1. Add your files
Drop files or a folder here
eval.json, runs.json, *.output_items.jsonl, dataset *.jsonl, prompt object *.json, or plain .txt/.md prompts
No files loaded.
Need the export first? The command-line tool in the project README downloads your evals, runs, output items and datasets with read-only GET requests: OPENAI_API_KEY=... openai-exit-kit export --out evals-export. Or use the data you already downloaded from the dashboard.
2. Review and download
Warnings
-What to trust
- Every grader is labeled exact, approximate or manual in the report. Manual ones are written as deliberately failing stubs so they cannot pass silently.
- No model is called and no score is computed here. Run
promptfoo evalyourself and compare to the exportedbaseline.json. - Agent Builder workflows are not converted. Export those from the Agent Builder UI.
- Not affiliated with OpenAI or promptfoo.