huggingface/transformers · error · HTTPException

Missing `model` field in the request body.

Error message

Missing `model` field in the request body.

What it means

Raised as an HTTP 422 by the /load_model endpoint of the transformers CLI serving server. The endpoint expects a JSON body such as {"model": "gpt2"} and refuses to proceed when the 'model' key is absent or null. It exists so clients fail fast with a clear message instead of triggering a KeyError deep inside model loading.

Source

Thrown at src/transformers/cli/serving/server.py:125

    @app.post("/v1/completions")
    async def completions(request: Request, body: dict):
        return await completion_handler.handle_request(body, request.state.request_id)

    @app.post("/v1/responses")
    async def responses(request: Request, body: dict):
        return await response_handler.handle_request(body, request.state.request_id)

    @app.post("/v1/audio/transcriptions")
    async def audio_transcriptions(request: Request):
        return await transcription_handler.handle_request(request)

    @app.post("/load_model")
    async def load_model(body: dict):
        from fastapi import HTTPException

        model = body.get("model")
        if model is None:
            raise HTTPException(status_code=422, detail="Missing `model` field in the request body.")
        model_id_and_revision = model_manager.process_model_name(model)
        return StreamingResponse(
            model_manager.load_model_streaming(model_id_and_revision), media_type="text/event-stream"
        )

    @app.post("/reset")
    def reset():
        model_manager.shutdown()
        return JSONResponse({"status": "ok"})

    @app.get("/v1/models")
    @app.options("/v1/models")
    def list_models():
        return JSONResponse({"object": "list", "data": model_manager.get_gen_models()})

    @app.get("/health")
    def health():
        if not generation_state.is_cb_alive():

View on GitHub (pinned to a597f97485)

Solutions

1. Send a JSON body containing the model field: curl -X POST http://host:port/load_model -H 'Content-Type: application/json' -d '{"model": "openai-community/gpt2"}'
2. Check that the variable holding the model id is set before building the request body
3. Verify you are hitting /load_model with POST and a JSON Content-Type, not a form or query string

Example fix

// before
curl -X POST http://localhost:8000/load_model -d '{}'
// after
curl -X POST http://localhost:8000/load_model -H 'Content-Type: application/json' -d '{"model": "openai-community/gpt2"}'

body = {"model": model_id}
assert isinstance(body.get("model"), str) and body["model"], "POST /load_model requires a non-empty 'model' string"

resp = requests.post(f'{base}/load_model', json=body)
if resp.status_code == 422 and 'Missing `model` field' in resp.text:
    raise ValueError('request body must include the model id') from None

When it happens

Trigger: POST /load_model with an empty body, a JSON body lacking the "model" key, or "model": null. Any HTTP client (curl, requests, fetch) that forgets to serialize the model id into the payload.

Common situations: Scripts that build the request body dynamically and skip the model field when a variable is unset; clients assuming the server has a default model; typos like "model_name" or "modelId" instead of "model".

Related errors

• 'input' must be a string or list
• Unsupported input item type: {item_type!r}
• Unexpected fields in the request: {unexpected}
• Expected file upload, got string
• Expected model name as string

AI-assisted analysis of huggingface/transformers@a597f97485 (2026-08-14). Data as JSON: /api/errors/cc343ed7f930c9d0. Report an issue: GitHub.