abi/screenshot-to-code · error · ValueError

Unexpected response from {context}: {result}

Error message

Unexpected response from {context}: {result}

What it means

Raised by _extract_output_url when the prediction's `output` field doesn't contain anything URL-shaped. The function accepts: a plain string, a dict with a non-empty `url` key, or a non-empty list whose first element is a string or a url-bearing mapping. Anything else — None, an empty list, a list of non-strings, a dict without url — raises this ValueError with the offending payload embedded.

Source

Thrown at backend/image_generation/replicate.py:122

def _extract_output_url(result: Any, context: str) -> str:
    if isinstance(result, str):
        return result

    if isinstance(result, dict):
        url = cast(Any, result.get("url"))
        if isinstance(url, str) and url:
            return url

    if isinstance(result, list) and len(result) > 0:
        first = cast(Any, result[0])
        if isinstance(first, str) and first:
            return first
        if isinstance(first, Mapping):
            url = cast(Any, first.get("url"))
            if isinstance(url, str) and url:
                return url

    raise ValueError(f"Unexpected response from {context}: {result}")


async def call_replicate_model(
    model_path: str, input: dict[str, Any], api_token: str
) -> Any:
    return await _run_prediction(
        f"{REPLICATE_API_BASE_URL}/models/{model_path}/predictions",
        {"input": input},
        api_token,
    )


async def call_replicate_version(
    version: str, input: dict[str, Any], api_token: str
) -> Any:
    return await _run_prediction(
        f"{REPLICATE_API_BASE_URL}/predictions",
        {"version": version, "input": input},

View on GitHub (pinned to d026163f58)

Solutions

1. Look at the embedded `result` in the message — it shows exactly what the model returned.
2. Check the model's docs on Replicate for its output schema and adapt _extract_output_url if it uses a different key.
3. If output is empty/None, treat it as a model-side failure and retry or adjust input.
4. Test the model once via the Replicate playground to see its real output shape.

Example fix

# before — model returns {"image": "https://..."}
# _extract_output_url raises because there is no "url" key

# after — extend the dict branch
url = cast(Any, result.get("url") or result.get("image"))

from typing import Any, Mapping

def output_has_extractable_url(output: Any) -> bool:
    if isinstance(output, str) and output:
        return True
    if isinstance(output, Mapping):
        return isinstance(output.get("url"), str) and bool(output["url"])
    if isinstance(output, list) and output:
        first = output[0]
        if isinstance(first, str) and first:
            return True
        if isinstance(first, Mapping):
            return isinstance(first.get("url"), str) and bool(first["url"])
    return False

try:
    url = _extract_output_url(result, context="sdxl")
except ValueError as e:
    if "Unexpected response" in str(e):
        log.error("Model output schema not URL-shaped: %r", result)
    raise

When it happens

Trigger: The prediction succeeded but output is null (some models emit output artifacts elsewhere, e.g. only in `data` or `logs`); output is a dict with the image under a different key (e.g. "image" instead of "url"); output is an empty list because the model returned zero images.

Common situations: Switching call_replicate_model to a new model whose output schema differs (e.g. returns a list of dicts without "url", or a nested structure); models that stream results to a different field; output empty because the model produced no image.

Related errors

• Prediction ID not found in initial response.
• Invalid prediction status response.
• Invalid prediction creation response.
• Inference errored out: {error_message}
• Inference failed

AI-assisted analysis of abi/screenshot-to-code@d026163f58 (2026-08-14). Data as JSON: /api/errors/047c7748b16f1597. Report an issue: GitHub.