What causes "OpenAI API key not found" in abi/screenshot-to-code?

Set OPENAI_API_KEY in the environment of the eval process and restart it.

abi/screenshot-to-code · critical · Exception

OpenAI API key not found

Error message

OpenAI API key not found

What it means

Exception raised by the eval runner core when model is in OPENAI_MODELS and the module-level OPENAI_API_KEY is empty. It fires before the AgentRunRecorder is created, so no partial run record is left behind.

Source

Thrown at backend/evals/core.py:46

    eval_session_id: str | None,
    input_file: str | None,
) -> str:
    async def send_message(
        _: str,
        __: str | None,
        ___: int,
        ____: dict[str, Any] | None = None,
        _____: str | None = None,
    ) -> None:
        # Evals do not stream tool/assistant messages to a frontend.
        return None

    if model in ANTHROPIC_MODELS and not ANTHROPIC_API_KEY:
        raise Exception("Anthropic API key not found")
    if model in GEMINI_MODELS and not GEMINI_API_KEY:
        raise Exception("Gemini API key not found")
    if model in OPENAI_MODELS and not OPENAI_API_KEY:
        raise Exception("OpenAI API key not found")

    print(f"[EVALS] Using agent runner for model: {model.value}")

    recorder = AgentRunRecorder(
        generation_id=(
            f"gen_{datetime.now().strftime('%Y%m%d_%H%M%S')}_{uuid.uuid4().hex[:8]}"
        ),
        variant_index=0,
        entry_point="eval",
        stack=str(stack),
        input_mode=input_mode,
        generation_type="create",
        eval_session=eval_session_id,
        eval_set=eval_set,
        input_file=input_file,
    )
    runner = Agent(
        send_message=send_message,

View on GitHub (pinned to d026163f58)

Solutions

Set OPENAI_API_KEY in the environment of the eval process and restart it.
Note this path does NOT read the in-app Settings key — env/.env is the only source here.
Or select an Anthropic/Gemini model matching an available key.

Example fix

# backend/.env
OPENAI_API_KEY=sk-...

Defensive patterns

Strategy: validation

Validate before calling

import os
if model in OPENAI_MODELS and not os.environ.get("OPENAI_API_KEY"):
    raise SystemExit("OPENAI_API_KEY not set; cannot run evals on OpenAI models")

Try / catch

try:
    await run_eval(...)
except Exception as e:
    if "API key not found" in str(e):
        abort_run("Configure OPENAI_API_KEY and restart the eval worker")

Prevention

Note evals read env only — keys entered in the browser Settings do not satisfy this path.
Add env validation to eval-worker startup so misconfiguration is caught before sessions start.

When it happens

Trigger: Starting an eval run with an OpenAI model while OPENAI_API_KEY is unset in the eval-hosting process.

Common situations: backend/.env missing the OpenAI key, eval worker started in a different shell/environment, or the key only configured in the browser UI (evals core reads env only).

Related errors

AI-assisted analysis of abi/screenshot-to-code@d026163f58 (2026-08-14). Data as JSON: /api/errors/3da92545b57f5762. Report an issue: GitHub.