Skip to main content
This section will help you upload your data to Docent. Our recommended path is to use the Docent Agent to ingest logs after running your evaluations. You can also trace your agents to capture data as they run. Before you start, the Data Models Overview covers Docent’s core abstractions (collections, agent runs, transcripts, metadata) and what belongs in each.

​
Quickstart

To ingest your data, point the Docent Agent to a directory containing your trajectories. For best results, sort your trajectories by format before invoking the agent. 
/docent Ingest the trajectories at <PATH_TO_DATA>
The agent may ask about how you want to organize your data within Docent. See the Data Models Overview for how Docent structures your data into collections, agent runs, transcripts, and metadata.
Using Inspect, Harbor, or NeMo-Gym? The integration guides handle those formats directly.

​
Best practices

  • Ingest one trajectory format at a time. If you have multiple formats, sort your trajectories by the scaffold that generated them (e.g., openhands/, mini-swe-agent/, custom/) before invoking /docent.
  • Include metadata relevant to your analysis. Fields like reward, model name, and task ID enable downstream filtering, DQL queries, and rubric evaluation.
  • Verify your uploaded data in the web interface. After upload, check that transcripts display correctly and metadata appears where you expect. It’s normal to iterate a few times on different organization structures.

​
Other paths

Docent ingestion SDK

The Python SDK the Docent Agent writes under the hood. Useful if you want to write an ingest script by hand, understand what the agent produced, or correct it.

Supported frameworks

Dedicated guides for Inspect, Harbor, and NeMo-Gym.