Recipes · Recipe

Connect a repository

Point Chiron at a GitHub or Azure DevOps repo and watch it index.

Goal: point Chiron at a repository and watch it become a queryable knowledge source.

Everything in a connected repo gets read into the ontology; you separately choose which repositories agents may push to. This recipe covers the connect step and how to confirm ingestion succeeded.

1. Add the repository to a project

From the project's settings, add the repository URL. Chiron detects the provider from the host:

text

https://github.com/acme/platform        -> github
https://dev.azure.com/acme/_git/platform -> azure_devops
https://gitlab.com/acme/platform         -> gitlab

Both HTTPS and SSH-style URLs work. The repo needs credentials with read access — a PAT for the provider — stored encrypted.

2. Watch it index

The project's setup_status moves pending -> analyzing -> ready. Under analyzing, Chiron parses code into modules and symbols, builds the graph, generates docs, and begins syncing commits, PRs, and work items.

bash

curl "$ONTOLOGY_URL/projects/$PROJECT_ID/knowledge/discover" \
  -H "X-API-Key: $INTERNAL_SERVICE_KEY" | jq '.repos, .sync'

3. Confirm each source is healthy

The sync array in discover reports a status per (provider, source):

text

ok       synced successfully
running  in progress
queued   accepted, no worker yet
error    check last_error (often an expired PAT)

An error on commits with an empty work_items is the classic "credential expired" signature — the sync failed, so the source looks empty. See Sync & freshness.

Once ready, run your first search.

Next: Trace a feature from work item to code.