Skip to content

Repository Intelligence API ​

Source-linked repository metadata, observed activity, published release revisions and original TNL release reports. Initial coverage is a curated public GitHub watchlist, not a complete repository crawl.

RouteReturns
GET /v1/intel/repositoriesLatest collected repository snapshots
GET /v1/intel/repositories/:idProject context, latest activity, AI/ADES status, paper/model references
GET /v1/intel/repositories/:id/versionsRetained repository snapshots
GET /v1/intel/repositories/:id/releasesPublished release-note revisions, including labeled prereleases
GET /v1/intel/repositories/:id/activityTimestamped counters, bounded samples and observed deltas
GET /v1/intel/repositories/:id/reportsOriginal grounded release reports

Use the existing TNL Bearer API key, email verification and plan quota. IDs are gh- followed by GitHub's numeric repository ID, so identity survives renames. Discover IDs in the collection; do not substitute an owner/repository path.

bash
curl 'https://theneuralledger.com/v1/intel/repositories?owner=ollama&limit=10' \
  -H "Authorization: Bearer $TNL_API_KEY"

Collection, versions and reports accept owner, language, q, since, before, limit (1–100) and cursor. Releases accept q, dates and paging; activity accepts dates and paging only. Dates filter observation time for snapshots/activity, release publication time for releases, and TNL publication time for reports. Responses contain data, nextCursor, and schemaVersion. Invalid filters return 400, unauthorized requests 401, unknown repositories 404, and temporary failure 503.

Repository snapshots retain the observed default-branch commit, declared license identifier, language, topics, archive/fork status and source links. A README is fetched at the exact commit. Current project context is not necessarily the code or documentation at a historical release tag. Raw README and release-note bodies remain internal; APIs expose normalized metadata and original TNL prose.

Release identity uses GitHub's numeric release ID plus an immutable content revision. The release collection returns retained revisions, so one release can appear more than once after edits. Drafts are excluded. Prereleases are labeled and retained but do not trigger automatic reports. A push or an ordinary tag is not treated as a published release. Upstream deletion or editing never removes a TNL publication.

Activity includes stars, forks and openIssuesIncludingPullRequests, plus samples of up to 20 contributor entries and five recent default-branch commits. These are not total contributors, active users, all development activity or adoption. delta is null until two distinct ordered observations exist. It then reports net changes and the actual observation interval, including negative changes. No unobserved historical growth or future growth is invented.

Every accepted snapshot, release and activity record is registered for AI/ADES digestion. intelligence exposes pending/processing/retry/complete status, source evidence, entities and conditional potential effects. Shared /v1/intel/digests filters include repository_snapshot, repository_release, and repository_activity. Registration is not completed analysis.

Model and paper links are explicit references from project documentation, not ownership claims. Exact matches to already collected models/papers include their TNL intelligence IDs; unmatched links remain unresolved.

Original reports are available at Open-Source Intelligence. Reports require completed digestion of both repository context and release notes, substantive evidence, original synthesis and a separate grounding review. TNL has not independently executed, benchmarked or security-audited the code. Earlier reports and supporting source revisions remain permanently available.

The Neural Ledger API