fix: save convo messages, docsgpt provider format

application/api/answer/routes.py

@@ -141,17 +141,17 @@ def save_conversation(conversation_id, question, response, source_log_docs, llm)
                 "role": "assistant",
                 "content": "Summarise following conversation in no more than 3 "
                 "words, respond ONLY with the summary, use the same "
-                "language as the system \n\nUser: "
-                + question
-                + "\n\n"
-                + "AI: "
-                + response,
+                "language as the system",
             },
             {
                 "role": "user",
                 "content": "Summarise following conversation in no more than 3 words, "
                 "respond ONLY with the summary, use the same language as the "
-                "system",
+                "system \n\nUser: "
+                + question
+                + "\n\n"
+                + "AI: "
+                + response,
             },
         ]
 

application/llm/docsgpt_provider.py

@@ -9,35 +9,25 @@ class DocsGPTAPILLM(BaseLLM):
         super().__init__(*args, **kwargs)
         self.api_key = api_key
         self.user_api_key = user_api_key
-        self.endpoint = "https://llm.docsgpt.co.uk"
+        self.endpoint = "https://llm.arc53.com"
 
     def _raw_gen(self, baseself, model, messages, stream=False, *args, **kwargs):
-        context = messages[0]["content"]
-        user_question = messages[-1]["content"]
-        prompt = f"### Instruction \n {user_question} \n ### Context \n {context} \n ### Answer \n"
-
         response = requests.post(
-            f"{self.endpoint}/answer", json={"prompt": prompt, "max_new_tokens": 30}
+            f"{self.endpoint}/answer", json={"messages": messages, "max_new_tokens": 30}
         )
         response_clean = response.json()["a"].replace("###", "")
 
         return response_clean
 
     def _raw_gen_stream(self, baseself, model, messages, stream=True, *args, **kwargs):
-        context = messages[0]["content"]
-        user_question = messages[-1]["content"]
-        prompt = f"### Instruction \n {user_question} \n ### Context \n {context} \n ### Answer \n"
-
-        # send prompt to endpoint /stream
         response = requests.post(
             f"{self.endpoint}/stream",
-            json={"prompt": prompt, "max_new_tokens": 256},
+            json={"messages": messages, "max_new_tokens": 256},
             stream=True,
         )
 
         for line in response.iter_lines():
             if line:
-                # data = json.loads(line)
                 data_str = line.decode("utf-8")
                 if data_str.startswith("data: "):
                     data = json.loads(data_str[6:])