# Teahann Knowledge Base

> Reference pages, structured datasets, and a directed graph of their links.

## Start here

- [AI access](https://teahann.com/ai): download and connection instructions.
- [Manifest](https://teahann.com/ai/v1/manifest.json): snapshot, coverage, sizes, SHA-256 hashes, and domain downloads.
- [Subject packs](https://teahann.com/ai/v1/topics.json): small plain-text files for chat applications.
- [Sources and licences](https://teahann.com/sources): original dataset provenance and terms.

## Retrieval

Use the public read-only MCP endpoint https://teahann.com/mcp with tools search, get, batch, neighbors, and context.
Alternatively, GET /api/v1/search?q=SUBJECT or /ai/context?q=SUBJECT&max_chars=24000.
Fetch pages by persistent ID or canonical site path using /api/v1/entity?id=ID.
Follow next_offset when has_more is true. Context bundles are selections; inspect truncated and omitted_candidates.
Character limits are not tokenizer-specific token limits. Source citations are inherited from pages, not independently verified by the exporter.
Graph edges distinguish links_to, related_to, and in_category. Hyperlinks do not imply scientific relations.
Page IDs identify documents; different domain views are not automatically merged into real-world entities.

## Bulk access

- [SQLite](https://teahann.com/ai/v1/knowledge.sqlite.gz): decompress for offline full-text search and graph queries.
- [Schema](https://teahann.com/ai/v1/schema.json): tables and record conventions.
- [Original datasets](https://teahann.com/ai/v1/datasets.jsonl.gz): original top-level JSON data, including supplemental fields.

The manifest also lists compressed entities, passages, edges, references, and per-domain bundles.