microsoft/autogen · error · ValueError

Run failed: {run.last_error}

Error message

Run failed: {run.last_error}

What it means

While polling a run on the Assistants API thread, OpenAIAssistantAgent checks run.status each iteration; a terminal status of 'failed' (server-side failure: expired run, model error, rate/quota issues, bad tool definitions) triggers this ValueError carrying run.last_error from OpenAI. The agent surfaces the API's own error detail rather than retrying.

Source

Thrown at python/packages/autogen-ext/src/autogen_ext/agents/openai/_openai_assistant_agent.py:438

                    thread_id=self._thread_id,
                    assistant_id=self._get_assistant_id,
                )
            )
        )

        # Wait for run completion by polling
        while True:
            run = await cancellation_token.link_future(
                asyncio.ensure_future(
                    self._client.beta.threads.runs.retrieve(  # type: ignore[reportDeprecated]
                        thread_id=self._thread_id,
                        run_id=run.id,
                    )
                )
            )

            if run.status == "failed":
                raise ValueError(f"Run failed: {run.last_error}")

            # If the run requires action (function calls), execute tools and continue
            if run.status == "requires_action" and run.required_action is not None:
                tool_calls: List[FunctionCall] = []
                for required_tool_call in run.required_action.submit_tool_outputs.tool_calls:
                    if required_tool_call.type == "function":
                        tool_calls.append(
                            FunctionCall(
                                id=required_tool_call.id,
                                name=required_tool_call.function.name,
                                arguments=required_tool_call.function.arguments,
                            )
                        )

                # Add tool call message to inner messages
                tool_call_msg = ToolCallRequestEvent(source=self.name, content=tool_calls)
                inner_messages.append(tool_call_msg)
                event_logger.debug(tool_call_msg)

View on GitHub (pinned to 027ecf0a37)

Solutions

1. Read run.last_error in the message — it contains OpenAI's error code/message (e.g. rate_limit_exceeded, invalid_model) and points at the true cause.
2. For quota/rate errors: wait and retry the on_messages call (the thread is preserved; new input starts a new run).
3. For invalid_model: update the assistant to a current model via client.beta.assistants.update.
4. For expired runs: shorten tool execution time or raise the run's expiration when creating it.

Example fix

# before
resp = await agent.on_messages([TextMessage(source="user", content="hi")], ct)

# after
try:
    resp = await agent.on_messages([TextMessage(source="user", content="hi")], ct)
except ValueError as e:
    if "Run failed" in str(e):
        print("OpenAI run error:", e)  # inspect run.last_error detail
        await asyncio.sleep(30)
        resp = await agent.on_messages([TextMessage(source="user", content="hi")], ct)

try:
    resp = await agent.on_messages(msgs, ct)
except ValueError as e:
    if "Run failed" not in str(e):
        raise
    detail = str(e)
    if "rate" in detail.lower() or "quota" in detail.lower():
        await asyncio.sleep(30)
        resp = await agent.on_messages(msgs, ct)
    else:
        raise

When it happens

Trigger: Any assistant run whose status becomes 'failed': invalid/inaccessible model, insufficient quota, run timeout/expiry (runs have a server-side time limit), malformed tool definitions, or content-policy failures.

Common situations: Using a deprecated model name on the assistant; hitting OpenAI rate limits or zero billing quota; long tool executions causing run expiry; stale vector stores or file references used by file_search.

Related errors

• Incorrect client passed to OpenAIAssistantAgent. Please use
• Unsupported tool type: {type(tool)}
• Assistant not initialized
• No tools are available.
• The tool '{tool_call.name}' is not available.

AI-assisted analysis of microsoft/autogen@027ecf0a37 (2026-08-15). Data as JSON: /api/errors/9976c8c6ec12c66a. Report an issue: GitHub.