xtekky/gpt4free · error · ResponseError

Invalid response: {result}

Error message

Invalid response: {result}

What it means

Raised as ResponseError after POSTing to {base_url}/{model}/predictions when the returned JSON has no 'id' key. Replicate normally returns a prediction object with an id used to open the streaming URL; anything else means the creation call did not produce a usable prediction.

Source

Thrown at g4f/Provider/needs_auth/Replicate.py:71

                    "prompt": format_prompt(messages),
                    **filter_none(
                        system_prompt=system_prompt,
                        max_new_tokens=max_tokens,
                        temperature=temperature,
                        top_p=top_p,
                        top_k=top_k,
                        stop_sequences=",".join(stop) if stop else None,
                    ),
                    **extra_body,
                },
            }
            url = f"{base_url.rstrip('/')}/{model}/predictions"
            async with session.post(url, json=data) as response:
                message = "Model not found" if response.status == 404 else None
                await raise_for_status(response, message)
                result = await response.json()
                if "id" not in result:
                    raise ResponseError(f"Invalid response: {result}")
                async with session.get(
                    result["urls"]["stream"], headers={"Accept": "text/event-stream"}
                ) as response:
                    await raise_for_status(response)
                    event = None
                    async for line in response.iter_lines():
                        if line.startswith(b"event: "):
                            event = line[7:]
                            if event == b"done":
                                break
                        elif event == b"output":
                            if line.startswith(b"data: "):
                                new_text = line[6:].decode()
                                if new_text:
                                    yield new_text
                                else:
                                    yield "\n"

View on GitHub (pinned to 973504e177)

Solutions

1. Inspect the raised message — it contains the full JSON body returned by Replicate
2. Verify the api_key is a valid Replicate token (starts with r8_)
3. Verify the model slug exists and is hosted on api.replicate.com when using a key

Example fix

# before
response = client.chat.completions.create(model='owner/model', provider=g4f.Provider.Replicate, api_key=key)

# after
try:
    response = client.chat.completions.create(model='owner/model', provider=g4f.Provider.Replicate, api_key=key)
except g4f.errors.ResponseError as e:
    logging.error('Replicate prediction creation failed: %s', e)
    raise

from g4f.errors import ResponseError
try:
    result = ...create(provider=g4f.Provider.Replicate)
except ResponseError as e:
    logging.error('Replicate body: %s', e)
    if 'auth' in str(e).lower():
        rotate_replicate_token()
    else:
        raise

When it happens

Trigger: The prediction-creation endpoint returns an error object (missing/invalid token with 200-style handling, model not deployable, malformed input) so the response JSON lacks 'id'.

Common situations: Replicate API token invalid so the POST returns an auth error document, model slug typo handled as JSON error, upstream schema change.

Related errors

• result
• api_key is missing
• Failed to upload file: {response.status} {error_text}
• Cohere API Error: {data['error']}
• Failed to upload file: {result}

AI-assisted analysis of xtekky/gpt4free@973504e177 (2026-08-14). Data as JSON: /api/errors/372e8dfd2ade07db. Report an issue: GitHub.