fix: add logging for LLM stream calls and remove debug logs in tool execution

src/unpredictable_lord/chat/chat.py

@@ -31,11 +31,13 @@ if USE_MODAL:
     _generate_stream = modal.Function.from_name(APP_NAME, "generate_stream")
 
     def generate_stream(input_tokens):
+        logger.info("Calling Modal LLM generate_stream")
         return _generate_stream.remote_gen(input_tokens)
 else:
     from unpredictable_lord.chat.llm_zerogpu import generate_stream as _generate_stream
 
     def generate_stream(input_tokens):
+        logger.info("Calling ZeroGPU LLM generate_stream")
         return _generate_stream(input_tokens)
 
 
@@ -128,8 +130,6 @@ def _stream_llm_response(messages: list[oh.Message], encoding):
     convo = oh.Conversation.from_messages(messages)
     input_tokens = encoding.render_conversation_for_completion(convo, oh.Role.ASSISTANT)
 
-    logger.debug(f"Input token count: {len(input_tokens)}")
-
     parser = oh.StreamableParser(encoding, role=oh.Role.ASSISTANT)
 
     response_text = ""
@@ -250,29 +250,14 @@ async def chat_with_mcp_tools(
 
             logger.info(f"Found {len(tool_calls)} tool call(s)")
 
-            # Log parsed messages for debugging
-            for i, pm in enumerate(parsed_messages):
-                logger.debug(
-                    f"Parsed message {i}: role={pm.author.role if pm.author else 'None'}, recipient={pm.recipient}, channel={pm.channel}"
-                )
-
             # Add parsed messages to conversation
             messages.extend(parsed_messages)
 
             # Execute tools via MCP
             tool_result_messages = await execute_tool_calls(tool_calls)
 
-            # Log tool result messages for debugging
-            for i, trm in enumerate(tool_result_messages):
-                logger.debug(
-                    f"Tool result message {i}: role={trm.author.role if trm.author else 'None'}, name={trm.author.name if trm.author else 'None'}, channel={trm.channel}"
-                )
-
             messages.extend(tool_result_messages)
 
-            # Log total message count
-            logger.debug(f"Total messages after tool execution: {len(messages)}")
-
         # Ensure final response is yielded (even if empty after tool calls)
         yield partial_history
 

src/unpredictable_lord/chat/chat_tools.py

@@ -138,9 +138,6 @@ async def _execute_tool(tool_name: str, params: dict) -> str:
     mcp_client = MCPClient()
     try:
         result = await mcp_client.call_tool(tool_name, params)
-        logger.debug(
-            f"Raw MCP result for {tool_name}: {result[:200] if result else 'None'}"
-        )
 
         # MCP returns text - try to parse it
         # First, try as JSON
@@ -189,9 +186,6 @@ async def execute_tool_calls(tool_calls: list[oh.Message]) -> list[oh.Message]:
         # Filter parameters to only include those defined in the tool schema
         params = _filter_tool_params(tool_name, raw_params)
 
-        if raw_params != params:
-            logger.debug(f"Filtered params for {tool_name}: {raw_params} -> {params}")
-
         logger.info(f"Executing MCP tool: {tool_name} with params: {params}")
 
         # Execute tool via MCP (returns JSON string)