Skip to main content

Explore a Run Dataset

pChronicle gives you one interface for Agent runs stored locally, in object storage, or behind a configured pin. The inspect, find, analysis, and query commands in this walkthrough are read-only.

1. Try pChronicle without preparing data​

pchronicle onboard

The walkthrough creates a temporary example Dataset and introduces the main commands. To jump directly to querying:

pchronicle onboard query

Neither command requires a source checkout or an existing Dataset.

2. Inspect a Dataset you already have​

The onboarding Dataset is temporary and is removed when the walkthrough ends. For a persistent query, point the commands below at a Dataset path you already own. If you do not have one yet, stop after the onboarding query and continue with Discover and query your own data.

A Dataset may be a local path, an object-store URI prefix, or a pin such as @prod:

pchronicle list ./trajectory-data
pchronicle stats overview ./trajectory-data

list (ls) shows the run data pChronicle can use. stats overview gives a stable summary without requiring SQL.

To locate content, use the unified find --match syntax:

pchronicle find ./trajectory-data --match "timeout" --format json
pchronicle find ./trajectory-data --match '#system("retry")'

3. Ask a specific question​

pchronicle query ./trajectory-data \
  --sql 'SELECT source, COUNT(*) AS steps
         FROM dataset.steps
         GROUP BY source
         ORDER BY source'

Queries are read-only and limited by explicit resource budgets. pChronicle normalizes supported run data formats into common runs, steps, and tool_calls tables where their semantics align.

What you completed​

You opened one Dataset, ran a built-in summary, queried a normalized table, and can now locate specific trajectories with FTS/JSONB. The Dataset was not modified.

Continue by task: