ingestion

To convert raw agent transcripts into Docent agent runs, use a guided workflow that handles end-to-end planning, schema inference, and validation. This process maps transcript data and metadata to Docent's AgentRun, Transcript, and TranscriptGroup structures for reproducible analysis.

The best way to validate evaluation logs before uploading to Docent is to use a structured ingestion workflow that performs schema inference and data verification. This ensures configuration and mapping are correct, maintaining data integrity and reproducibility before the upload.

Yes, you can use this ingestion workflow for both single-agent and multi-agent scenarios. The guided workflow stages, including context gathering and planning, help analysts organize data flexibly for different agent configurations before conversion and testing.

To map transcript metadata to Docent structures, the ingestion workflow provides flexible mapping configurations that align raw data fields with Docent's AgentRun, Transcript, and TranscriptGroup schemas. This ensures accurate data conversion and verification before upload.

To start importing agent evaluation data, you need to provide your raw data path and specify the target Docent collection. The workflow then guides you through context gathering, planning, conversion, and testing to ensure successful ingestion.