docs/DEBUG_TO_PRODUCTION.md

Debug to Production Guide

This document lists all debugging features added and how to remove them for production deployment.

Debug Features Added

1. Vectorization Query Logging (embeddings.py)

Location: backend/embeddings.py line 52 Debug Code:

logger.info(f"🔍 VECTORIZING QUERY STRING: '{query}'")

Purpose: Shows the exact string being vectorized for debugging vector search issues. To Remove: Delete line 52

2. Vector Search Results Display (chatbot_backend.py)

Location: Multiple locations in backend/chatbot_backend.py

a. Debug Method _format_vector_search_display (lines 80-117)

Purpose: Formats vector search results for display similar to initial RAG query To Remove: Delete the entire method

b. Debug Method _execute_tool_calls_with_display (lines 419-471)

Purpose: Executes tool calls and yields display content for vector search To Remove: Delete the entire method

c. Tool Execution Logic (lines 321-335)

Debug Code:

tool_results = []
for result in self._execute_tool_calls_with_display(tool_calls):
    if result.get("display"):
        # Yield display content
        yield {
            "type": "content",
            "content": result["display"],
            "done": False
        }
    else:
        # Collect tool result
        tool_results.append(result)

To Remove: Replace entire block with:

tool_results = self._execute_tool_calls(tool_calls)

d. Enhanced Status Message for vector_search (lines 264-267)

Debug Code:

if tool_name == "vector_search":
    versions = tool_args.get('versions', [])
    query = tool_args.get('query', '')
    status_msg = f"\n🔍 Vector searching {', '.join(versions)} for: '{query}'...\n"

Purpose: Shows more detailed status including the query string To Remove: Can keep this as it's useful UX, or simplify to:

if tool_name == "vector_search":
    status_msg = f"\n🔍 Using vector search tool...\n"

Production Deployment Steps

1. Remove debug logging in embeddings.py:

   • Delete line 52 that logs the vectorization query string

2. Clean up chatbot_backend.py:

   • Delete _format_vector_search_display method (lines 80-117)
   • Delete _execute_tool_calls_with_display method (lines 419-471)
   • Replace the debug tool execution block (lines 321-335) with:

     tool_results = self._execute_tool_calls(tool_calls)
     

3. Optional - Simplify status messages:

   • Optionally simplify the vector_search status message

4. Remove test file:

   • Delete test_debug_output.py

Summary

The debug features add visibility into:

1. The exact query string being vectorized (useful for debugging search quality)
2. The retrieved chunks from vector_search tool (similar to initial RAG display)

These features are helpful during development but add extra output that may not be desired in production. The core functionality remains unchanged - only the display/logging is affected.