release: v0.4.2 - fix critical negative cash position bug

Remove debug logging and update CHANGELOG for v0.4.2 release.

Fixed critical bug where trades calculated from initial $10,000 capital
instead of accumulating, allowing over-spending and negative cash balances.

Key changes:
- Extract position dict from CallToolResult.structuredContent
- Enable MCP service logging for better debugging
- Update tests to match production MCP behavior

All tests passing. Ready for production release.

CHANGELOG.md

@@ -7,6 +7,21 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.4.2] - 2025-11-07
+
+### Fixed
+- **Critical:** Fixed negative cash position bug where trades calculated from initial capital instead of accumulating
+  - Root cause: MCP tools return `CallToolResult` objects with position data in `structuredContent` field, but `ContextInjector` was checking `isinstance(result, dict)` which always failed
+  - Impact: Each trade checked cash against initial $10,000 instead of cumulative position, allowing over-spending and resulting in negative cash balances (e.g., -$8,768.68 after 11 trades totaling $18,768.68)
+  - Solution: Updated `ContextInjector` to extract position dict from `CallToolResult.structuredContent` before validation
+  - Fix ensures proper intra-day position tracking with cumulative cash checks preventing over-trading
+  - Updated unit tests to mock `CallToolResult` objects matching production MCP behavior
+  - Locations: `agent/context_injector.py:95-109`, `tests/unit/test_context_injector.py:26-53`
+- Enabled MCP service logging by redirecting stdout/stderr from `/dev/null` to main process for better debugging
+  - Previously, all MCP tool debug output was silently discarded
+  - Now visible in docker logs for diagnosing parameter injection and trade execution issues
+  - Location: `agent_tools/start_mcp_services.py:81-88`
+
 ### Fixed
 - **Critical:** Fixed stale jobs blocking new jobs after Docker container restart
   - Root cause: Jobs with status 'pending', 'downloading_data', or 'running' remained in database after container shutdown, preventing new job creation

agent/context_injector.py

@@ -88,28 +88,17 @@ class ContextInjector:
 
         # Update position state after successful trade
         if request.name in ["buy", "sell"]:
-            # Debug: Log result type and structure
-            print(f"[DEBUG ContextInjector] Trade result type: {type(result)}")
-            print(f"[DEBUG ContextInjector] Trade result: {result}")
-
             # Extract position dict from MCP result
             # MCP tools return CallToolResult objects with structuredContent field
             position_dict = None
             if hasattr(result, 'structuredContent') and result.structuredContent:
                 position_dict = result.structuredContent
-                print(f"[DEBUG ContextInjector] Extracted from structuredContent: {position_dict}")
             elif isinstance(result, dict):
                 position_dict = result
-                print(f"[DEBUG ContextInjector] Using result as dict: {position_dict}")
 
             # Check if position dict is valid (not an error) and update state
             if position_dict and "error" not in position_dict and "CASH" in position_dict:
                 # Update our tracked position with the new state
                 self._current_position = position_dict.copy()
-                print(f"[DEBUG ContextInjector] Updated _current_position: {self._current_position}")
-            else:
-                print(f"[DEBUG ContextInjector] Did NOT update _current_position - check failed")
-                print(f"[DEBUG ContextInjector] position_dict: {position_dict}")
-                print(f"[DEBUG ContextInjector] _current_position remains: {self._current_position}")
 
         return result

agent_tools/start_mcp_services.py

@@ -78,10 +78,11 @@ class MCPServiceManager:
             env['PYTHONPATH'] = str(Path.cwd())
 
             # Start service process (output goes to Docker logs)
+            # Enable stdout/stderr for debugging (previously sent to DEVNULL)
             process = subprocess.Popen(
                 [sys.executable, str(script_path)],
-                stdout=subprocess.DEVNULL,
-                stderr=subprocess.DEVNULL,
+                stdout=sys.stdout,  # Redirect to main process stdout
+                stderr=sys.stderr,  # Redirect to main process stderr
                 cwd=Path.cwd(),  # Use current working directory (/app)
                 env=env  # Pass environment with PYTHONPATH
             )

agent_tools/tool_trade.py

@@ -132,14 +132,11 @@ def _buy_impl(symbol: str, amount: int, signature: str = None, today_date: str =
         # Step 1: Get current position
         # Use injected position if available (for intra-day tracking),
         # otherwise query database for starting position
-        print(f"[DEBUG buy] _current_position received: {_current_position}")
         if _current_position is not None:
             current_position = _current_position
             next_action_id = 0  # Not used in new schema
-            print(f"[DEBUG buy] Using injected position: {current_position}")
         else:
             current_position, next_action_id = get_current_position_from_db(job_id, signature, today_date)
-            print(f"[DEBUG buy] Queried position from DB: {current_position}")
 
         # Step 2: Get stock price
         try:
@@ -192,8 +189,6 @@ def _buy_impl(symbol: str, amount: int, signature: str = None, today_date: str =
 
         conn.commit()
         print(f"[buy] {signature} bought {amount} shares of {symbol} at ${this_symbol_price}")
-        print(f"[DEBUG buy] Returning new_position: {new_position}")
-        print(f"[DEBUG buy] new_position keys: {list(new_position.keys())}")
         return new_position
 
     except Exception as e: