svg';
 import openNewChat from './assets/openNewChat.svg';
@@ -302,18 +303,20 @@ export default function Navigation({ navOpen, setNavOpen }: NavigationProps) {
       {
         <div className="absolute top-3 left-3 z-20 hidden transition-all duration-300 ease-in-out lg:block">
           <div className="flex items-center gap-3">
-            <button
-              onClick={() => {
-                setNavOpen(!navOpen);
-              }}
-              className="transition-transform duration-200 hover:scale-110"
-            >
-              <img
-                src={Expand}
-                alt="Toggle navigation menu"
-                className="m-auto transition-all duration-300 ease-in-out"
-              />
-            </button>
+            {!navOpen && (
+              <button
+                onClick={() => {
+                  setNavOpen(!navOpen);
+                }}
+                className="transition-transform duration-200 hover:scale-110"
+              >
+                <img
+                  src={PanelLeftOpen}
+                  alt="Open navigation menu"
+                  className="m-auto transition-all duration-300 ease-in-out"
+                />
+              </button>
+            )}
             {queries?.length > 0 && (
               <button
                 onClick={() => {
@@ -363,8 +366,8 @@ export default function Navigation({ navOpen, setNavOpen }: NavigationProps) {
             }}
           >
             <img
-              src={Expand}
-              alt="Toggle navigation menu"
+              src={navOpen ? PanelLeftClose : PanelLeftOpen}
+              alt={navOpen ? 'Collapse sidebar' : 'Expand sidebar'}
               className="m-auto transition-all duration-300 ease-in-out hover:scale-110"
             />
           </button>
@@ -408,7 +411,9 @@ export default function Navigation({ navOpen, setNavOpen }: NavigationProps) {
           {recentAgents?.length > 0 ? (
             <div>
               <div className="mx-4 my-auto mt-2 flex h-6 items-center">
-                <p className="mt-1 ml-4 text-sm font-semibold">Agents</p>
+                <p className="mt-1 ml-4 text-sm font-semibold">
+                  {t('navigation.agents')}
+                </p>
               </div>
               <div className="agents-container">
                 <div>
@@ -562,7 +567,7 @@ export default function Navigation({ navOpen, setNavOpen }: NavigationProps) {
               <div className="flex items-center gap-1 pr-4">
                 <NavLink
                   target="_blank"
-                  to={'https://discord.gg/WHJdfbQDR4'}
+                  to={'https://discord.gg/vN7YFfdMpj'}
                   className={
                     'rounded-full hover:bg-gray-100 dark:hover:bg-[#28292E]'
                   }

frontend/src/PageNotFound.tsx

@@ -1,13 +1,16 @@
 import { Link } from 'react-router-dom';
+import { useTranslation } from 'react-i18next';
 
 export default function PageNotFound() {
+  const { t } = useTranslation();
+
   return (
     <div className="dark:bg-raisin-black grid min-h-screen">
       <p className="text-jet dark:bg-outer-space mx-auto my-auto mt-20 flex w-full max-w-6xl flex-col place-items-center gap-6 rounded-3xl bg-gray-100 p-6 lg:p-10 xl:p-16 dark:text-gray-100">
-        <h1>404</h1>
-        <p>The page you are looking for does not exist.</p>
+        <h1>{t('pageNotFound.title')}</h1>
+        <p>{t('pageNotFound.message')}</p>
         <button className="pointer-cursor bg-blue-1000 hover:bg-blue-3000 mr-4 flex cursor-pointer items-center justify-center rounded-full px-4 py-2 text-white transition-colors duration-100">
-          <Link to="/">Go Back Home</Link>
+          <Link to="/">{t('pageNotFound.goHome')}</Link>
         </button>
       </p>
     </div>

frontend/src/agents/AgentLogs.tsx

@@ -1,4 +1,5 @@
 import { useEffect, useState } from 'react';
+import { useTranslation } from 'react-i18next';
 import { useSelector } from 'react-redux';
 import { useNavigate, useParams } from 'react-router-dom';
 
@@ -11,6 +12,7 @@ import Logs from '../settings/Logs';
 import { Agent } from './types';
 
 export default function AgentLogs() {
+  const { t } = useTranslation();
   const navigate = useNavigate();
   const { agentId } = useParams();
   const token = useSelector(selectToken);
@@ -45,12 +47,12 @@ export default function AgentLogs() {
           <img src={ArrowLeft} alt="left-arrow" className="h-3 w-3" />
         </button>
         <p className="text-eerie-black dark:text-bright-gray mt-px text-sm font-semibold">
-          Back to all agents
+          {t('agents.backToAll')}
         </p>
       </div>
       <div className="mt-5 flex w-full flex-wrap items-center justify-between gap-2 px-4">
         <h1 className="text-eerie-black m-0 text-[32px] font-bold md:text-[40px] dark:text-white">
-          Agent Logs
+          {t('agents.logs.title')}
         </h1>
       </div>
       <div className="mt-6 flex flex-col gap-3 px-4">
@@ -59,9 +61,10 @@ export default function AgentLogs() {
             <p className="text-[#28292E] dark:text-[#E0E0E0]">{agent.name}</p>
             <p className="text-xs text-[#28292E] dark:text-[#E0E0E0]/40">
               {agent.last_used_at
-                ? 'Last used at ' +
+                ? t('agents.logs.lastUsedAt') +
+                  ' ' +
                   new Date(agent.last_used_at).toLocaleString()
-                : 'No usage history'}
+                : t('agents.logs.noUsageHistory')}
             </p>
           </div>
         )}
@@ -79,7 +82,9 @@ export default function AgentLogs() {
           <Spinner />
         </div>
       ) : (
-        agent && <Logs agentId={agent.id} tableHeader="Agent endpoint logs" />
+        agent && (
+          <Logs agentId={agent.id} tableHeader={t('agents.logs.tableHeader')} />
+        )
       )}
     </div>
   );

frontend/src/agents/AgentPreview.tsx

@@ -1,4 +1,5 @@
 import { useCallback, useEffect, useRef, useState } from 'react';
+import { useTranslation } from 'react-i18next';
 import { useDispatch, useSelector } from 'react-redux';
 
 import MessageInput from '../components/MessageInput';
@@ -17,6 +18,7 @@ import { selectSelectedAgent } from '../preferences/preferenceSlice';
 import { AppDispatch } from '../store';
 
 export default function AgentPreview() {
+  const { t } = useTranslation();
   const dispatch = useDispatch<AppDispatch>();
 
   const queries = useSelector(selectPreviewQueries);
@@ -109,18 +111,18 @@ export default function AgentPreview() {
     } else setLastQueryReturnedErr(false);
   }, [queries]);
   return (
-    <div>
-      <div className="dark:bg-raisin-black flex h-full flex-col items-center justify-between gap-2 overflow-y-hidden">
-        <div className="h-[512px] w-full overflow-y-auto">
-          <ConversationMessages
-            handleQuestion={handleQuestion}
-            handleQuestionSubmission={handleQuestionSubmission}
-            queries={queries}
-            status={status}
-            showHeroOnEmpty={false}
-          />
-        </div>
-        <div className="flex w-[95%] max-w-[1500px] flex-col items-center gap-4 pb-2 md:w-9/12 lg:w-8/12 xl:w-8/12 2xl:w-6/12">
+    <div className="relative h-full w-full">
+      <div className="scrollbar-thin absolute inset-0 bottom-[180px] overflow-hidden px-4 pt-4 [&>div>div]:!w-full [&>div>div]:!max-w-none">
+        <ConversationMessages
+          handleQuestion={handleQuestion}
+          handleQuestionSubmission={handleQuestionSubmission}
+          queries={queries}
+          status={status}
+          showHeroOnEmpty={false}
+        />
+      </div>
+      <div className="absolute right-0 bottom-0 left-0 flex w-full flex-col gap-4 pb-2">
+        <div className="w-full px-4">
           <MessageInput
             onSubmit={(text) => handleQuestionSubmission(text)}
             loading={status === 'loading'}
@@ -128,11 +130,10 @@ export default function AgentPreview() {
             showToolButton={selectedAgent ? false : true}
             autoFocus={false}
           />
-          <p className="text-gray-4000 dark:text-sonic-silver w-full self-center bg-transparent pt-2 text-center text-xs md:inline">
-            This is a preview of the agent. You can publish it to start using it
-            in conversations.
-          </p>
         </div>
+        <p className="text-gray-4000 dark:text-sonic-silver w-full bg-transparent text-center text-xs md:inline">
+          {t('agents.preview.testMessage')}
+        </p>
       </div>
     </div>
   );

frontend/src/agents/AgentsList.tsx

@@ -1,4 +1,5 @@
 import { useEffect, useState } from 'react';
+import { useTranslation } from 'react-i18next';
 import { useDispatch, useSelector } from 'react-redux';
 import { useNavigate } from 'react-router-dom';
 
@@ -17,6 +18,7 @@ import { agentSectionsConfig } from './agents.config';
 import { Agent } from './types';
 
 export default function AgentsList() {
+  const { t } = useTranslation();
   const dispatch = useDispatch();
   const token = useSelector(selectToken);
   const selectedAgent = useSelector(selectSelectedAgent);
@@ -33,11 +35,10 @@ export default function AgentsList() {
   return (
     <div className="p-4 md:p-12">
       <h1 className="text-eerie-black mb-0 text-[32px] font-bold lg:text-[40px] dark:text-[#E0E0E0]">
-        Agents
+        {t('agents.title')}
       </h1>
       <p className="dark:text-gray-4000 mt-5 text-[15px] text-[#71717A]">
-        Discover and create custom versions of DocsGPT that combine
-        instructions, extra knowledge, and any combination of skills
+        {t('agents.description')}
       </p>
       {agentSectionsConfig.map((sectionConfig) => (
         <AgentSection key={sectionConfig.id} config={sectionConfig} />
@@ -51,6 +52,7 @@ function AgentSection({
 }: {
   config: (typeof agentSectionsConfig)[number];
 }) {
+  const { t } = useTranslation();
   const navigate = useNavigate();
   const dispatch = useDispatch();
   const token = useSelector(selectToken);
@@ -85,16 +87,18 @@ function AgentSection({
       <div className="flex w-full items-center justify-between">
         <div className="flex flex-col gap-2">
           <h2 className="text-[18px] font-semibold text-[#18181B] dark:text-[#E0E0E0]">
-            {config.title}
+            {t(`agents.sections.${config.id}.title`)}
           </h2>
-          <p className="text-[13px] text-[#71717A]">{config.description}</p>
+          <p className="text-[13px] text-[#71717A]">
+            {t(`agents.sections.${config.id}.description`)}
+          </p>
         </div>
         {config.showNewAgentButton && (
           <button
             className="bg-purple-30 hover:bg-violets-are-blue rounded-full px-4 py-2 text-sm text-white"
             onClick={() => navigate('/agents/new')}
           >
-            New Agent
+            {t('agents.newAgent')}
           </button>
         )}
       </div>
@@ -117,13 +121,13 @@ function AgentSection({
           </div>
         ) : (
           <div className="flex h-72 w-full flex-col items-center justify-center gap-3 text-base text-[#18181B] dark:text-[#E0E0E0]">
-            <p>{config.emptyStateDescription}</p>
+            <p>{t(`agents.sections.${config.id}.emptyState`)}</p>
             {config.showNewAgentButton && (
               <button
                 className="bg-purple-30 hover:bg-violets-are-blue ml-2 rounded-full px-4 py-2 text-sm text-white"
                 onClick={() => navigate('/agents/new')}
               >
-                New Agent
+                {t('agents.newAgent')}
               </button>
             )}
           </div>

frontend/src/agents/NewAgent.tsx

@@ -1,8 +1,10 @@
 import isEqual from 'lodash/isEqual';
 import React, { useCallback, useEffect, useRef, useState } from 'react';
+import { useTranslation } from 'react-i18next';
 import { useDispatch, useSelector } from 'react-redux';
 import { useNavigate, useParams } from 'react-router-dom';
 
+import modelService from '../api/services/modelService';
 import userService from '../api/services/userService';
 import ArrowLeft from '../assets/arrow-left.svg';
 import SourceIcon from '../assets/source.svg';
@@ -25,11 +27,13 @@ import { UserToolType } from '../settings/types';
 import AgentPreview from './AgentPreview';
 import { Agent, ToolSummary } from './types';
 
+import type { Model } from '../models/types';
 const embeddingsName =
   import.meta.env.VITE_EMBEDDINGS_NAME ||
   'huggingface_sentence-transformers/all-mpnet-base-v2';
 
 export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
+  const { t } = useTranslation();
   const navigate = useNavigate();
   const dispatch = useDispatch();
   const { agentId } = useParams();
@@ -57,18 +61,25 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
     token_limit: undefined,
     limited_request_mode: false,
     request_limit: undefined,
+    models: [],
+    default_model_id: '',
   });
   const [imageFile, setImageFile] = useState<File | null>(null);
   const [prompts, setPrompts] = useState<
     { name: string; id: string; type: string }[]
   >([]);
   const [userTools, setUserTools] = useState<OptionType[]>([]);
+  const [availableModels, setAvailableModels] = useState<Model[]>([]);
   const [isSourcePopupOpen, setIsSourcePopupOpen] = useState(false);
   const [isToolsPopupOpen, setIsToolsPopupOpen] = useState(false);
+  const [isModelsPopupOpen, setIsModelsPopupOpen] = useState(false);
   const [selectedSourceIds, setSelectedSourceIds] = useState<
     Set<string | number>
   >(new Set());
   const [selectedTools, setSelectedTools] = useState<ToolSummary[]>([]);
+  const [selectedModelIds, setSelectedModelIds] = useState<Set<string>>(
+    new Set(),
+  );
   const [deleteConfirmation, setDeleteConfirmation] =
     useState<ActiveState>('INACTIVE');
   const [agentDetails, setAgentDetails] = useState<ActiveState>('INACTIVE');
@@ -84,11 +95,12 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
   const initialAgentRef = useRef<Agent | null>(null);
   const sourceAnchorButtonRef = useRef<HTMLButtonElement>(null);
   const toolAnchorButtonRef = useRef<HTMLButtonElement>(null);
+  const modelAnchorButtonRef = useRef<HTMLButtonElement>(null);
 
   const modeConfig = {
     new: {
-      heading: 'New Agent',
-      buttonText: 'Publish',
+      heading: t('agents.form.headings.new'),
+      buttonText: t('agents.form.buttons.publish'),
       showDelete: false,
       showSaveDraft: true,
       showLogs: false,
@@ -96,8 +108,8 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
       trackChanges: false,
     },
     edit: {
-      heading: 'Edit Agent',
-      buttonText: 'Save',
+      heading: t('agents.form.headings.edit'),
+      buttonText: t('agents.form.buttons.save'),
       showDelete: true,
       showSaveDraft: false,
       showLogs: true,
@@ -105,8 +117,8 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
       trackChanges: true,
     },
     draft: {
-      heading: 'New Agent (Draft)',
-      buttonText: 'Publish',
+      heading: t('agents.form.headings.draft'),
+      buttonText: t('agents.form.buttons.publish'),
       showDelete: true,
       showSaveDraft: true,
       showLogs: false,
@@ -116,8 +128,8 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
   };
   const chunks = ['0', '2', '4', '6', '8', '10'];
   const agentTypes = [
-    { label: 'Classic', value: 'classic' },
-    { label: 'ReAct', value: 'react' },
+    { label: t('agents.form.agentTypes.classic'), value: 'classic' },
+    { label: t('agents.form.agentTypes.react'), value: 'react' },
   ];
 
   const isPublishable = () => {
@@ -198,13 +210,19 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
 
     if (agent.limited_token_mode && agent.token_limit) {
       formData.append('limited_token_mode', 'True');
-      formData.append('token_limit', JSON.stringify(agent.token_limit));
-    } else formData.append('token_limit', '0');
+      formData.append('token_limit', agent.token_limit.toString());
+    } else {
+      formData.append('limited_token_mode', 'False');
+      formData.append('token_limit', '0');
+    }
 
     if (agent.limited_request_mode && agent.request_limit) {
       formData.append('limited_request_mode', 'True');
-      formData.append('request_limit', JSON.stringify(agent.request_limit));
-    } else formData.append('request_limit', '0');
+      formData.append('request_limit', agent.request_limit.toString());
+    } else {
+      formData.append('limited_request_mode', 'False');
+      formData.append('request_limit', '0');
+    }
 
     if (imageFile) formData.append('image', imageFile);
 
@@ -216,6 +234,13 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
       formData.append('json_schema', JSON.stringify(agent.json_schema));
     }
 
+    if (agent.models && agent.models.length > 0) {
+      formData.append('models', JSON.stringify(agent.models));
+    }
+    if (agent.default_model_id) {
+      formData.append('default_model_id', agent.default_model_id);
+    }
+
     try {
       setDraftLoading(true);
       const response =
@@ -295,15 +320,29 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
       formData.append('json_schema', JSON.stringify(agent.json_schema));
     }
 
+    // Always send the limited mode fields
     if (agent.limited_token_mode && agent.token_limit) {
       formData.append('limited_token_mode', 'True');
-      formData.append('token_limit', JSON.stringify(agent.token_limit));
-    } else formData.append('token_limit', '0');
+      formData.append('token_limit', agent.token_limit.toString());
+    } else {
+      formData.append('limited_token_mode', 'False');
+      formData.append('token_limit', '0');
+    }
 
     if (agent.limited_request_mode && agent.request_limit) {
       formData.append('limited_request_mode', 'True');
-      formData.append('request_limit', JSON.stringify(agent.request_limit));
-    } else formData.append('request_limit', '0');
+      formData.append('request_limit', agent.request_limit.toString());
+    } else {
+      formData.append('limited_request_mode', 'False');
+      formData.append('request_limit', '0');
+    }
+
+    if (agent.models && agent.models.length > 0) {
+      formData.append('models', JSON.stringify(agent.models));
+    }
+    if (agent.default_model_id) {
+      formData.append('default_model_id', agent.default_model_id);
+    }
 
     try {
       setPublishLoading(true);
@@ -373,8 +412,16 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
       const data = await response.json();
       setPrompts(data);
     };
+    const getModels = async () => {
+      const response = await modelService.getModels(null);
+      if (!response.ok) throw new Error('Failed to fetch models');
+      const data = await response.json();
+      const transformed = modelService.transformModels(data.models || []);
+      setAvailableModels(transformed);
+    };
     getTools();
     getPrompts();
+    getModels();
   }, [token]);
 
   // Auto-select default source if none selected
@@ -447,6 +494,34 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
     }
   }, [agentId, mode, token]);
 
+  useEffect(() => {
+    if (agent.models && agent.models.length > 0 && availableModels.length > 0) {
+      const agentModelIds = new Set(agent.models);
+      if (agentModelIds.size > 0 && selectedModelIds.size === 0) {
+        setSelectedModelIds(agentModelIds);
+      }
+    }
+  }, [agent.models, availableModels.length]);
+
+  useEffect(() => {
+    const modelsArray = Array.from(selectedModelIds);
+    if (modelsArray.length > 0) {
+      setAgent((prev) => ({
+        ...prev,
+        models: modelsArray,
+        default_model_id: modelsArray.includes(prev.default_model_id || '')
+          ? prev.default_model_id
+          : modelsArray[0],
+      }));
+    } else {
+      setAgent((prev) => ({
+        ...prev,
+        models: [],
+        default_model_id: '',
+      }));
+    }
+  }, [selectedModelIds]);
+
   useEffect(() => {
     const selectedSources = Array.from(selectedSourceIds)
       .map((id) =>
@@ -534,7 +609,7 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
     setHasChanges(isChanged);
   }, [agent, dispatch, effectiveMode, imageFile, jsonSchemaText]);
   return (
-    <div className="p-4 md:p-12">
+    <div className="flex flex-col px-4 pt-4 pb-2 max-[1179px]:min-h-[100dvh] min-[1180px]:h-[100dvh] md:px-12 md:pt-12 md:pb-3">
       <div className="flex items-center gap-3 px-4">
         <button
           className="rounded-full border p-3 text-sm text-gray-400 dark:border-0 dark:bg-[#28292D] dark:text-gray-500 dark:hover:bg-[#2E2F34]"
@@ -543,7 +618,7 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
           <img src={ArrowLeft} alt="left-arrow" className="h-3 w-3" />
         </button>
         <p className="text-eerie-black dark:text-bright-gray mt-px text-sm font-semibold">
-          Back to all agents
+          {t('agents.backToAll')}
         </p>
       </div>
       <div className="mt-5 flex w-full flex-wrap items-center justify-between gap-2 px-4">
@@ -555,7 +630,7 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
             className="text-purple-30 dark:text-light-gray mr-4 rounded-3xl py-2 text-sm font-medium dark:bg-transparent"
             onClick={handleCancel}
           >
-            Cancel
+            {t('agents.form.buttons.cancel')}
           </button>
           {modeConfig[effectiveMode].showDelete && agent.id && (
             <button
@@ -563,7 +638,7 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
               onClick={() => setDeleteConfirmation('ACTIVE')}
             >
               <span className="block h-4 w-4 bg-[url('/src/assets/red-trash.svg')] bg-contain bg-center bg-no-repeat transition-all group-hover:bg-[url('/src/assets/white-trash.svg')]" />
-              Delete
+              {t('agents.form.buttons.delete')}
             </button>
           )}
           {modeConfig[effectiveMode].showSaveDraft && (
@@ -578,7 +653,7 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
                 {draftLoading ? (
                   <Spinner size="small" color="#976af3" />
                 ) : (
-                  'Save Draft'
+                  t('agents.form.buttons.saveDraft')
                 )}
               </span>
             </button>
@@ -589,7 +664,7 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
               onClick={() => navigate(`/agents/logs/${agent.id}`)}
             >
               <span className="block h-5 w-5 bg-[url('/src/assets/monitoring-purple.svg')] bg-contain bg-center bg-no-repeat transition-all group-hover:bg-[url('/src/assets/monitoring-white.svg')]" />
-              Logs
+              {t('agents.form.buttons.logs')}
             </button>
           )}
           {modeConfig[effectiveMode].showAccessDetails && (
@@ -597,7 +672,7 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
               className="hover:bg-vi</button>olets-are-blue border-violets-are-blue text-violets-are-blue hover:bg-violets-are-blue rounded-3xl border border-solid px-5 py-2 text-sm font-medium transition-colors hover:text-white"
               onClick={() => setAgentDetails('ACTIVE')}
             >
-              Access Details
+              {t('agents.form.buttons.accessDetails')}
             </button>
           )}
           <button
@@ -615,20 +690,22 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
           </button>
         </div>
       </div>
-      <div className="mt-5 flex w-full grid-cols-5 flex-col gap-10 min-[1180px]:grid min-[1180px]:gap-5">
-        <div className="col-span-2 flex flex-col gap-5">
-          <div className="rounded-[30px] bg-[#F6F6F6] px-6 py-3 dark:bg-[#383838] dark:text-[#E0E0E0]">
-            <h2 className="text-lg font-semibold">Meta</h2>
+      <div className="mt-3 flex w-full flex-1 grid-cols-5 flex-col gap-10 rounded-[30px] bg-[#F6F6F6] p-5 max-[1179px]:overflow-visible min-[1180px]:grid min-[1180px]:gap-5 min-[1180px]:overflow-hidden dark:bg-[#383838]">
+        <div className="scrollbar-thin col-span-2 flex flex-col gap-5 max-[1179px]:overflow-visible min-[1180px]:max-h-full min-[1180px]:overflow-y-auto min-[1180px]:pr-3">
+          <div className="dark:bg-raisin-black rounded-[30px] bg-white px-6 py-3 dark:text-[#E0E0E0]">
+            <h2 className="text-lg font-semibold">
+              {t('agents.form.sections.meta')}
+            </h2>
             <input
               className="border-silver text-jet dark:bg-raisin-black dark:text-bright-gray dark:placeholder:text-silver mt-3 w-full rounded-3xl border bg-white px-5 py-3 text-sm outline-hidden placeholder:text-gray-400 dark:border-[#7E7E7E]"
               type="text"
               value={agent.name}
-              placeholder="Agent name"
+              placeholder={t('agents.form.placeholders.agentName')}
               onChange={(e) => setAgent({ ...agent, name: e.target.value })}
             />
             <textarea
               className="border-silver text-jet dark:bg-raisin-black dark:text-bright-gray dark:placeholder:text-silver mt-3 h-32 w-full rounded-xl border bg-white px-5 py-4 text-sm outline-hidden placeholder:text-gray-400 dark:border-[#7E7E7E]"
-              placeholder="Describe your agent"
+              placeholder={t('agents.form.placeholders.describeAgent')}
               value={agent.description}
               onChange={(e) =>
                 setAgent({ ...agent, description: e.target.value })
@@ -641,17 +718,22 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
                 onUpload={handleUpload}
                 onRemove={() => setImageFile(null)}
                 uploadText={[
-                  { text: 'Click to upload', colorClass: 'text-[#7D54D1]' },
                   {
-                    text: ' or drag and drop',
+                    text: t('agents.form.upload.clickToUpload'),
+                    colorClass: 'text-[#7D54D1]',
+                  },
+                  {
+                    text: t('agents.form.upload.dragAndDrop'),
                     colorClass: 'text-[#525252]',
                   },
                 ]}
               />
             </div>
           </div>
-          <div className="rounded-[30px] bg-[#F6F6F6] px-6 py-3 dark:bg-[#383838] dark:text-[#E0E0E0]">
-            <h2 className="text-lg font-semibold">Source</h2>
+          <div className="dark:bg-raisin-black rounded-[30px] bg-white px-6 py-3 dark:text-[#E0E0E0]">
+            <h2 className="text-lg font-semibold">
+              {t('agents.form.sections.source')}
+            </h2>
             <div className="mt-3">
               <div className="flex flex-wrap items-center gap-1">
                 <button
@@ -672,11 +754,13 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
                               source.name === id ||
                               source.retriever === id,
                           );
-                          return matchedDoc?.name || `External KB`;
+                          return (
+                            matchedDoc?.name || t('agents.form.externalKb')
+                          );
                         })
                         .filter(Boolean)
                         .join(', ')
-                    : 'Select sources'}
+                    : t('agents.form.placeholders.selectSources')}
                 </button>
                 <MultiSelectPopup
                   isOpen={isSourcePopupOpen}
@@ -720,9 +804,13 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
                       setSelectedSourceIds(newSelectedIds);
                     }
                   }}
-                  title="Select Sources"
-                  searchPlaceholder="Search sources..."
-                  noOptionsMessage="No sources available"
+                  title={t('agents.form.sourcePopup.title')}
+                  searchPlaceholder={t(
+                    'agents.form.sourcePopup.searchPlaceholder',
+                  )}
+                  noOptionsMessage={t(
+                    'agents.form.sourcePopup.noOptionsMessage',
+                  )}
                 />
               </div>
               <div className="mt-3">
@@ -737,14 +825,14 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
                   border="border"
                   buttonClassName="bg-white dark:bg-[#222327] border-silver dark:border-[#7E7E7E]"
                   optionsClassName="bg-white dark:bg-[#383838] border-silver dark:border-[#7E7E7E]"
-                  placeholder="Chunks per query"
+                  placeholder={t('agents.form.placeholders.chunksPerQuery')}
                   placeholderClassName="text-gray-400 dark:text-silver"
                   contentSize="text-sm"
                 />
               </div>
             </div>
           </div>
-          <div className="rounded-[30px] bg-[#F6F6F6] px-6 py-3 dark:bg-[#383838] dark:text-[#E0E0E0]">
+          <div className="dark:bg-raisin-black rounded-[30px] bg-white px-6 py-3 dark:text-[#E0E0E0]">
             <div className="flex flex-wrap items-end gap-1">
               <div className="min-w-20 grow basis-full sm:basis-0">
                 <Prompts
@@ -757,7 +845,7 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
                     setAgent({ ...agent, prompt_id: id })
                   }
                   setPrompts={setPrompts}
-                  title="Prompt"
+                  title={t('agents.form.sections.prompt')}
                   titleClassName="text-lg font-semibold dark:text-[#E0E0E0]"
                   showAddButton={false}
                   dropdownProps={{
@@ -777,12 +865,14 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
                 className="border-violets-are-blue text-violets-are-blue hover:bg-violets-are-blue w-20 shrink-0 basis-full rounded-3xl border-2 border-solid px-5 py-[11px] text-sm transition-colors hover:text-white sm:basis-auto"
                 onClick={() => setAddPromptModal('ACTIVE')}
               >
-                Add
+                {t('agents.form.buttons.add')}
               </button>
             </div>
           </div>
-          <div className="rounded-[30px] bg-[#F6F6F6] px-6 py-3 dark:bg-[#383838] dark:text-[#E0E0E0]">
-            <h2 className="text-lg font-semibold">Tools</h2>
+          <div className="dark:bg-raisin-black rounded-[30px] bg-white px-6 py-3 dark:text-[#E0E0E0]">
+            <h2 className="text-lg font-semibold">
+              {t('agents.form.sections.tools')}
+            </h2>
             <div className="mt-3 flex flex-wrap items-center gap-1">
               <button
                 ref={toolAnchorButtonRef}
@@ -798,7 +888,7 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
                       .map((tool) => tool.display_name || tool.name)
                       .filter(Boolean)
                       .join(', ')
-                  : 'Select tools'}
+                  : t('agents.form.placeholders.selectTools')}
               </button>
               <MultiSelectPopup
                 isOpen={isToolsPopupOpen}
@@ -817,14 +907,18 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
                       })),
                   )
                 }
-                title="Select Tools"
-                searchPlaceholder="Search tools..."
-                noOptionsMessage="No tools available"
+                title={t('agents.form.toolsPopup.title')}
+                searchPlaceholder={t(
+                  'agents.form.toolsPopup.searchPlaceholder',
+                )}
+                noOptionsMessage={t('agents.form.toolsPopup.noOptionsMessage')}
               />
             </div>
           </div>
-          <div className="rounded-[30px] bg-[#F6F6F6] px-6 py-3 dark:bg-[#383838] dark:text-[#E0E0E0]">
-            <h2 className="text-lg font-semibold">Agent type</h2>
+          <div className="dark:bg-raisin-black rounded-[30px] bg-white px-6 py-3 dark:text-[#E0E0E0]">
+            <h2 className="text-lg font-semibold">
+              {t('agents.form.sections.agentType')}
+            </h2>
             <div className="mt-3">
               <Dropdown
                 options={agentTypes}
@@ -842,13 +936,89 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
                 border="border"
                 buttonClassName="bg-white dark:bg-[#222327] border-silver dark:border-[#7E7E7E]"
                 optionsClassName="bg-white dark:bg-[#383838] border-silver dark:border-[#7E7E7E]"
-                placeholder="Select type"
+                placeholder={t('agents.form.placeholders.selectType')}
                 placeholderClassName="text-gray-400 dark:text-silver"
                 contentSize="text-sm"
               />
             </div>
           </div>
-          <div className="rounded-[30px] bg-[#F6F6F6] px-6 py-3 dark:bg-[#383838] dark:text-[#E0E0E0]">
+          <div className="dark:bg-raisin-black rounded-[30px] bg-white px-6 py-3 dark:text-[#E0E0E0]">
+            <h2 className="text-lg font-semibold">
+              {t('agents.form.sections.models')}
+            </h2>
+            <div className="mt-3 flex flex-col gap-3">
+              <button
+                ref={modelAnchorButtonRef}
+                onClick={() => setIsModelsPopupOpen(!isModelsPopupOpen)}
+                className={`border-silver dark:bg-raisin-black w-full truncate rounded-3xl border bg-white px-5 py-3 text-left text-sm dark:border-[#7E7E7E] ${
+                  selectedModelIds.size > 0
+                    ? 'text-jet dark:text-bright-gray'
+                    : 'dark:text-silver text-gray-400'
+                }`}
+              >
+                {selectedModelIds.size > 0
+                  ? availableModels
+                      .filter((m) => selectedModelIds.has(m.id))
+                      .map((m) => m.display_name)
+                      .join(', ')
+                  : t('agents.form.placeholders.selectModels')}
+              </button>
+              <MultiSelectPopup
+                isOpen={isModelsPopupOpen}
+                onClose={() => setIsModelsPopupOpen(false)}
+                anchorRef={modelAnchorButtonRef}
+                options={availableModels.map((model) => ({
+                  id: model.id,
+                  label: model.display_name,
+                }))}
+                selectedIds={selectedModelIds}
+                onSelectionChange={(newSelectedIds: Set<string | number>) =>
+                  setSelectedModelIds(
+                    new Set(Array.from(newSelectedIds).map(String)),
+                  )
+                }
+                title={t('agents.form.modelsPopup.title')}
+                searchPlaceholder={t(
+                  'agents.form.modelsPopup.searchPlaceholder',
+                )}
+                noOptionsMessage={t('agents.form.modelsPopup.noOptionsMessage')}
+              />
+              {selectedModelIds.size > 0 && (
+                <div>
+                  <label className="mb-2 block text-sm font-medium">
+                    {t('agents.form.labels.defaultModel')}
+                  </label>
+                  <Dropdown
+                    options={availableModels
+                      .filter((m) => selectedModelIds.has(m.id))
+                      .map((m) => ({
+                        label: m.display_name,
+                        value: m.id,
+                      }))}
+                    selectedValue={
+                      availableModels.find(
+                        (m) => m.id === agent.default_model_id,
+                      )?.display_name || null
+                    }
+                    onSelect={(option: { label: string; value: string }) =>
+                      setAgent({ ...agent, default_model_id: option.value })
+                    }
+                    size="w-full"
+                    rounded="3xl"
+                    border="border"
+                    buttonClassName="bg-white dark:bg-[#222327] border-silver dark:border-[#7E7E7E]"
+                    optionsClassName="bg-white dark:bg-[#383838] border-silver dark:border-[#7E7E7E]"
+                    placeholder={t(
+                      'agents.form.placeholders.selectDefaultModel',
+                    )}
+                    placeholderClassName="text-gray-400 dark:text-silver"
+                    contentSize="text-sm"
+                  />
+                </div>
+              )}
+            </div>
+          </div>
+          <div className="dark:bg-raisin-black rounded-[30px] bg-white px-6 py-3 dark:text-[#E0E0E0]">
             <button
               onClick={() =>
                 setIsAdvancedSectionExpanded(!isAdvancedSectionExpanded)
@@ -856,7 +1026,9 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
               className="flex w-full items-center justify-between text-left focus:outline-none"
             >
               <div>
-                <h2 className="text-lg font-semibold">Advanced</h2>
+                <h2 className="text-lg font-semibold">
+                  {t('agents.form.sections.advanced')}
+                </h2>
               </div>
               <div className="ml-4 flex items-center">
                 <svg
@@ -879,9 +1051,11 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
             {isAdvancedSectionExpanded && (
               <div className="mt-3">
                 <div>
-                  <h2 className="text-sm font-medium">JSON response schema</h2>
+                  <h2 className="text-sm font-medium">
+                    {t('agents.form.advanced.jsonSchema')}
+                  </h2>
                   <p className="mt-1 text-xs text-gray-600 dark:text-gray-400">
-                    Define a JSON schema to enforce structured output format
+                    {t('agents.form.advanced.jsonSchemaDescription')}
                   </p>
                 </div>
                 <textarea
@@ -915,17 +1089,19 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
                       }`}
                     />
                     {jsonSchemaValid
-                      ? 'Valid JSON'
-                      : 'Invalid JSON - fix to enable saving'}
+                      ? t('agents.form.advanced.validJson')
+                      : t('agents.form.advanced.invalidJson')}
                   </div>
                 )}
 
                 <div className="mt-6">
                   <div className="flex items-center justify-between">
                     <div>
-                      <h2 className="text-sm font-medium">Token limiting</h2>
+                      <h2 className="text-sm font-medium">
+                        {t('agents.form.advanced.tokenLimiting')}
+                      </h2>
                       <p className="mt-1 text-xs text-gray-600 dark:text-gray-400">
-                        Limit daily total tokens that can be used by this agent
+                        {t('agents.form.advanced.tokenLimitingDescription')}
                       </p>
                     </div>
                     <button
@@ -965,7 +1141,7 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
                       })
                     }
                     disabled={!agent.limited_token_mode}
-                    placeholder="Enter token limit"
+                    placeholder={t('agents.form.placeholders.enterTokenLimit')}
                     className={`border-silver text-jet dark:bg-raisin-black dark:text-bright-gray dark:placeholder:text-silver mt-2 w-full rounded-3xl border bg-white px-5 py-3 text-sm outline-hidden placeholder:text-gray-400 dark:border-[#7E7E7E] ${
                       !agent.limited_token_mode
                         ? 'cursor-not-allowed opacity-50'
@@ -977,10 +1153,11 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
                 <div className="mt-6">
                   <div className="flex items-center justify-between">
                     <div>
-                      <h2 className="text-sm font-medium">Request limiting</h2>
+                      <h2 className="text-sm font-medium">
+                        {t('agents.form.advanced.requestLimiting')}
+                      </h2>
                       <p className="mt-1 text-xs text-gray-600 dark:text-gray-400">
-                        Limit daily total requests that can be made to this
-                        agent
+                        {t('agents.form.advanced.requestLimitingDescription')}
                       </p>
                     </div>
                     <button
@@ -1020,7 +1197,9 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
                       })
                     }
                     disabled={!agent.limited_request_mode}
-                    placeholder="Enter request limit"
+                    placeholder={t(
+                      'agents.form.placeholders.enterRequestLimit',
+                    )}
                     className={`border-silver text-jet dark:bg-raisin-black dark:text-bright-gray dark:placeholder:text-silver mt-2 w-full rounded-3xl border bg-white px-5 py-3 text-sm outline-hidden placeholder:text-gray-400 dark:border-[#7E7E7E] ${
                       !agent.limited_request_mode
                         ? 'cursor-not-allowed opacity-50'
@@ -1032,21 +1211,25 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
             )}
           </div>
         </div>
-        <div className="col-span-3 flex flex-col gap-3 rounded-[30px] bg-[#F6F6F6] px-6 py-3 dark:bg-[#383838] dark:text-[#E0E0E0]">
-          <h2 className="text-lg font-semibold">Preview</h2>
-          <AgentPreviewArea />
+        <div className="col-span-3 flex flex-col gap-2 max-[1179px]:h-auto max-[1179px]:px-0 max-[1179px]:py-0 min-[1180px]:h-full min-[1180px]:py-2 dark:text-[#E0E0E0]">
+          <h2 className="text-lg font-semibold">
+            {t('agents.form.sections.preview')}
+          </h2>
+          <div className="flex-1 max-[1179px]:overflow-visible min-[1180px]:min-h-0 min-[1180px]:overflow-hidden">
+            <AgentPreviewArea />
+          </div>
         </div>
       </div>
       <ConfirmationModal
-        message="Are you sure you want to delete this agent?"
+        message={t('agents.deleteConfirmation')}
         modalState={deleteConfirmation}
         setModalState={setDeleteConfirmation}
-        submitLabel="Delete"
+        submitLabel={t('agents.form.buttons.delete')}
         handleSubmit={() => {
           handleDelete(agent.id || '');
           setDeleteConfirmation('INACTIVE');
         }}
-        cancelLabel="Cancel"
+        cancelLabel={t('agents.form.buttons.cancel')}
         variant="danger"
       />
       <AgentDetailsModal
@@ -1069,18 +1252,19 @@ export default function NewAgent({ mode }: { mode: 'new' | 'edit' | 'draft' }) {
 }
 
 function AgentPreviewArea() {
+  const { t } = useTranslation();
   const selectedAgent = useSelector(selectSelectedAgent);
   return (
-    <div className="dark:bg-raisin-black h-full w-full rounded-[30px] border border-[#F6F6F6] bg-white max-[1180px]:h-192 dark:border-[#7E7E7E]">
+    <div className="dark:bg-raisin-black w-full rounded-[30px] border border-[#F6F6F6] bg-white max-[1179px]:h-[600px] min-[1180px]:h-full dark:border-[#7E7E7E]">
       {selectedAgent?.status === 'published' ? (
-        <div className="flex h-full w-full flex-col justify-end overflow-auto rounded-[30px]">
+        <div className="flex h-full w-full flex-col overflow-hidden rounded-[30px]">
           <AgentPreview />
         </div>
       ) : (
         <div className="flex h-full w-full flex-col items-center justify-center gap-2">
           <span className="block h-12 w-12 bg-[url('/src/assets/science-spark.svg')] bg-contain bg-center bg-no-repeat transition-all dark:bg-[url('/src/assets/science-spark-dark.svg')]" />{' '}
           <p className="dark:text-gray-4000 text-xs text-[#18181B]">
-            Published agents can be previewed here
+            {t('agents.form.preview.publishedPreview')}
           </p>
         </div>
       )}

frontend/src/agents/SharedAgent.tsx

@@ -144,7 +144,7 @@ export default function SharedAgent() {
             className="mx-auto mb-6 h-32 w-32"
           />
           <p className="dark:text-gray-4000 text-center text-lg text-[#71717A]">
-            No agent found. Please ensure the agent is shared.
+            {t('agents.shared.notFound')}
           </p>
         </div>
       </div>
@@ -177,13 +177,15 @@ export default function SharedAgent() {
           />
         </div>
         <div className="flex w-[95%] max-w-[1500px] flex-col items-center pb-2 md:w-9/12 lg:w-8/12 xl:w-8/12 2xl:w-6/12">
-          <MessageInput
-            onSubmit={(text) => handleQuestionSubmission(text)}
-            loading={status === 'loading'}
-            showSourceButton={sharedAgent ? false : true}
-            showToolButton={sharedAgent ? false : true}
-            autoFocus={false}
-          />
+          <div className="w-full px-2">
+            <MessageInput
+              onSubmit={(text) => handleQuestionSubmission(text)}
+              loading={status === 'loading'}
+              showSourceButton={sharedAgent ? false : true}
+              showToolButton={sharedAgent ? false : true}
+              autoFocus={false}
+            />
+          </div>
           <p className="text-gray-4000 dark:text-sonic-silver hidden w-screen self-center bg-transparent py-2 text-center text-xs md:inline md:w-full">
             {t('tagline')}
           </p>

frontend/src/agents/agentPreviewSlice.ts

@@ -52,6 +52,10 @@ export const fetchPreviewAnswer = createAsyncThunk<
     }
 
     if (state.preference) {
+      const modelId =
+        state.preference.selectedAgent?.default_model_id ||
+        state.preference.selectedModel?.id;
+
       if (API_STREAMING) {
         await handleFetchAnswerSteaming(
           question,
@@ -120,22 +124,23 @@ export const fetchPreviewAnswer = createAsyncThunk<
           indx,
           state.preference.selectedAgent?.id,
           attachmentIds,
-          false, // Don't save preview conversations
+          false,
+          modelId,
         );
       } else {
-        // Non-streaming implementation
         const answer = await handleFetchAnswer(
           question,
           signal,
           state.preference.token,
           state.preference.selectedDocs,
-          null, // No conversation ID for previews
+          null,
           state.preference.prompt.id,
           state.preference.chunks,
           state.preference.token_limit,
           state.preference.selectedAgent?.id,
           attachmentIds,
-          false, // Don't save preview conversations
+          false,
+          modelId,
         );
 
         if (answer) {

frontend/src/agents/types/index.ts

@@ -32,4 +32,6 @@ export type Agent = {
   token_limit?: number;
   limited_request_mode?: boolean;
   request_limit?: number;
+  models?: string[];
+  default_model_id?: string;
 };

frontend/src/api/endpoints.ts

@@ -2,6 +2,7 @@ const endpoints = {
   USER: {
     CONFIG: '/api/config',
     NEW_TOKEN: '/api/generate_token',
+    MODELS: '/api/models',
     DOCS: '/api/sources',
     DOCS_PAGINATED: '/api/sources/paginated',
     API_KEYS: '/api/get_api_keys',

frontend/src/api/services/modelService.ts

@@ -0,0 +1,25 @@
+import apiClient from '../client';
+import endpoints from '../endpoints';
+
+import type { AvailableModel, Model } from '../../models/types';
+
+const modelService = {
+  getModels: (token: string | null): Promise<Response> =>
+    apiClient.get(endpoints.USER.MODELS, token, {}),
+
+  transformModels: (models: AvailableModel[]): Model[] =>
+    models.map((model) => ({
+      id: model.id,
+      value: model.id,
+      provider: model.provider,
+      display_name: model.display_name,
+      description: model.description,
+      context_window: model.context_window,
+      supported_attachment_types: model.supported_attachment_types,
+      supports_tools: model.supports_tools,
+      supports_structured_output: model.supports_structured_output,
+      supports_streaming: model.supports_streaming,
+    })),
+};
+
+export default modelService;

frontend/src/assets/book.svg

@@ -0,0 +1,3 @@
+<svg width="12" height="14" viewBox="0 0 12 14" fill="none" xmlns="http://www.w3.org/2000/svg">
+<path d="M10.2857 14H2.57143C1.15179 14 0 12.8242 0 11.375V2.625C0 1.17578 1.15179 0 2.57143 0H10.7143C11.4241 0 12 0.587891 12 1.3125V9.1875C12 9.75898 11.6411 10.2457 11.1429 10.4262V12.25C11.617 12.25 12 12.641 12 13.125C12 13.609 11.617 14 11.1429 14H10.2857ZM2.57143 10.5C2.09732 10.5 1.71429 10.891 1.71429 11.375C1.71429 11.859 2.09732 12.25 2.57143 12.25H9.42857V10.5H2.57143ZM3.42857 4.15625C3.42857 4.51992 3.71518 4.8125 4.07143 4.8125H8.78571C9.14196 4.8125 9.42857 4.51992 9.42857 4.15625C9.42857 3.79258 9.14196 3.5 8.78571 3.5H4.07143C3.71518 3.5 3.42857 3.79258 3.42857 4.15625ZM4.07143 6.125C3.71518 6.125 3.42857 6.41758 3.42857 6.78125C3.42857 7.14492 3.71518 7.4375 4.07143 7.4375H8.78571C9.14196 7.4375 9.42857 7.14492 9.42857 6.78125C9.42857 6.41758 9.14196 6.125 8.78571 6.125H4.07143Z" fill="#6A4DF4"/>
+</svg>

frontend/src/assets/expand.svg

@@ -1,4 +0,0 @@
-<svg width="27" height="26" viewBox="0 0 27 26" fill="none" xmlns="http://www.w3.org/2000/svg">
-<path d="M4.03371 5.27275L4.1915 20.9162C4.20021 21.7802 4.90766 22.4735 5.77162 22.4648L21.4151 22.307C22.2791 22.2983 22.9724 21.5909 22.9637 20.7269L22.8059 5.0834C22.7972 4.21944 22.0897 3.52612 21.2258 3.53483L5.58228 3.69262C4.71831 3.70134 4.02499 4.40878 4.03371 5.27275Z" stroke="#949494" stroke-width="2.08591" stroke-linejoin="round"/>
-<path d="M9.42289 22.428L9.23354 3.65585M17.6924 15.0436L15.5856 12.9788L17.6504 10.872M6.29419 22.4596L12.5516 22.3965M6.10484 3.68741L12.3622 3.62429" stroke="#949494" stroke-width="2.08591" stroke-linecap="round" stroke-linejoin="round"/>
-</svg>

frontend/src/assets/no-files.svg

@@ -1,5 +1,5 @@
 <svg width="113" height="124" viewBox="0 0 113 124" fill="none" xmlns="http://www.w3.org/2000/svg">
-<circle cx="55.5" cy="71" r="53" fill="#F1F1F1" fill-opacity="0.5"/>
+<circle cx="55.5" cy="71" r="53" fill="#E8E3F3" fill-opacity="0.6"/>
 <rect x="-0.599797" y="0.654564" width="43.9445" height="61.5222" rx="4.39444" transform="matrix(-0.999048 0.0436194 0.0436194 0.999048 68.9873 43.3176)" fill="#EEEEEE" stroke="#999999" stroke-width="1.25556"/>
 <rect x="0.704349" y="-0.540466" width="46.4556" height="64.0333" rx="5.65" transform="matrix(-0.991445 -0.130526 -0.130526 0.991445 96.3673 40.893)" fill="#FAFAFA" stroke="#999999" stroke-width="1.25556"/>
 <path d="M94.3796 45.7849C94.7417 43.0349 92.8059 40.5122 90.0559 40.1501L55.2011 35.5614C52.4511 35.1994 49.9284 37.1352 49.5663 39.8851L48.3372 49.2212L93.1505 55.121L94.3796 45.7849Z" fill="#EEEEEE"/>

frontend/src/assets/panel-left-close.svg

@@ -0,0 +1 @@
+<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="#949494" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-panel-left-close-icon lucide-panel-left-close"><rect width="18" height="18" x="3" y="3" rx="2"/><path d="M9 3v18"/><path d="m16 15-3-3 3-3"/></svg>

frontend/src/assets/panel-left-open.svg

@@ -0,0 +1 @@
+<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="#949494" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-panel-left-open-icon lucide-panel-left-open"><rect width="18" height="18" x="3" y="3" rx="2"/><path d="M9 3v18"/><path d="m14 9 3 3-3 3"/></svg>

frontend/src/assets/rounded-tick.svg

@@ -0,0 +1,3 @@
+<svg width="20" height="21" viewBox="0 0 20 21" fill="none" xmlns="http://www.w3.org/2000/svg">
+<path d="M10 0.75C4.62391 0.75 0.25 5.12391 0.25 10.5C0.25 15.8761 4.62391 20.25 10 20.25C15.3761 20.25 19.75 15.8761 19.75 10.5C19.75 5.12391 15.3761 0.75 10 0.75ZM15.0742 7.23234L8.77422 14.7323C8.70511 14.8147 8.61912 14.8812 8.52207 14.9273C8.42502 14.9735 8.31918 14.9983 8.21172 15H8.19906C8.09394 15 7.99 14.9778 7.89398 14.935C7.79797 14.8922 7.71202 14.8297 7.64172 14.7516L4.94172 11.7516C4.87315 11.6788 4.81981 11.5931 4.78483 11.4995C4.74986 11.4059 4.73395 11.3062 4.73805 11.2063C4.74215 11.1064 4.76617 11.0084 4.8087 10.9179C4.85124 10.8275 4.91142 10.7464 4.98572 10.6796C5.06002 10.6127 5.14694 10.5614 5.24136 10.5286C5.33579 10.4958 5.43581 10.4822 5.53556 10.4886C5.63531 10.495 5.73277 10.5213 5.82222 10.5659C5.91166 10.6106 5.99128 10.6726 6.05641 10.7484L8.17938 13.1072L13.9258 6.26766C14.0547 6.11863 14.237 6.02631 14.4335 6.01066C14.6299 5.99501 14.8246 6.05728 14.9754 6.18402C15.1263 6.31075 15.2212 6.49176 15.2397 6.68793C15.2582 6.8841 15.1988 7.07966 15.0742 7.23234Z" fill="#B5B5B5"/>
+</svg>

frontend/src/components/ActionButtons.tsx

@@ -45,7 +45,7 @@ export default function ActionButtons({
       <div className={`flex items-center gap-2 sm:gap-4 ${className}`}>
         {showNewChat && (
           <button
-            title="Open New Chat"
+            title={t('actionButtons.openNewChat')}
             onClick={newChat}
             className="hover:bg-bright-gray flex items-center gap-1 rounded-full p-2 lg:hidden dark:hover:bg-[#28292E]"
           >
@@ -62,7 +62,7 @@ export default function ActionButtons({
         {showShare && conversationId && (
           <>
             <button
-              title="Share"
+              title={t('actionButtons.share')}
               onClick={() => setShareModalState(true)}
               className="hover:bg-bright-gray rounded-full p-2 dark:hover:bg-[#28292E]"
             >

frontend/src/components/Chunks.tsx

@@ -136,33 +136,34 @@ const Chunks: React.FC<ChunksProps> = ({
 
   const pathParts = path ? path.split('/') : [];
 
-  const fetchChunks = () => {
+  const fetchChunks = async () => {
     setLoading(true);
     try {
-      userService
-        .getDocumentChunks(documentId, page, perPage, token, path, searchTerm)
-        .then((response) => {
-          if (!response.ok) {
-            setLoading(false);
-            setPaginatedChunks([]);
-            throw new Error('Failed to fetch chunks data');
-          }
-          return response.json();
-        })
-        .then((data) => {
-          setPage(data.page);
-          setPerPage(data.per_page);
-          setTotalChunks(data.total);
-          setPaginatedChunks(data.chunks);
-          setLoading(false);
-        })
-        .catch((error) => {
-          setLoading(false);
-          setPaginatedChunks([]);
-        });
-    } catch (e) {
-      setLoading(false);
+      const response = await userService.getDocumentChunks(
+        documentId,
+        page,
+        perPage,
+        token,
+        path,
+        searchTerm,
+      );
+
+      if (!response.ok) {
+        throw new Error('Failed to fetch chunks data');
+      }
+
+      const data = await response.json();
+
+      setPage(data.page);
+      setPerPage(data.per_page);
+      setTotalChunks(data.total);
+      setPaginatedChunks(data.chunks);
+    } catch (error) {
       setPaginatedChunks([]);
+      console.error(error);
+    } finally {
+      // ✅ always runs, success or failure
+      setLoading(false);
     }
   };
 

frontend/src/components/ConnectorTree.tsx

@@ -38,7 +38,7 @@ interface DirectoryStructure {
   [key: string]: FileNode;
 }
 
-interface ConnectorTreeComponentProps {
+interface ConnectorTreeProps {
   docId: string;
   sourceName: string;
   onBackToDocuments: () => void;
@@ -50,7 +50,7 @@ interface SearchResult {
   isFile: boolean;
 }
 
-const ConnectorTreeComponent: React.FC<ConnectorTreeComponentProps> = ({
+const ConnectorTree: React.FC<ConnectorTreeProps> = ({
   docId,
   sourceName,
   onBackToDocuments,
@@ -744,4 +744,4 @@ const ConnectorTreeComponent: React.FC<ConnectorTreeComponentProps> = ({
   );
 };
 
-export default ConnectorTreeComponent;
+export default ConnectorTree;

frontend/src/components/CopyButton.tsx

@@ -20,14 +20,9 @@ type CopyButtonProps = {
 const DEFAULT_ICON_SIZE = 'w-4 h-4';
 const DEFAULT_PADDING = 'p-2';
 const DEFAULT_COPIED_DURATION = 2000;
-const DEFAULT_BG_LIGHT = '#FFFFFF';
-const DEFAULT_BG_DARK = 'transparent';
-const DEFAULT_HOVER_BG_LIGHT = '#EEEEEE';
-const DEFAULT_HOVER_BG_DARK = '#464152';
 
 export default function CopyButton({
   textToCopy,
-
   iconSize = DEFAULT_ICON_SIZE,
   padding = DEFAULT_PADDING,
   showText = false,
@@ -43,9 +38,8 @@ export default function CopyButton({
   const iconWrapperClasses = clsx(
     'flex items-center justify-center rounded-full transition-colors duration-150 ease-in-out',
     padding,
-    `bg-[${DEFAULT_BG_LIGHT}] dark:bg-[${DEFAULT_BG_DARK}]`,
     {
-      [`hover:bg-[${DEFAULT_HOVER_BG_LIGHT}] dark:hover:bg-[${DEFAULT_HOVER_BG_DARK}]`]:
+      [`bg-[#FFFFFF}] dark:bg-transparent hover:bg-[#EEEEEE] dark:hover:bg-purple-taupe`]:
         !isCopied,
       'bg-green-100 dark:bg-green-900 hover:bg-green-100 dark:hover:bg-green-900':
         isCopied,

frontend/src/components/Dropdown.tsx

@@ -60,7 +60,7 @@ function Dropdown<T extends DropdownOption>({
         }`}
       >
         {typeof selectedValue === 'string' ? (
-          <span className="dark:text-bright-gray truncate">
+          <span className={`dark:text-bright-gray truncate ${contentSize}`}>
             {selectedValue}
           </span>
         ) : (

frontend/src/components/DropdownModel.tsx

@@ -0,0 +1,138 @@
+import React, { useEffect } from 'react';
+import { useDispatch, useSelector } from 'react-redux';
+
+import modelService from '../api/services/modelService';
+import Arrow2 from '../assets/dropdown-arrow.svg';
+import RoundedTick from '../assets/rounded-tick.svg';
+import {
+  selectAvailableModels,
+  selectSelectedModel,
+  setAvailableModels,
+  setModelsLoading,
+  setSelectedModel,
+} from '../preferences/preferenceSlice';
+
+import type { Model } from '../models/types';
+
+export default function DropdownModel() {
+  const dispatch = useDispatch();
+  const selectedModel = useSelector(selectSelectedModel);
+  const availableModels = useSelector(selectAvailableModels);
+  const dropdownRef = React.useRef<HTMLDivElement>(null);
+  const [isOpen, setIsOpen] = React.useState(false);
+
+  useEffect(() => {
+    const loadModels = async () => {
+      if ((availableModels?.length ?? 0) > 0) {
+        return;
+      }
+      dispatch(setModelsLoading(true));
+      try {
+        const response = await modelService.getModels(null);
+        if (!response.ok) {
+          throw new Error(`API error: ${response.status}`);
+        }
+        const data = await response.json();
+        const models = data.models || [];
+        const transformed = modelService.transformModels(models);
+
+        dispatch(setAvailableModels(transformed));
+        if (!selectedModel && transformed.length > 0) {
+          const defaultModel =
+            transformed.find((m) => m.id === data.default_model_id) ||
+            transformed[0];
+          dispatch(setSelectedModel(defaultModel));
+        } else if (selectedModel && transformed.length > 0) {
+          const isValid = transformed.find((m) => m.id === selectedModel.id);
+          if (!isValid) {
+            const defaultModel =
+              transformed.find((m) => m.id === data.default_model_id) ||
+              transformed[0];
+            dispatch(setSelectedModel(defaultModel));
+          }
+        }
+      } catch (error) {
+        console.error('Failed to load models:', error);
+      } finally {
+        dispatch(setModelsLoading(false));
+      }
+    };
+
+    loadModels();
+  }, [availableModels?.length, dispatch, selectedModel]);
+
+  const handleClickOutside = (event: MouseEvent) => {
+    if (
+      dropdownRef.current &&
+      !dropdownRef.current.contains(event.target as Node)
+    ) {
+      setIsOpen(false);
+    }
+  };
+
+  useEffect(() => {
+    document.addEventListener('mousedown', handleClickOutside);
+    return () => {
+      document.removeEventListener('mousedown', handleClickOutside);
+    };
+  }, []);
+
+  return (
+    <div ref={dropdownRef}>
+      <div
+        className={`bg-gray-1000 dark:bg-dark-charcoal mx-auto flex w-full cursor-pointer justify-between p-1 dark:text-white ${isOpen ? 'rounded-t-3xl' : 'rounded-3xl'}`}
+        onClick={() => setIsOpen(!isOpen)}
+      >
+        {selectedModel?.display_name ? (
+          <p className="mx-4 my-3 truncate overflow-hidden whitespace-nowrap">
+            {selectedModel.display_name}
+          </p>
+        ) : (
+          <p className="mx-4 my-3 truncate overflow-hidden whitespace-nowrap">
+            Select Model
+          </p>
+        )}
+        <img
+          src={Arrow2}
+          alt="arrow"
+          className={`${
+            isOpen ? 'rotate-360' : 'rotate-270'
+          } mr-3 w-3 transition-all select-none`}
+        />
+      </div>
+      {isOpen && (
+        <div className="no-scrollbar dark:bg-dark-charcoal absolute right-0 left-0 z-20 -mt-1 max-h-52 w-full overflow-y-auto rounded-b-3xl bg-white shadow-md">
+          {availableModels && (availableModels?.length ?? 0) > 0 ? (
+            availableModels.map((model: Model) => (
+              <div
+                key={model.id}
+                onClick={() => {
+                  dispatch(setSelectedModel(model));
+                  setIsOpen(false);
+                }}
+                className={`border-gray-3000/75 dark:border-purple-taupe/50 hover:bg-gray-3000/75 dark:hover:bg-purple-taupe flex h-10 w-full cursor-pointer items-center justify-between border-t`}
+              >
+                <div className="flex w-full items-center justify-between">
+                  <p className="overflow-hidden py-3 pr-2 pl-5 overflow-ellipsis whitespace-nowrap">
+                    {model.display_name}
+                  </p>
+                  {model.id === selectedModel?.id ? (
+                    <img
+                      src={RoundedTick}
+                      alt="selected"
+                      className="mr-3.5 h-4 w-4"
+                    />
+                  ) : null}
+                </div>
+              </div>
+            ))
+          ) : (
+            <div className="h-10 w-full border-x-2 border-b-2">
+              <p className="ml-5 py-3 text-gray-500">No models available</p>
+            </div>
+          )}
+        </div>
+      )}
+    </div>
+  );
+}

frontend/src/components/FilePicker.tsx

@@ -1,4 +1,5 @@
 import React, { useState, useEffect, useCallback, useRef } from 'react';
+import { useTranslation } from 'react-i18next';
 import { formatBytes } from '../utils/stringUtils';
 import { formatDate } from '../utils/dateTimeUtils';
 import {
@@ -66,6 +67,7 @@ export const FilePicker: React.FC<CloudFilePickerProps> = ({
     );
   };
 
+  const { t } = useTranslation();
   const [files, setFiles] = useState<CloudFile[]>([]);
   const [selectedFiles, setSelectedFiles] =
     useState<string[]>(initialSelectedFiles);
@@ -417,7 +419,7 @@ export const FilePicker: React.FC<CloudFilePickerProps> = ({
               <div className="mb-3 max-w-md">
                 <Input
                   type="text"
-                  placeholder="Search files and folders..."
+                  placeholder={t('filePicker.searchPlaceholder')}
                   value={searchQuery}
                   onChange={(e) => handleSearchChange(e.target.value)}
                   colorVariant="silver"
@@ -431,7 +433,9 @@ export const FilePicker: React.FC<CloudFilePickerProps> = ({
 
               {/* Selected Files Message */}
               <div className="pb-3 text-sm text-gray-600 dark:text-gray-400">
-                {selectedFiles.length + selectedFolders.length} selected
+                {t('filePicker.itemsSelected', {
+                  count: selectedFiles.length + selectedFolders.length,
+                })}
               </div>
             </div>
 
@@ -448,9 +452,15 @@ export const FilePicker: React.FC<CloudFilePickerProps> = ({
                       <TableHead>
                         <TableRow>
                           <TableHeader width="40px"></TableHeader>
-                          <TableHeader width="60%">Name</TableHeader>
-                          <TableHeader width="20%">Last Modified</TableHeader>
-                          <TableHeader width="20%">Size</TableHeader>
+                          <TableHeader width="60%">
+                            {t('filePicker.name')}
+                          </TableHeader>
+                          <TableHeader width="20%">
+                            {t('filePicker.lastModified')}
+                          </TableHeader>
+                          <TableHeader width="20%">
+                            {t('filePicker.size')}
+                          </TableHeader>
                         </TableRow>
                       </TableHead>
                       <TableBody>

frontend/src/components/FileTree.tsx

@@ -36,7 +36,7 @@ interface DirectoryStructure {
   [key: string]: FileNode;
 }
 
-interface FileTreeComponentProps {
+interface FileTreeProps {
   docId: string;
   sourceName: string;
   onBackToDocuments: () => void;
@@ -48,7 +48,7 @@ interface SearchResult {
   isFile: boolean;
 }
 
-const FileTreeComponent: React.FC<FileTreeComponentProps> = ({
+const FileTree: React.FC<FileTreeProps> = ({
   docId,
   sourceName,
   onBackToDocuments,
@@ -871,4 +871,4 @@ const FileTreeComponent: React.FC<FileTreeComponentProps> = ({
   );
 };
 
-export default FileTreeComponent;
+export default FileTree;

frontend/src/components/FileUpload.tsx

@@ -1,4 +1,5 @@
 import React, { useCallback, useState } from 'react';
+import { useTranslation } from 'react-i18next';
 import { useDropzone } from 'react-dropzone';
 import { twMerge } from 'tailwind-merge';
 
@@ -44,13 +45,14 @@ export const FileUpload = ({
   activeClassName = 'border-blue-500 bg-blue-50',
   acceptClassName = 'border-green-500 dark:border-green-500 bg-green-50 dark:bg-green-50/10',
   rejectClassName = 'border-red-500 bg-red-50 dark:bg-red-500/10 dark:border-red-500',
-  uploadText = 'Click to upload or drag and drop',
-  dragActiveText = 'Drop the files here',
-  fileTypeText = 'PNG, JPG, JPEG up to',
-  sizeLimitText = 'MB',
+  uploadText,
+  dragActiveText,
+  fileTypeText,
+  sizeLimitText,
   disabled = false,
   validator,
 }: FileUploadProps) => {
+  const { t } = useTranslation();
   const [errors, setErrors] = useState<string[]>([]);
   const [preview, setPreview] = useState<string | null>(null);
   const [currentFile, setCurrentFile] = useState<File | null>(null);
@@ -71,7 +73,9 @@ export const FileUpload = ({
     if (file.size > maxSize) {
       return {
         isValid: false,
-        error: `File exceeds ${maxSize / 1024 / 1024}MB limit`,
+        error: t('components.fileUpload.fileSizeError', {
+          size: maxSize / 1024 / 1024,
+        }),
       };
     }
 
@@ -178,7 +182,11 @@ export const FileUpload = ({
         </p>
       );
     }
-    return <p className="text-sm font-semibold">{uploadText}</p>;
+    return (
+      <p className="text-sm font-semibold">
+        {uploadText || t('components.fileUpload.clickToUpload')}
+      </p>
+    );
   };
 
   const defaultContent = (
@@ -196,14 +204,17 @@ export const FileUpload = ({
       <div className="text-center">
         <div className="text-sm font-medium">
           {isDragActive ? (
-            <p className="text-sm font-semibold">{dragActiveText}</p>
+            <p className="text-sm font-semibold">
+              {dragActiveText || t('components.fileUpload.dropFiles')}
+            </p>
           ) : (
             renderUploadText()
           )}
         </div>
         <p className="mt-1 text-xs text-[#A3A3A3]">
-          {fileTypeText} {maxSize / 1024 / 1024}
-          {sizeLimitText}
+          {fileTypeText || t('components.fileUpload.fileTypes')}{' '}
+          {maxSize / 1024 / 1024}
+          {sizeLimitText || t('components.fileUpload.sizeLimitUnit')}
         </p>
       </div>
     </div>

frontend/src/components/Input.tsx

@@ -20,6 +20,7 @@ const Input = ({
   onChange,
   onPaste,
   onKeyDown,
+  edgeRoundness = 'rounded-full',
 }: InputProps) => {
   const colorStyles = {
     silver: 'border-silver dark:border-silver/40',
@@ -43,7 +44,7 @@ const Input = ({
     <div className={`relative ${className}`}>
       <input
         ref={inputRef}
-        className={`peer text-jet dark:text-bright-gray h-[42px] w-full rounded-full bg-transparent ${leftIcon ? 'pl-10' : 'px-3'} py-1 placeholder-transparent outline-hidden ${colorStyles[colorVariant]} ${borderStyles[borderVariant]} ${textSizeStyles[textSize]} [&:-webkit-autofill]:appearance-none [&:-webkit-autofill]:bg-transparent [&:-webkit-autofill_selected]:bg-transparent`}
+        className={`peer text-jet dark:text-bright-gray h-[42px] w-full ${edgeRoundness} bg-transparent ${leftIcon ? 'pl-10' : 'px-3'} py-1 placeholder-transparent outline-hidden ${colorStyles[colorVariant]} ${borderStyles[borderVariant]} ${textSizeStyles[textSize]} [&:-webkit-autofill]:appearance-none [&:-webkit-autofill]:bg-transparent [&:-webkit-autofill_selected]:bg-transparent`}
         type={type}
         id={id}
         name={name}

frontend/src/components/MermaidRenderer.tsx

@@ -1,4 +1,5 @@
 import React, { useEffect, useRef, useState } from 'react';
+import { useTranslation } from 'react-i18next';
 import mermaid from 'mermaid';
 import CopyButton from './CopyButton';
 import { Prism as SyntaxHighlighter } from 'react-syntax-highlighter';
@@ -15,6 +16,7 @@ const MermaidRenderer: React.FC<MermaidRendererProps> = ({
   code,
   isLoading,
 }) => {
+  const { t } = useTranslation();
   const [isDarkTheme] = useDarkTheme();
   const diagramId = useRef(
     `mermaid-${Date.now()}-${Math.random().toString(36).substring(2)}`,
@@ -273,7 +275,7 @@ const MermaidRenderer: React.FC<MermaidRendererProps> = ({
               <button
                 onClick={() => setShowDownloadMenu(!showDownloadMenu)}
                 className="flex h-full items-center rounded-sm bg-gray-100 px-2 py-1 text-xs dark:bg-gray-700"
-                title="Download options"
+                title={t('mermaid.downloadOptions')}
               >
                 Download <span className="ml-1">▼</span>
               </button>
@@ -307,7 +309,7 @@ const MermaidRenderer: React.FC<MermaidRendererProps> = ({
                   ? 'bg-blue-200 dark:bg-blue-800'
                   : 'bg-gray-100 dark:bg-gray-700'
               }`}
-              title="View Code"
+              title={t('mermaid.viewCode')}
             >
               Code
             </button>
@@ -353,7 +355,7 @@ const MermaidRenderer: React.FC<MermaidRendererProps> = ({
                       setZoomFactor((prev) => Math.max(1, prev - 0.5))
                     }
                     className="rounded px-1 hover:bg-gray-600"
-                    title="Decrease zoom"
+                    title={t('mermaid.decreaseZoom')}
                   >
                     -
                   </button>
@@ -362,7 +364,7 @@ const MermaidRenderer: React.FC<MermaidRendererProps> = ({
                     onClick={() => {
                       setZoomFactor(2);
                     }}
-                    title="Reset zoom"
+                    title={t('mermaid.resetZoom')}
                   >
                     {zoomFactor.toFixed(1)}x
                   </span>
@@ -371,7 +373,7 @@ const MermaidRenderer: React.FC<MermaidRendererProps> = ({
                       setZoomFactor((prev) => Math.min(6, prev + 0.5))
                     }
                     className="rounded px-1 hover:bg-gray-600"
-                    title="Increase zoom"
+                    title={t('mermaid.increaseZoom')}
                   >
                     +
                   </button>

frontend/src/components/MessageInput.tsx

@@ -1,4 +1,6 @@
-import { useEffect, useRef, useState } from 'react';
+import { useCallback, useEffect, useRef, useState } from 'react';
+import { createPortal } from 'react-dom';
+import { useDropzone } from 'react-dropzone';
 import { useTranslation } from 'react-i18next';
 import { useDispatch, useSelector } from 'react-redux';
 
@@ -6,6 +8,7 @@ import endpoints from '../api/endpoints';
 import userService from '../api/services/userService';
 import AlertIcon from '../assets/alert.svg';
 import ClipIcon from '../assets/clip.svg';
+import DragFileUpload from '../assets/DragFileUpload.svg';
 import ExitIcon from '../assets/exit.svg';
 import SendArrowIcon from './SendArrowIcon';
 import SourceIcon from '../assets/source.svg';
@@ -16,6 +19,7 @@ import {
   removeAttachment,
   selectAttachments,
   updateAttachment,
+  reorderAttachments,
 } from '../upload/uploadSlice';
 
 import { ActiveState } from '../models/misc';
@@ -53,6 +57,7 @@ export default function MessageInput({
   const [isToolsPopupOpen, setIsToolsPopupOpen] = useState(false);
   const [uploadModalState, setUploadModalState] =
     useState<ActiveState>('INACTIVE');
+  const [handleDragActive, setHandleDragActive] = useState<boolean>(false);
 
   const selectedDocs = useSelector(selectSelectedDocs);
   const token = useSelector(selectToken);
@@ -72,7 +77,7 @@ export default function MessageInput({
         (browserOS === 'mac' && event.metaKey && event.key === 'k')
       ) {
         event.preventDefault();
-        setIsSourcesPopupOpen(!isSourcesPopupOpen);
+        setIsSourcesPopupOpen((s) => !s);
       }
     };
 
@@ -82,77 +87,360 @@ export default function MessageInput({
     };
   }, [browserOS]);
 
-  const handleFileAttachment = (e: React.ChangeEvent<HTMLInputElement>) => {
-    if (!e.target.files || e.target.files.length === 0) return;
+  const uploadFiles = useCallback(
+    (files: File[]) => {
+      if (!files || files.length === 0) return;
 
-    const file = e.target.files[0];
-    const formData = new FormData();
-    formData.append('file', file);
+      const apiHost = import.meta.env.VITE_API_HOST;
 
-    const apiHost = import.meta.env.VITE_API_HOST;
-    const xhr = new XMLHttpRequest();
+      if (files.length > 1) {
+        const formData = new FormData();
+        const indexToUiId: Record<number, string> = {};
 
-    const newAttachment = {
-      fileName: file.name,
-      progress: 0,
-      status: 'uploading' as const,
-      taskId: '',
-    };
-
-    dispatch(addAttachment(newAttachment));
-
-    xhr.upload.addEventListener('progress', (event) => {
-      if (event.lengthComputable) {
-        const progress = Math.round((event.loaded / event.total) * 100);
-        dispatch(
-          updateAttachment({
-            taskId: newAttachment.taskId,
-            updates: { progress },
-          }),
-        );
-      }
-    });
-
-    xhr.onload = () => {
-      if (xhr.status === 200) {
-        const response = JSON.parse(xhr.responseText);
-        if (response.task_id) {
+        files.forEach((file, i) => {
+          formData.append('file', file);
+          const uiId = crypto.randomUUID();
+          indexToUiId[i] = uiId;
           dispatch(
-            updateAttachment({
-              taskId: newAttachment.taskId,
-              updates: {
-                taskId: response.task_id,
-                status: 'processing',
-                progress: 10,
-              },
+            addAttachment({
+              id: uiId,
+              fileName: file.name,
+              progress: 0,
+              status: 'uploading' as const,
+              taskId: '',
             }),
           );
-        }
-      } else {
-        dispatch(
-          updateAttachment({
-            taskId: newAttachment.taskId,
-            updates: { status: 'failed' },
-          }),
-        );
+        });
+
+        const xhr = new XMLHttpRequest();
+
+        xhr.upload.addEventListener('progress', (event) => {
+          if (event.lengthComputable) {
+            const progress = Math.round((event.loaded / event.total) * 100);
+            Object.values(indexToUiId).forEach((uiId) =>
+              dispatch(
+                updateAttachment({
+                  id: uiId,
+                  updates: { progress },
+                }),
+              ),
+            );
+          }
+        });
+
+        xhr.onload = () => {
+          const status = xhr.status;
+          if (status === 200) {
+            try {
+              const response = JSON.parse(xhr.responseText);
+
+              if (Array.isArray(response?.tasks)) {
+                const tasks = response.tasks as Array<{
+                  task_id?: string;
+                  filename?: string;
+                  attachment_id?: string;
+                  path?: string;
+                }>;
+
+                tasks.forEach((t, idx) => {
+                  const uiId = indexToUiId[idx];
+                  if (!uiId) return;
+                  if (t?.task_id) {
+                    dispatch(
+                      updateAttachment({
+                        id: uiId,
+                        updates: {
+                          taskId: t.task_id,
+                          status: 'processing',
+                          progress: 10,
+                        },
+                      }),
+                    );
+                  } else {
+                    dispatch(
+                      updateAttachment({
+                        id: uiId,
+                        updates: { status: 'failed' },
+                      }),
+                    );
+                  }
+                });
+
+                if (tasks.length < files.length) {
+                  for (let i = tasks.length; i < files.length; i++) {
+                    const uiId = indexToUiId[i];
+                    if (uiId) {
+                      dispatch(
+                        updateAttachment({
+                          id: uiId,
+                          updates: { status: 'failed' },
+                        }),
+                      );
+                    }
+                  }
+                }
+              } else if (response?.task_id) {
+                if (files.length === 1) {
+                  const uiId = indexToUiId[0];
+                  if (uiId) {
+                    dispatch(
+                      updateAttachment({
+                        id: uiId,
+                        updates: {
+                          taskId: response.task_id,
+                          status: 'processing',
+                          progress: 10,
+                        },
+                      }),
+                    );
+                  }
+                } else {
+                  console.warn(
+                    'Server returned a single task_id for multiple files. Update backend to return tasks[].',
+                  );
+                  const firstUi = indexToUiId[0];
+                  if (firstUi) {
+                    dispatch(
+                      updateAttachment({
+                        id: firstUi,
+                        updates: {
+                          taskId: response.task_id,
+                          status: 'processing',
+                          progress: 10,
+                        },
+                      }),
+                    );
+                  }
+                  for (let i = 1; i < files.length; i++) {
+                    const uiId = indexToUiId[i];
+                    if (uiId) {
+                      dispatch(
+                        updateAttachment({
+                          id: uiId,
+                          updates: { status: 'failed' },
+                        }),
+                      );
+                    }
+                  }
+                }
+              } else {
+                console.error('Unexpected upload response shape', response);
+                Object.values(indexToUiId).forEach((id) =>
+                  dispatch(
+                    updateAttachment({
+                      id,
+                      updates: { status: 'failed' },
+                    }),
+                  ),
+                );
+              }
+            } catch (err) {
+              console.error(
+                'Failed to parse upload response',
+                err,
+                xhr.responseText,
+              );
+              Object.values(indexToUiId).forEach((id) =>
+                dispatch(
+                  updateAttachment({
+                    id,
+                    updates: { status: 'failed' },
+                  }),
+                ),
+              );
+            }
+          } else {
+            console.error('Upload failed', status, xhr.responseText);
+            Object.values(indexToUiId).forEach((id) =>
+              dispatch(
+                updateAttachment({
+                  id,
+                  updates: { status: 'failed' },
+                }),
+              ),
+            );
+          }
+        };
+
+        xhr.onerror = () => {
+          console.error('Upload network error');
+          Object.values(indexToUiId).forEach((id) =>
+            dispatch(
+              updateAttachment({
+                id,
+                updates: { status: 'failed' },
+              }),
+            ),
+          );
+        };
+
+        xhr.open('POST', `${apiHost}${endpoints.USER.STORE_ATTACHMENT}`);
+        if (token) xhr.setRequestHeader('Authorization', `Bearer ${token}`);
+        xhr.send(formData);
+        return;
       }
-    };
 
-    xhr.onerror = () => {
-      dispatch(
-        updateAttachment({
-          taskId: newAttachment.taskId,
-          updates: { status: 'failed' },
-        }),
-      );
-    };
+      // Single-file path: upload each file individually (original repo behavior)
+      files.forEach((file) => {
+        const formData = new FormData();
+        formData.append('file', file);
+        const xhr = new XMLHttpRequest();
+        const uniqueId = crypto.randomUUID();
 
-    xhr.open('POST', `${apiHost}${endpoints.USER.STORE_ATTACHMENT}`);
-    xhr.setRequestHeader('Authorization', `Bearer ${token}`);
-    xhr.send(formData);
+        const newAttachment = {
+          id: uniqueId,
+          fileName: file.name,
+          progress: 0,
+          status: 'uploading' as const,
+          taskId: '',
+        };
+
+        dispatch(addAttachment(newAttachment));
+
+        xhr.upload.addEventListener('progress', (event) => {
+          if (event.lengthComputable) {
+            const progress = Math.round((event.loaded / event.total) * 100);
+            dispatch(
+              updateAttachment({
+                id: uniqueId,
+                updates: { progress },
+              }),
+            );
+          }
+        });
+
+        xhr.onload = () => {
+          if (xhr.status === 200) {
+            try {
+              const response = JSON.parse(xhr.responseText);
+              if (response.task_id) {
+                dispatch(
+                  updateAttachment({
+                    id: uniqueId,
+                    updates: {
+                      taskId: response.task_id,
+                      status: 'processing',
+                      progress: 10,
+                    },
+                  }),
+                );
+              } else {
+                // If backend returned tasks[] for single-file, handle gracefully:
+                if (
+                  Array.isArray(response?.tasks) &&
+                  response.tasks[0]?.task_id
+                ) {
+                  dispatch(
+                    updateAttachment({
+                      id: uniqueId,
+                      updates: {
+                        taskId: response.tasks[0].task_id,
+                        status: 'processing',
+                        progress: 10,
+                      },
+                    }),
+                  );
+                } else {
+                  dispatch(
+                    updateAttachment({
+                      id: uniqueId,
+                      updates: { status: 'failed' },
+                    }),
+                  );
+                }
+              }
+            } catch (err) {
+              console.error(
+                'Failed to parse upload response',
+                err,
+                xhr.responseText,
+              );
+              dispatch(
+                updateAttachment({
+                  id: uniqueId,
+                  updates: { status: 'failed' },
+                }),
+              );
+            }
+          } else {
+            dispatch(
+              updateAttachment({
+                id: uniqueId,
+                updates: { status: 'failed' },
+              }),
+            );
+          }
+        };
+
+        xhr.onerror = () => {
+          dispatch(
+            updateAttachment({
+              id: uniqueId,
+              updates: { status: 'failed' },
+            }),
+          );
+        };
+
+        xhr.open('POST', `${apiHost}${endpoints.USER.STORE_ATTACHMENT}`);
+        if (token) xhr.setRequestHeader('Authorization', `Bearer ${token}`);
+        xhr.send(formData);
+      });
+    },
+    [dispatch, token],
+  );
+
+  const handleFileAttachment = (e: React.ChangeEvent<HTMLInputElement>) => {
+    if (!e.target.files || e.target.files.length === 0) return;
+    const files = Array.from(e.target.files);
+    uploadFiles(files);
+    // clear input so same file can be selected again
     e.target.value = '';
   };
 
+  // Drag & drop via react-dropzone
+  const onDrop = useCallback(
+    (acceptedFiles: File[]) => {
+      uploadFiles(acceptedFiles);
+      setHandleDragActive(false);
+    },
+    [uploadFiles],
+  );
+
+  const { getRootProps, getInputProps } = useDropzone({
+    onDrop,
+    noClick: true,
+    noKeyboard: true,
+    multiple: true,
+    onDragEnter: () => {
+      setHandleDragActive(true);
+    },
+    onDragLeave: () => {
+      setHandleDragActive(false);
+    },
+    maxSize: 25000000,
+    accept: {
+      'application/pdf': ['.pdf'],
+      'text/plain': ['.txt'],
+      'text/x-rst': ['.rst'],
+      'text/x-markdown': ['.md'],
+      'application/zip': ['.zip'],
+      'application/vnd.openxmlformats-officedocument.wordprocessingml.document':
+        ['.docx'],
+      'application/json': ['.json'],
+      'text/csv': ['.csv'],
+      'text/html': ['.html'],
+      'application/epub+zip': ['.epub'],
+      'application/vnd.openxmlformats-officedocument.spreadsheetml.sheet': [
+        '.xlsx',
+      ],
+      'application/vnd.openxmlformats-officedocument.presentationml.presentation':
+        ['.pptx'],
+      'image/png': ['.png'],
+      'image/jpeg': ['.jpeg'],
+      'image/jpg': ['.jpg'],
+    },
+  });
+
   useEffect(() => {
     const checkTaskStatus = () => {
       const processingAttachments = attachments.filter(
@@ -167,7 +455,7 @@ export default function MessageInput({
             if (data.status === 'SUCCESS') {
               dispatch(
                 updateAttachment({
-                  taskId: attachment.taskId!,
+                  id: attachment.id,
                   updates: {
                     status: 'completed',
                     progress: 100,
@@ -179,14 +467,14 @@ export default function MessageInput({
             } else if (data.status === 'FAILURE') {
               dispatch(
                 updateAttachment({
-                  taskId: attachment.taskId!,
+                  id: attachment.id,
                   updates: { status: 'failed' },
                 }),
               );
             } else if (data.status === 'PROGRESS' && data.result?.current) {
               dispatch(
                 updateAttachment({
-                  taskId: attachment.taskId!,
+                  id: attachment.id,
                   updates: { progress: data.result.current },
                 }),
               );
@@ -195,7 +483,7 @@ export default function MessageInput({
           .catch(() => {
             dispatch(
               updateAttachment({
-                taskId: attachment.taskId!,
+                id: attachment.id,
                 updates: { status: 'failed' },
               }),
             );
@@ -259,90 +547,134 @@ export default function MessageInput({
     handleAbort();
   };
 
+  const [draggingId, setDraggingId] = useState<string | null>(null);
+
+  const findIndexById = (id: string) =>
+    attachments.findIndex((a) => a.id === id);
+
+  const handleDragStart = (e: React.DragEvent, id: string) => {
+    setDraggingId(id);
+    try {
+      e.dataTransfer.setData('text/plain', id);
+      e.dataTransfer.effectAllowed = 'move';
+    } catch (err) {
+      // ignore
+    }
+  };
+
+  const handleDragOver = (e: React.DragEvent) => {
+    e.preventDefault();
+    e.dataTransfer.dropEffect = 'move';
+  };
+
+  const handleDropOn = (e: React.DragEvent, targetId: string) => {
+    e.preventDefault();
+    const sourceId = e.dataTransfer.getData('text/plain');
+    if (!sourceId || sourceId === targetId) return;
+
+    const sourceIndex = findIndexById(sourceId);
+    const destIndex = findIndexById(targetId);
+    if (sourceIndex === -1 || destIndex === -1) return;
+
+    dispatch(reorderAttachments({ sourceIndex, destinationIndex: destIndex }));
+    setDraggingId(null);
+  };
+
   return (
-    <div className="mx-2 flex w-full flex-col">
+    <div {...getRootProps()} className="flex w-full flex-col">
+      {/* react-dropzone input (for drag/drop) */}
+      <input {...getInputProps()} />
+
       <div className="border-dark-gray bg-lotion dark:border-grey relative flex w-full flex-col rounded-[23px] border dark:bg-transparent">
         <div className="flex flex-wrap gap-1.5 px-2 py-2 sm:gap-2 sm:px-3">
-          {attachments.map((attachment, index) => (
-            <div
-              key={index}
-              className={`group dark:text-bright-gray relative flex items-center rounded-xl bg-[#EFF3F4] px-2 py-1 text-[12px] text-[#5D5D5D] sm:px-3 sm:py-1.5 sm:text-[14px] dark:bg-[#393B3D] ${
-                attachment.status !== 'completed' ? 'opacity-70' : 'opacity-100'
-              }`}
-              title={attachment.fileName}
-            >
-              <div className="bg-purple-30 mr-2 items-center justify-center rounded-lg p-[5.5px]">
-                {attachment.status === 'completed' && (
-                  <img
-                    src={DocumentationDark}
-                    alt="Attachment"
-                    className="h-[15px] w-[15px] object-fill"
-                  />
-                )}
-
-                {attachment.status === 'failed' && (
-                  <img
-                    src={AlertIcon}
-                    alt="Failed"
-                    className="h-[15px] w-[15px] object-fill"
-                  />
-                )}
-
-                {(attachment.status === 'uploading' ||
-                  attachment.status === 'processing') && (
-                  <div className="flex h-[15px] w-[15px] items-center justify-center">
-                    <svg className="h-[15px] w-[15px]" viewBox="0 0 24 24">
-                      <circle
-                        className="opacity-0"
-                        cx="12"
-                        cy="12"
-                        r="10"
-                        stroke="transparent"
-                        strokeWidth="4"
-                        fill="none"
-                      />
-                      <circle
-                        className="text-[#ECECF1]"
-                        cx="12"
-                        cy="12"
-                        r="10"
-                        stroke="currentColor"
-                        strokeWidth="4"
-                        fill="none"
-                        strokeDasharray="62.83"
-                        strokeDashoffset={
-                          62.83 * (1 - attachment.progress / 100)
-                        }
-                        transform="rotate(-90 12 12)"
-                      />
-                    </svg>
-                  </div>
-                )}
-              </div>
-
-              <span className="max-w-[120px] truncate font-medium sm:max-w-[150px]">
-                {attachment.fileName}
-              </span>
-
-              <button
-                className="ml-1.5 flex items-center justify-center rounded-full p-1"
-                onClick={() => {
-                  if (attachment.id) {
-                    dispatch(removeAttachment(attachment.id));
-                  } else if (attachment.taskId) {
-                    dispatch(removeAttachment(attachment.taskId));
-                  }
-                }}
-                aria-label={t('conversation.attachments.remove')}
+          {attachments.map((attachment) => {
+            return (
+              <div
+                key={attachment.id}
+                draggable={true}
+                onDragStart={(e) => handleDragStart(e, attachment.id)}
+                onDragOver={handleDragOver}
+                onDrop={(e) => handleDropOn(e, attachment.id)}
+                className={`group dark:text-bright-gray relative flex items-center rounded-xl bg-[#EFF3F4] px-2 py-1 text-[12px] text-[#5D5D5D] sm:px-3 sm:py-1.5 sm:text-[14px] dark:bg-[#393B3D] ${
+                  attachment.status !== 'completed'
+                    ? 'opacity-70'
+                    : 'opacity-100'
+                } ${
+                  draggingId === attachment.id
+                    ? 'ring-dashed opacity-60 ring-2 ring-purple-200'
+                    : ''
+                }`}
+                title={attachment.fileName}
               >
-                <img
-                  src={ExitIcon}
-                  alt={t('conversation.attachments.remove')}
-                  className="h-2.5 w-2.5 filter dark:invert"
-                />
-              </button>
-            </div>
-          ))}
+                <div className="bg-purple-30 mr-2 flex h-8 w-8 items-center justify-center rounded-md p-1">
+                  {attachment.status === 'completed' && (
+                    <img
+                      src={DocumentationDark}
+                      alt="Attachment"
+                      className="h-[15px] w-[15px] object-fill"
+                    />
+                  )}
+
+                  {attachment.status === 'failed' && (
+                    <img
+                      src={AlertIcon}
+                      alt="Failed"
+                      className="h-[15px] w-[15px] object-fill"
+                    />
+                  )}
+
+                  {(attachment.status === 'uploading' ||
+                    attachment.status === 'processing') && (
+                    <div className="flex h-[15px] w-[15px] items-center justify-center">
+                      <svg className="h-[15px] w-[15px]" viewBox="0 0 24 24">
+                        <circle
+                          className="opacity-0"
+                          cx="12"
+                          cy="12"
+                          r="10"
+                          stroke="transparent"
+                          strokeWidth="4"
+                          fill="none"
+                        />
+                        <circle
+                          className="text-[#ECECF1]"
+                          cx="12"
+                          cy="12"
+                          r="10"
+                          stroke="currentColor"
+                          strokeWidth="4"
+                          fill="none"
+                          strokeDasharray="62.83"
+                          strokeDashoffset={
+                            62.83 * (1 - attachment.progress / 100)
+                          }
+                          transform="rotate(-90 12 12)"
+                        />
+                      </svg>
+                    </div>
+                  )}
+                </div>
+
+                <span className="max-w-[120px] truncate font-medium sm:max-w-[150px]">
+                  {attachment.fileName}
+                </span>
+
+                <button
+                  className="ml-1.5 flex items-center justify-center rounded-full p-1"
+                  onClick={() => {
+                    dispatch(removeAttachment(attachment.id));
+                  }}
+                  aria-label={t('conversation.attachments.remove')}
+                >
+                  <img
+                    src={ExitIcon}
+                    alt={t('conversation.attachments.remove')}
+                    className="h-2.5 w-2.5 filter dark:invert"
+                  />
+                </button>
+              </div>
+            );
+          })}
         </div>
 
         <div className="w-full">
@@ -424,6 +756,7 @@ export default function MessageInput({
               <input
                 type="file"
                 className="hidden"
+                multiple
                 onChange={handleFileAttachment}
               />
             </label>
@@ -483,6 +816,20 @@ export default function MessageInput({
           close={() => setUploadModalState('INACTIVE')}
         />
       )}
+
+      {handleDragActive &&
+        createPortal(
+          <div className="dark:bg-gray-alpha/50 pointer-events-none fixed top-0 left-0 z-50 flex size-full flex-col items-center justify-center bg-white/85">
+            <img className="filter dark:invert" src={DragFileUpload} />
+            <span className="text-outer-space dark:text-silver px-2 text-2xl font-bold">
+              {t('modals.uploadDoc.drag.title')}
+            </span>
+            <span className="text-s text-outer-space dark:text-silver w-48 p-2 text-center">
+              {t('modals.uploadDoc.drag.description')}
+            </span>
+          </div>,
+          document.body,
+        )}
     </div>
   );
 }

frontend/src/components/Notification.tsx

@@ -1,3 +1,4 @@
+import { useTranslation } from 'react-i18next';
 import close from '../assets/cross.svg';
 import rightArrow from '../assets/arrow-full-right.svg';
 import bg from '../assets/notification-bg.jpg';
@@ -13,13 +14,14 @@ export default function Notification({
   notificationLink,
   handleCloseNotification,
 }: NotificationProps) {
+  const { t } = useTranslation();
   return (
     <a
       className="absolute right-2 bottom-6 z-20 flex w-3/4 items-center justify-center gap-2 rounded-lg bg-cover bg-center bg-no-repeat px-2 py-4 sm:right-4 md:w-2/5 lg:w-1/3 xl:w-1/4 2xl:w-1/5"
       style={{ backgroundImage: `url(${bg})` }}
       href={notificationLink}
       target="_blank"
-      aria-label="Notification"
+      aria-label={t('notification.ariaLabel')}
       rel="noreferrer"
     >
       <p className="text-white-3000 text-xs leading-6 font-semibold xl:text-sm xl:leading-7">
@@ -31,7 +33,7 @@ export default function Notification({
 
       <button
         className="absolute top-2 right-2 z-30 h-4 w-4 hover:opacity-70"
-        aria-label="Close notification"
+        aria-label={t('notification.closeAriaLabel')}
         onClick={(e) => {
           e.stopPropagation();
           e.preventDefault();

frontend/src/components/SettingsBar.tsx

@@ -24,6 +24,7 @@ interface SettingsBarProps {
 }
 
 const SettingsBar = ({ setActiveTab, activeTab }: SettingsBarProps) => {
+  const { t } = useTranslation();
   const [hiddenGradient, setHiddenGradient] =
     useState<HiddenGradientType>('left');
   const containerRef = useRef<null | HTMLDivElement>(null);
@@ -60,7 +61,7 @@ const SettingsBar = ({ setActiveTab, activeTab }: SettingsBarProps) => {
         <button
           onClick={() => scrollTabs(-1)}
           className="flex h-6 w-6 items-center justify-center rounded-full transition-all hover:bg-gray-200 dark:hover:bg-gray-700"
-          aria-label="Scroll tabs left"
+          aria-label={t('settings.scrollTabsLeft')}
         >
           <img src={ArrowLeft} alt="left-arrow" className="h-3" />
         </button>
@@ -69,7 +70,7 @@ const SettingsBar = ({ setActiveTab, activeTab }: SettingsBarProps) => {
         ref={containerRef}
         className="no-scrollbar flex snap-x flex-nowrap overflow-x-auto scroll-smooth md:space-x-4"
         role="tablist"
-        aria-label="Settings tabs"
+        aria-label={t('settings.tabsAriaLabel')}
       >
         {tabs.map((tab, index) => (
           <button
@@ -93,7 +94,7 @@ const SettingsBar = ({ setActiveTab, activeTab }: SettingsBarProps) => {
         <button
           onClick={() => scrollTabs(1)}
           className="flex h-6 w-6 items-center justify-center rounded-full hover:bg-gray-200 dark:hover:bg-gray-700"
-          aria-label="Scroll tabs right"
+          aria-label={t('settings.scrollTabsRight')}
         >
           <img src={ArrowRight} alt="right-arrow" className="h-3" />
         </button>

frontend/src/components/Sidebar.tsx

@@ -33,7 +33,7 @@ export default function Sidebar({
   return (
     <div ref={sidebarRef} className="h-vh relative">
       <div
-        className={`dark:bg-chinese-black fixed top-0 right-0 z-50 h-full w-72 transform bg-white shadow-xl transition-all duration-300 sm:w-96 ${
+        className={`dark:bg-chinese-black fixed top-0 right-0 z-50 h-full w-64 transform bg-white shadow-xl transition-all duration-300 sm:w-80 ${
           isOpen ? 'translate-x-[10px]' : 'translate-x-full'
         } border-l border-[#9ca3af]/10`}
       >

frontend/src/components/SourcesPopup.tsx

@@ -1,4 +1,5 @@
 import React, { useRef, useEffect, useState, useLayoutEffect } from 'react';
+import { createPortal } from 'react-dom';
 import { useTranslation } from 'react-i18next';
 import { useDispatch, useSelector } from 'react-redux';
 import { Doc } from '../models/misc';
@@ -107,7 +108,7 @@ export default function SourcesPopup({
     onClose();
   };
 
-  return (
+  const popupContent = (
     <div
       ref={popupRef}
       className="bg-lotion dark:bg-charleston-green-2 fixed z-50 flex flex-col rounded-xl shadow-[0px_9px_46px_8px_#0000001F,0px_24px_38px_3px_#00000024,0px_11px_15px_-7px_#00000033]"
@@ -171,11 +172,7 @@ export default function SourcesPopup({
                                     : doc.date !== option.date,
                                 )
                               : [];
-                          dispatch(
-                            setSelectedDocs(
-                              updatedDocs.length > 0 ? updatedDocs : null,
-                            ),
-                          );
+                          dispatch(setSelectedDocs(updatedDocs));
                           handlePostDocumentSelect(
                             updatedDocs.length > 0 ? updatedDocs : null,
                           );
@@ -218,7 +215,7 @@ export default function SourcesPopup({
             </>
           ) : (
             <div className="dark:text-bright-gray p-4 text-center text-gray-500 dark:text-[14px]">
-              {t('noSourcesAvailable')}
+              {t('conversation.sources.noSourcesAvailable')}
             </div>
           )}
         </div>
@@ -245,4 +242,6 @@ export default function SourcesPopup({
       </div>
     </div>
   );
+
+  return createPortal(popupContent, document.body);
 }

frontend/src/components/TextToSpeechButton.tsx

@@ -1,94 +1,202 @@
-import { useState, useRef } from 'react';
+import { useState, useRef, useEffect } from 'react';
 import Speaker from '../assets/speaker.svg?react';
 import Stopspeech from '../assets/stopspeech.svg?react';
 import LoadingIcon from '../assets/Loading.svg?react'; // Add a loading icon SVG here
+
 const apiHost = import.meta.env.VITE_API_HOST || 'https://docsapi.arc53.com';
 
-export default function SpeakButton({
-  text,
-  colorLight,
-  colorDark,
-}: {
-  text: string;
-  colorLight?: string;
-  colorDark?: string;
-}) {
+let currentlyPlayingAudio: {
+  audio: HTMLAudioElement;
+  stopCallback: () => void;
+} | null = null;
+
+let currentLoadingRequest: {
+  abortController: AbortController;
+  stopLoadingCallback: () => void;
+} | null = null;
+
+// LRU Cache for audio
+const audioCache = new Map<string, string>();
+const MAX_CACHE_SIZE = 10;
+
+function getCachedAudio(text: string): string | undefined {
+  const cached = audioCache.get(text);
+  if (cached) {
+    audioCache.delete(text);
+    audioCache.set(text, cached);
+  }
+  return cached;
+}
+
+function setCachedAudio(text: string, audioBase64: string) {
+  if (audioCache.has(text)) {
+    audioCache.delete(text);
+  }
+  if (audioCache.size >= MAX_CACHE_SIZE) {
+    const firstKey = audioCache.keys().next().value;
+    if (firstKey !== undefined) {
+      audioCache.delete(firstKey);
+    }
+  }
+
+  audioCache.set(text, audioBase64);
+}
+
+export default function SpeakButton({ text }: { text: string }) {
   const [isSpeaking, setIsSpeaking] = useState(false);
   const [isLoading, setIsLoading] = useState(false);
-  const [isSpeakHovered, setIsSpeakHovered] = useState(false);
   const audioRef = useRef<HTMLAudioElement | null>(null);
+  const abortControllerRef = useRef<AbortController | null>(null);
+
+  useEffect(() => {
+    return () => {
+      // Abort any pending fetch request
+      if (abortControllerRef.current) {
+        abortControllerRef.current.abort();
+        abortControllerRef.current = null;
+      }
+
+      // Stop any playing audio
+      if (audioRef.current) {
+        audioRef.current.pause();
+        if (currentlyPlayingAudio?.audio === audioRef.current) {
+          currentlyPlayingAudio = null;
+        }
+        audioRef.current = null;
+      }
+
+      // Clear global loading request if it's this component's
+      if (currentLoadingRequest) {
+        currentLoadingRequest = null;
+      }
+    };
+  }, []);
 
   const handleSpeakClick = async () => {
     if (isSpeaking) {
-      // Stop audio if it's currently playing
       audioRef.current?.pause();
       audioRef.current = null;
+      currentlyPlayingAudio = null;
       setIsSpeaking(false);
       return;
     }
 
+    // Stop any currently playing audio
+    if (currentlyPlayingAudio) {
+      currentlyPlayingAudio.audio.pause();
+      currentlyPlayingAudio.stopCallback();
+      currentlyPlayingAudio = null;
+    }
+
+    // Abort any pending loading request
+    if (currentLoadingRequest) {
+      currentLoadingRequest.abortController.abort();
+      currentLoadingRequest.stopLoadingCallback();
+      currentLoadingRequest = null;
+    }
+
     try {
-      // Set loading state and initiate TTS request
       setIsLoading(true);
+      const cachedAudio = getCachedAudio(text);
+      let audioBase64: string;
 
-      const response = await fetch(apiHost + '/api/tts', {
-        method: 'POST',
-        headers: { 'Content-Type': 'application/json' },
-        body: JSON.stringify({ text }),
-      });
-
-      const data = await response.json();
-
-      if (data.success && data.audio_base64) {
-        // Create and play the audio
-        const audio = new Audio(`data:audio/mp3;base64,${data.audio_base64}`);
-        audioRef.current = audio;
-
-        audio.play().then(() => {
-          setIsSpeaking(true);
-          setIsLoading(false);
-
-          // Reset when audio ends
-          audio.onended = () => {
-            setIsSpeaking(false);
-            audioRef.current = null;
-          };
-        });
-      } else {
-        console.error('Failed to retrieve audio.');
+      if (cachedAudio) {
+        audioBase64 = cachedAudio;
         setIsLoading(false);
+      } else {
+        const abortController = new AbortController();
+        abortControllerRef.current = abortController;
+
+        currentLoadingRequest = {
+          abortController,
+          stopLoadingCallback: () => {
+            setIsLoading(false);
+          },
+        };
+
+        const response = await fetch(apiHost + '/api/tts', {
+          method: 'POST',
+          headers: { 'Content-Type': 'application/json' },
+          body: JSON.stringify({ text }),
+          signal: abortController.signal,
+        });
+
+        const data = await response.json();
+        abortControllerRef.current = null;
+        currentLoadingRequest = null;
+
+        if (data.success && data.audio_base64) {
+          audioBase64 = data.audio_base64;
+          // Store in cache
+          setCachedAudio(text, audioBase64);
+          setIsLoading(false);
+        } else {
+          console.error('Failed to retrieve audio.');
+          setIsLoading(false);
+          return;
+        }
+      }
+
+      const audio = new Audio(`data:audio/mp3;base64,${audioBase64}`);
+      audioRef.current = audio;
+
+      currentlyPlayingAudio = {
+        audio,
+        stopCallback: () => {
+          setIsSpeaking(false);
+          audioRef.current = null;
+        },
+      };
+
+      audio.play().then(() => {
+        setIsSpeaking(true);
+        setIsLoading(false);
+
+        audio.onended = () => {
+          setIsSpeaking(false);
+          audioRef.current = null;
+          if (currentlyPlayingAudio?.audio === audio) {
+            currentlyPlayingAudio = null;
+          }
+        };
+      });
+    } catch (error: any) {
+      abortControllerRef.current = null;
+      currentLoadingRequest = null;
+
+      if (error.name === 'AbortError') {
+        return;
       }
-    } catch (error) {
       console.error('Error fetching audio from TTS endpoint', error);
       setIsLoading(false);
     }
   };
 
   return (
-    <div
-      className={`flex items-center justify-center rounded-full p-2 ${
-        isSpeakHovered
-          ? `dark:bg-purple-taupe bg-[#EEEEEE]`
-          : `bg-[${colorLight ? colorLight : '#FFFFFF'}] dark:bg-[${colorDark ? colorDark : 'transparent'}]`
+    <button
+      type="button"
+      className={`flex cursor-pointer items-center justify-center rounded-full p-2 ${
+        isSpeaking || isLoading
+          ? 'dark:bg-purple-taupe bg-[#EEEEEE]'
+          : 'bg-white-3000 dark:hover:bg-purple-taupe hover:bg-[#EEEEEE] dark:bg-transparent'
       }`}
+      onClick={handleSpeakClick}
+      aria-label={
+        isLoading
+          ? 'Loading audio'
+          : isSpeaking
+            ? 'Stop speaking'
+            : 'Speak text'
+      }
+      disabled={isLoading}
     >
       {isLoading ? (
         <LoadingIcon className="animate-spin" />
       ) : isSpeaking ? (
-        <Stopspeech
-          className="cursor-pointer fill-none"
-          onClick={handleSpeakClick}
-          onMouseEnter={() => setIsSpeakHovered(true)}
-          onMouseLeave={() => setIsSpeakHovered(false)}
-        />
+        <Stopspeech className="fill-none" />
       ) : (
-        <Speaker
-          className="cursor-pointer fill-none"
-          onClick={handleSpeakClick}
-          onMouseEnter={() => setIsSpeakHovered(true)}
-          onMouseLeave={() => setIsSpeakHovered(false)}
-        />
+        <Speaker className="fill-none" />
       )}
-    </div>
+    </button>
   );
 }

frontend/src/components/ToolsPopup.tsx

@@ -1,4 +1,5 @@
 import React, { useEffect, useRef, useState, useLayoutEffect } from 'react';
+import { createPortal } from 'react-dom';
 import { useTranslation } from 'react-i18next';
 import { useSelector } from 'react-redux';
 import { selectToken } from '../preferences/preferenceSlice';
@@ -133,10 +134,10 @@ export default function ToolsPopup({
     tool.displayName.toLowerCase().includes(searchTerm.toLowerCase()),
   );
 
-  return (
+  const popupContent = (
     <div
       ref={popupRef}
-      className="border-light-silver bg-lotion dark:border-dim-gray dark:bg-charleston-green-2 fixed z-9999 rounded-lg border shadow-[0px_9px_46px_8px_#0000001F,0px_24px_38px_3px_#00000024,0px_11px_15px_-7px_#00000033]"
+      className="border-light-silver bg-lotion dark:border-dim-gray dark:bg-charleston-green-2 fixed z-50 rounded-lg border shadow-[0px_9px_46px_8px_#0000001F,0px_24px_38px_3px_#00000024,0px_11px_15px_-7px_#00000033]"
       style={{
         top: popupPosition.showAbove ? popupPosition.top : undefined,
         bottom: popupPosition.showAbove
@@ -242,4 +243,6 @@ export default function ToolsPopup({
       </div>
     </div>
   );
+
+  return createPortal(popupContent, document.body);
 }

frontend/src/components/UploadToast.tsx

@@ -44,7 +44,10 @@ export default function UploadToast() {
   };
 
   return (
-    <div className="fixed right-4 bottom-4 z-50 flex max-w-md flex-col gap-2">
+    <div
+      className="fixed right-4 bottom-4 z-50 flex max-w-md flex-col gap-2"
+      onMouseDown={(e) => e.stopPropagation()}
+    >
       {uploadTasks
         .filter((task) => !task.dismissed)
         .map((task) => {

frontend/src/components/types/index.ts

@@ -23,6 +23,7 @@ export type InputProps = {
     e: React.KeyboardEvent<HTMLTextAreaElement | HTMLInputElement>,
   ) => void;
   leftIcon?: React.ReactNode;
+  edgeRoundness?: string;
 };
 
 export type MermaidRendererProps = {

frontend/src/conversation/Conversation.tsx

@@ -1,20 +1,16 @@
 import { useCallback, useEffect, useRef, useState } from 'react';
-import { useDropzone } from 'react-dropzone';
 import { useTranslation } from 'react-i18next';
 import { useDispatch, useSelector } from 'react-redux';
 
 import SharedAgentCard from '../agents/SharedAgentCard';
-import DragFileUpload from '../assets/DragFileUpload.svg';
 import MessageInput from '../components/MessageInput';
 import { useMediaQuery } from '../hooks';
-import { ActiveState } from '../models/misc';
 import {
   selectConversationId,
   selectSelectedAgent,
   selectToken,
 } from '../preferences/preferenceSlice';
 import { AppDispatch } from '../store';
-import Upload from '../upload/Upload';
 import { handleSendFeedback } from './conversationHandlers';
 import ConversationMessages from './ConversationMessages';
 import { FEEDBACK, Query } from './conversationModels';
@@ -45,53 +41,12 @@ export default function Conversation() {
   const selectedAgent = useSelector(selectSelectedAgent);
   const completedAttachments = useSelector(selectCompletedAttachments);
 
-  const [uploadModalState, setUploadModalState] =
-    useState<ActiveState>('INACTIVE');
-  const [files, setFiles] = useState<File[]>([]);
   const [lastQueryReturnedErr, setLastQueryReturnedErr] =
     useState<boolean>(false);
   const [isShareModalOpen, setShareModalState] = useState<boolean>(false);
-  const [handleDragActive, setHandleDragActive] = useState<boolean>(false);
 
   const fetchStream = useRef<any>(null);
 
-  const onDrop = useCallback((acceptedFiles: File[]) => {
-    setUploadModalState('ACTIVE');
-    setFiles(acceptedFiles);
-    setHandleDragActive(false);
-  }, []);
-
-  const { getRootProps, getInputProps } = useDropzone({
-    onDrop,
-    noClick: true,
-    multiple: true,
-    onDragEnter: () => {
-      setHandleDragActive(true);
-    },
-    onDragLeave: () => {
-      setHandleDragActive(false);
-    },
-    maxSize: 25000000,
-    accept: {
-      'application/pdf': ['.pdf'],
-      'text/plain': ['.txt'],
-      'text/x-rst': ['.rst'],
-      'text/x-markdown': ['.md'],
-      'application/zip': ['.zip'],
-      'application/vnd.openxmlformats-officedocument.wordprocessingml.document':
-        ['.docx'],
-      'application/json': ['.json'],
-      'text/csv': ['.csv'],
-      'text/html': ['.html'],
-      'application/epub+zip': ['.epub'],
-      'application/vnd.openxmlformats-officedocument.spreadsheetml.sheet': [
-        '.xlsx',
-      ],
-      'application/vnd.openxmlformats-officedocument.presentationml.presentation':
-        ['.pptx'],
-    },
-  });
-
   const handleFetchAnswer = useCallback(
     ({ question, index }: { question: string; index?: number }) => {
       fetchStream.current = dispatch(fetchAnswer({ question, indx: index }));
@@ -175,7 +130,7 @@ export default function Conversation() {
           }),
         );
         handleQuestion({
-          question: queries[queries.length - 1].prompt,
+          question: question,
           isRetry: true,
         });
       } else {
@@ -222,14 +177,7 @@ export default function Conversation() {
       />
 
       <div className="bg-opacity-0 z-3 flex h-auto w-full max-w-[1300px] flex-col items-end self-center rounded-2xl py-1 md:w-9/12 lg:w-8/12 xl:w-8/12 2xl:w-6/12">
-        <div
-          {...getRootProps()}
-          className="flex w-full items-center rounded-[40px]"
-        >
-          <label htmlFor="file-upload" className="sr-only">
-            {t('modals.uploadDoc.label')}
-          </label>
-          <input {...getInputProps()} id="file-upload" />
+        <div className="flex w-full items-center rounded-[40px] px-2">
           <MessageInput
             onSubmit={(text) => {
               handleQuestionSubmission(text);
@@ -244,26 +192,6 @@ export default function Conversation() {
           {t('tagline')}
         </p>
       </div>
-      {handleDragActive && (
-        <div className="bg-opacity-50 dark:bg-gray-alpha pointer-events-none fixed top-0 left-0 z-30 flex size-full flex-col items-center justify-center bg-white">
-          <img className="filter dark:invert" src={DragFileUpload} />
-          <span className="text-outer-space dark:text-silver px-2 text-2xl font-bold">
-            {t('modals.uploadDoc.drag.title')}
-          </span>
-          <span className="text-s text-outer-space dark:text-silver w-48 p-2 text-center">
-            {t('modals.uploadDoc.drag.description')}
-          </span>
-        </div>
-      )}
-      {uploadModalState === 'ACTIVE' && (
-        <Upload
-          receivedFile={files}
-          setModalState={setUploadModalState}
-          isOnboarding={false}
-          renderTab={'file'}
-          close={() => setUploadModalState('INACTIVE')}
-        ></Upload>
-      )}
     </div>
   );
 }

frontend/src/conversation/ConversationBubble.module.css

@@ -3,9 +3,9 @@
 }
 
 .list li:not(:first-child) {
-  margin-top: 1em;
+  margin-top: 0.5em;
 }
 
 .list li > .list {
-  margin-top: 1em;
+  margin-top: 0.5em;
 }