Versioned public contract

RepoRadar Data API v1

Machine-readable, partitioned decision intelligence with explicit schemas, build identity, generation time, exact retained GitHub star observations, full change events, dated snapshots with explicit coverage, and lazy item details.

2520entities

1853latest events

49dated snapshots

2026-06-16history begins

Endpoints

GET /data/v1/manifest.json
GET /data/v1/curated.json
GET /data/v1/verification-cohort.json
GET /data/v1/changes/latest.json
GET /data/v1/changes/stream.json
GET /data/v1/changes/archive/index.json
GET /data/v1/comparisons/index.json
GET /data/v1/snapshots/index.json
GET /data/v1/history/gh-openai-codex.json
GET /data/v1/test-logs/run-<hex-encoded-run-id>.txt
GET /data/v1/schema.json

Every core and detail record exposes its exact history_url; entity IDs are not URL templates. Each search-core record also names its lazy detail shard. Consumers must check schema_version and build_id before joining artifacts.

Open manifest →

History coverage

This history is not contiguous. It holds 49 dated snapshots beginning 2026-06-16, but 10 dates in that range have no snapshot, because no catalog was produced on those days. They are not reconstructable and RepoRadar does not interpolate them; a comparison that spans one of these dates is measuring across a gap, not across a quiet period.

2026-07-03
2026-07-05
2026-07-15
2026-07-16
2026-07-22
2026-07-23
2026-07-24
2026-07-25
2026-07-26
2026-07-27

The same list is machine-readable as snapshot_missing_dates in the manifest and as coverage.missing_dates in /data/v1/snapshots/index.json.