Guides¶
Task- and concept-oriented walkthroughs of Refindery's capabilities. Each guide explains how the feature works and links to the precise API contract in the Reference.
- Ingesting pages — capture → index → enrich, the page lifecycle, and revisit semantics.
- Watches — poll RSS/Atom feeds and YouTube playlists/channels on a schedule and auto-ingest new items.
- Searching — the hybrid retrieval pipeline, filters, recency decay, and A/B compare.
- MCP for agents — connect an LLM agent, read vs mutating tools, and grounding.
- Evaluation — score and replay retrieval quality from the query log.
- Clustering — resurface reading themes with stable cluster IDs.
- Entities — extraction chains and corpus-internal canonicalization.
- Embedding models — register, backfill, activate, and compare models.
- Deletion & blacklist — purge content and block re-ingestion atomically.
The Architecture overview ties these together into the ingest → index → cluster data flow and the search pipeline.