Export an eval run as a fine-tuning JSONL dataset
Export a completed eval run’s scored results as a fine-tuning JSONL dataset in the OpenAI, Anthropic, or HuggingFace shape (query format, which is required). The completion side defaults to the agent’s judged actual output (distill the best transcripts) or the golden expected_output; optional min_score filters to judge-passed rows only. Returns the JSONL as a file download (no JSON envelope). 404 when the run does not exist for this agent; 422 when no rows qualify after filtering. Orbit produces the dataset only — the customer fine-tunes on their own provider account.
Authorizations
Dashboard JWT token from Clerk
Headers
Sandbox opt-in for Clerk-session-authenticated requests. Set to true to route the call through the test-mode pipeline: no real provider delivery, no credits deducted, response meta.test_mode: true. Ignored for live API keys (dv_live_sk_*) — server-to-server clients must use a test-prefixed key (dv_test_sk_*) to exercise sandbox. Test-prefixed keys unconditionally enable sandbox regardless of this header.
true, false Response
Successful response.