add tool call from fe

main.py

@@ -363,6 +363,10 @@ class FollowupQueryModel(BaseModel):
     )
     conversation_id: str = Field(default_factory=lambda: str(uuid4()), description="Unique identifier for the conversation")
     user_id: str = Field(..., description="Unique identifier for the user")
+    tool_call: Literal["web", "news", "auto"] = Field(
+        default="auto",
+        description="Type of tool to call (web, news, auto)"
+    )
 
     class Config:
         schema_extra = {
@@ -370,7 +374,8 @@ class FollowupQueryModel(BaseModel):
                 "query": "How can I improve my productivity?",
                 "model_id": "openai/gpt-4o-mini",
                 "conversation_id": "123e4567-e89b-12d3-a456-426614174000",
-                "user_id": "user123"
+                "user_id": "user123",
+                "tool_call": "auto"
             }
         }
 
@@ -519,33 +524,45 @@ def followup_agent(query: FollowupQueryModel, background_tasks: BackgroundTasks,
     
     def process_response():
         full_response = ""
-        for content in chat_with_llama_stream(limited_conversation, model=query.model_id):
-            yield content
-            full_response += content
-        
-        logger.info(f"LLM RAW response for query: {query.query}: {full_response}")
-        response_content, interact, tools = parse_followup_and_tools(full_response)
-        
-        result = {
-            "clarification": interact,
-            "tools": tools
-        }
-        
-        yield "<json>"+ json.dumps(result)+"</json>"
         
+        # Check if tool_call is specified and call the tool directly
+        if query.tool_call in ["web", "news"]:
+            search_query = query.query
+            search_response = search_assistant_api(search_query, query.tool_call, model=query.model_id)
+            
+            yield "<report>"
+            for content in search_response():
+                yield content
+                full_response += content
+            yield "</report>"
+        else:
+            for content in chat_with_llama_stream(limited_conversation, model=query.model_id):
+                yield content
+                full_response += content
         
-        # Process tool if present
-        if tools and len(tools) > 0:
-            tool = tools[0]  # Assume only one tool is present
-            if tool["name"] in ["news", "web"]:
-                search_query = tool["input"]
-                search_response = search_assistant_api(search_query, tool["name"], model=query.model_id)
-                
-                yield "<report>"
-                for content in search_response():
-                    yield content
-                    full_response += content
-                yield "</report>"
+            logger.info(f"LLM RAW response for query: {query.query}: {full_response}")
+            response_content, interact, tools = parse_followup_and_tools(full_response)
+            
+            result = {
+                "clarification": interact,
+                "tools": tools
+            }
+            
+            yield "<json>"+ json.dumps(result)+"</json>"
+            
+            
+            # Process tool if present
+            if tools and len(tools) > 0:
+                tool = tools[0]  # Assume only one tool is present
+                if tool["name"] in ["news", "web"]:
+                    search_query = tool["input"]
+                    search_response = search_assistant_api(search_query, tool["name"], model=query.model_id)
+                    
+                    yield "<report>"
+                    for content in search_response():
+                        yield content
+                        full_response += content
+                    yield "</report>"
         
         # Add the assistant's response to the conversation history
         conversations[query.conversation_id].append({"role": "assistant", "content": full_response})