Skip to content

Personal AI, On Personal Devices

Python

revert: remove _fix_tool_call_arguments from OpenAI-compatible engines

Jon Saad-Falcond97e1dd266ac
diff --git a/src/openjarvis/engine/_openai_compat.py b/src/openjarvis/engine/_openai_compat.py
index ba7d3870..b859bfd9 100644
--- a/src/openjarvis/engine/_openai_compat.py
+++ b/src/openjarvis/engine/_openai_compat.py
@@ -32,26 +32,6 @@ class _OpenAICompatibleEngine(InferenceEngine):
# -- InferenceEngine interface ------------------------------------------
- @staticmethod
- def _fix_tool_call_arguments(msg_dicts: list) -> list:
- """Ensure tool_call arguments are dicts, not JSON strings.
-
- OpenAI-compatible servers (vLLM, SGLang, llama.cpp, etc.) expect
- tool_call arguments as JSON objects. ``messages_to_dicts`` may
- serialize them as strings, which causes 400 errors on multi-turn
- tool-calling conversations.
- """
- for md in msg_dicts:
- for tc in md.get("tool_calls", []):
- fn = tc.get("function", {})
- args = fn.get("arguments")
- if isinstance(args, str):
- try:
- fn["arguments"] = json.loads(args)
- except (json.JSONDecodeError, TypeError):
- pass
- return msg_dicts
-
def generate(
self,
messages: Sequence[Message],
@@ -61,10 +41,9 @@ class _OpenAICompatibleEngine(InferenceEngine):
max_tokens: int = 1024,
**kwargs: Any,
) -> Dict[str, Any]:
- msg_dicts = self._fix_tool_call_arguments(messages_to_dicts(messages))
payload: Dict[str, Any] = {
"model": model,
- "messages": msg_dicts,
+ "messages": messages_to_dicts(messages),
"temperature": temperature,
"max_tokens": max_tokens,
"stream": False,
@@ -125,10 +104,9 @@ class _OpenAICompatibleEngine(InferenceEngine):
max_tokens: int = 1024,
**kwargs: Any,
) -> AsyncIterator[str]:
- msg_dicts = self._fix_tool_call_arguments(messages_to_dicts(messages))
payload: Dict[str, Any] = {
"model": model,
- "messages": msg_dicts,
+ "messages": messages_to_dicts(messages),
"temperature": temperature,
"max_tokens": max_tokens,
"stream": True,