Machine-readable access

Encode Wiki publishes the same public content in deterministic, machine-readable forms. These interfaces are generated during the static build and require no account or API key.

Start with the index

/api/v1/index.json enumerates every public document and article. Each record contains its stable ID, canonical page URL, clean Markdown and JSON retrieval URLs, status, tags, dates, authors, source and license metadata, headings, outbound references, byte count, and SHA-256 checksum.

Retrieve an individual record using the URL in its apiUrl field. The corresponding markdownUrl returns plain Markdown with a short attribution header. Do not construct URLs from titles; follow the URLs in the index so nested identifiers remain opaque to clients.

Contracts and release detection

• /api/v1/openapi.json describes the read-only API using OpenAPI 3.1.
• /api/v1/schema.json is the JSON Schema for an individual corpus record.
• /api/v1/release.json reports the Git content version, deterministic timestamp, corpus checksum, counts, and knowledge-bundle checksums.

The API path changes only for breaking contract revisions. Compare schemaVersion before parsing and contentVersion or contentChecksum to detect a new release. Page checksums cover cleaned Markdown only.

Bulk and agent-oriented files

• /llms.txt is a compact curated map of high-value pages and complete interfaces.
• /llms-docs.txt and /llms-articles.txt contain full, cleaned collection text.
• /llms-full.txt combines both collections and may be too large for smaller context windows.
• /api/v1/knowledge.ndjson.gz is the complete corpus as gzip-compressed newline-delimited JSON for ingestion by future search, Ask AI, or MCP services.

All API and text responses permit cross-origin reads and use short shared-cache lifetimes. The downloadable .gz file is an actual gzip payload, not an HTTP content-encoding wrapper. Its uncompressed and compressed checksums are in the release record.

Attribution and content status

Machine clients must retain each record’s canonicalUrl, license, sources, provenance, and modification notice when redistributing migrated material. Treat needs-review content as potentially stale and archived content as historical. The API does not turn those records into current recommendations.