fix: handle invalid_tool_calls args normalization for DeepSeek

Extended ToolCallArgsParsingWrapper to handle both tool_calls and
invalid_tool_calls args formatting inconsistencies from DeepSeek:

- tool_calls.args: string -> dict (for successful calls)
- invalid_tool_calls.args: dict -> string (for failed calls)

The wrapper now normalizes both types before AIMessage construction,
preventing Pydantic validation errors in both success and error cases.

Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

CHANGELOG.md

@@ -10,7 +10,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ## [0.4.1] - 2025-11-05
 
 ### Fixed
-- Fixed Pydantic validation error for tool_calls when using DeepSeek and other AI providers that return `args` as JSON strings instead of dictionaries. Added `ToolCallArgsParsingWrapper` that monkey-patches ChatOpenAI's `_create_chat_result` method to parse string arguments before AIMessage construction.
+- Fixed Pydantic validation errors for both `tool_calls` and `invalid_tool_calls` when using DeepSeek and other AI providers:
+  - `tool_calls.args`: Converts JSON strings to dictionaries (for successful tool calls)
+  - `invalid_tool_calls.args`: Converts dictionaries to JSON strings (for failed tool calls)
+- Added `ToolCallArgsParsingWrapper` that monkey-patches ChatOpenAI's `_create_chat_result` method to normalize arguments before AIMessage construction.
 
 ## [0.4.0] - 2025-11-05
 

agent/chat_model_wrapper.py

@@ -50,10 +50,14 @@ class ToolCallArgsParsingWrapper:
 
             if 'choices' in response_dict:
                 for choice in response_dict['choices']:
-                    if 'message' in choice and 'tool_calls' in choice['message']:
-                        tool_calls = choice['message']['tool_calls']
-                        if tool_calls:
-                            for tool_call in tool_calls:
+                    if 'message' not in choice:
+                        continue
+
+                    message = choice['message']
+
+                    # Fix regular tool_calls: string args -> dict
+                    if 'tool_calls' in message and message['tool_calls']:
+                        for tool_call in message['tool_calls']:
                             if 'function' in tool_call and 'arguments' in tool_call['function']:
                                 args = tool_call['function']['arguments']
                                 # Parse string arguments to dict
@@ -64,6 +68,19 @@ class ToolCallArgsParsingWrapper:
                                         # Keep as string if parsing fails
                                         pass
 
+                    # Fix invalid_tool_calls: dict args -> string
+                    if 'invalid_tool_calls' in message and message['invalid_tool_calls']:
+                        for invalid_call in message['invalid_tool_calls']:
+                            if 'args' in invalid_call:
+                                args = invalid_call['args']
+                                # Convert dict arguments to JSON string
+                                if isinstance(args, dict):
+                                    try:
+                                        invalid_call['args'] = json.dumps(args)
+                                    except (TypeError, ValueError):
+                                        # Keep as-is if serialization fails
+                                        pass
+
             # Call original method with fixed response
             return original_create_chat_result(response_dict, generation_info)