feat: add ReadWebpageTool for fetching and converting webpage content to Markdown

This commit is contained in:
Alex
2025-05-15 12:56:06 +01:00
parent d9fc623dcb
commit 45aace8966
4 changed files with 96 additions and 5 deletions

View File

@@ -0,0 +1,83 @@
import requests
from markdownify import markdownify
from application.agents.tools.base import Tool
from urllib.parse import urlparse
class ReadWebpageTool(Tool):
"""
Read Webpage (browser)
A tool to fetch the HTML content of a URL and convert it to Markdown.
"""
def __init__(self, config=None):
"""
Initializes the tool.
:param config: Optional configuration dictionary. Not used by this tool.
"""
self.config = config
def execute_action(self, action_name: str, **kwargs) -> str:
"""
Executes the specified action. For this tool, the only action is 'read_webpage'.
:param action_name: The name of the action to execute. Should be 'read_webpage'.
:param kwargs: Keyword arguments, must include 'url'.
:return: The Markdown content of the webpage or an error message.
"""
if action_name != "read_webpage":
return f"Error: Unknown action '{action_name}'. This tool only supports 'read_webpage'."
url = kwargs.get("url")
if not url:
return "Error: URL parameter is missing."
# Ensure the URL has a scheme (if not, default to http)
parsed_url = urlparse(url)
if not parsed_url.scheme:
url = "http://" + url
try:
response = requests.get(url, timeout=10, headers={'User-Agent': 'DocsGPT-Agent/1.0'})
response.raise_for_status() # Raise an exception for HTTP errors (4xx or 5xx)
html_content = response.text
#soup = BeautifulSoup(html_content, 'html.parser')
markdown_content = markdownify(html_content, heading_style="ATX", newline_style="BACKSLASH")
return markdown_content
except requests.exceptions.RequestException as e:
return f"Error fetching URL {url}: {e}"
except Exception as e:
return f"Error processing URL {url}: {e}"
def get_actions_metadata(self):
"""
Returns metadata for the actions supported by this tool.
"""
return [
{
"name": "read_webpage",
"description": "Fetches the HTML content of a given URL and returns it as clean Markdown text. Input must be a valid URL.",
"parameters": {
"type": "object",
"properties": {
"url": {
"type": "string",
"description": "The fully qualified URL of the webpage to read (e.g., 'https://www.example.com').",
}
},
"required": ["url"],
"additionalProperties": False,
},
}
]
def get_config_requirements(self):
"""
Returns a dictionary describing the configuration requirements for the tool.
This tool does not require any specific configuration.
"""
return {}

View File

@@ -472,6 +472,7 @@ class Stream(Resource):
decoded_token = request.decoded_token
else:
decoded_token = {"sub": data_key.get("user")}
is_shared_usage = False
elif "active_docs" in data:
source = {"active_docs": data["active_docs"]}
@@ -524,7 +525,8 @@ class Stream(Resource):
user_api_key=user_api_key,
decoded_token=decoded_token,
)
is_shared_usage_val = data.get("is_shared_usage", False)
is_shared_token_val = data.get("shared_token", None)
return Response(
complete_stream(
question=question,
@@ -537,8 +539,8 @@ class Stream(Resource):
index=index,
should_save_conversation=save_conv,
agent_id=agent_id,
is_shared_usage=is_shared_usage,
shared_token=shared_token,
is_shared_usage=is_shared_usage_val,
shared_token=is_shared_token_val,
),
mimetype="text/event-stream",
)

View File

@@ -1,9 +1,14 @@
You are a helpful AI assistant, DocsGPT, specializing in document assistance, designed to offer detailed and informative responses.
You are a helpful AI assistant, DocsGPT. You are proactive and helpful. Try to use tools, if they are available to you,
be proactive and fill in missing information.
Users can Upload documents for your context as attachments or sources via UI using the Conversation input box.
If appropriate, your answers can include code examples, formatted as follows:
```(language)
(code)
```
Users are also able to see charts and diagrams if you use them with valid mermaid syntax in your responses.
Try to respond with mermaid charts if visualization helps with users queries.
You effectively utilize chat history, ensuring relevant and tailored responses.
If a question doesn't align with your context, you provide friendly and helpful replies.
Try to use additional provided context if its available, otherwise use your knowledge and tool capabilities.
----------------
Possible additional context from uploaded sources:
{summaries}

View File

@@ -0,0 +1 @@
<svg xmlns="http://www.w3.org/2000/svg" height="24px" viewBox="0 -960 960 960" width="24px" fill="#e3e3e3"><path d="M480-80q-82 0-155-31.5t-127.5-86Q143-252 111.5-325T80-480q0-83 31.5-155.5t86-127Q252-817 325-848.5T480-880q83 0 155.5 31.5t127 86q54.5 54.5 86 127T880-480q0 82-31.5 155t-86 127.5q-54.5 54.5-127 86T480-80Zm0-82q26-36 45-75t31-83H404q12 44 31 83t45 75Zm-104-16q-18-33-31.5-68.5T322-320H204q29 50 72.5 87t99.5 55Zm208 0q56-18 99.5-55t72.5-87H638q-9 38-22.5 73.5T584-178ZM170-400h136q-3-20-4.5-39.5T300-480q0-21 1.5-40.5T306-560H170q-5 20-7.5 39.5T160-480q0 21 2.5 40.5T170-400Zm216 0h188q3-20 4.5-39.5T580-480q0-21-1.5-40.5T574-560H386q-3 20-4.5 39.5T380-480q0 21 1.5 40.5T386-400Zm268 0h136q5-20 7.5-39.5T800-480q0-21-2.5-40.5T790-560H654q3 20 4.5 39.5T660-480q0 21-1.5 40.5T654-400Zm-16-240h118q-29-50-72.5-87T584-782q18 33 31.5 68.5T638-640Zm-234 0h152q-12-44-31-83t-45-75q-26 36-45 75t-31 83Zm-200 0h118q9-38 22.5-73.5T376-782q-56 18-99.5 55T204-640Z"/></svg>

After

Width:  |  Height:  |  Size: 976 B