Generate training pairs
Generate instruction fine-tuning pairs from your data for use in alignment jobs.
Authorizations
Your Seekr API key, sent in the Authorization header with no 'Bearer' prefix.
Body
Task description/instructions for the alignment task
List of file ids to use for alignment
Optional system prompt baked into the generated dataset's system role on every conversation. If omitted, one is generated from instructions and the recursive document summary.
Type of alignment task (principle or context_data)
principle, context_data Optional vector database id to use for context retrieval during context_data alignment
Response
Successful Response
pending, queued, running, cancelled, failed, completed, stopped The processing step the job is currently working on. null until the job starts running.
Progress through the current step. null when no step is in progress.
Overall completion of the job across all steps. null until progress can be calculated.
Projected completion timestamp (UTC). null until enough progress data is available to produce an estimate.
Estimated number of minutes remaining until the job completes. null until an estimate is available.
Human-readable description of the job's current state, suitable for display to end users.