Connectors

Why This Matters

Without Attest’s connectors, here’s what you’d build for each data source:

• API fetching + pagination — Slack cursors, Gmail page tokens, Jira JQL offsets, Graph API paging
• Claim extraction — LLM prompt engineering to turn messages and documents into structured (subject, predicate, object) triples
• Entity normalization — Unicode NFKD, Greek letter expansion, whitespace collapse, deduplication across sources
• Provenance tracking — record which source, which page, which message produced each fact
• Contradiction detection — when Slack says X and Confluence says not-X, flag it instead of silently overwriting
• Embedding updates — re-compute or re-sync a separate vector store every time new data arrives

With Attest, one line handles all of it:

db.connect("slack", token="xoxb-...", channels=["#research"]).run(db)

Each connector automatically runs the full pipeline: fetch → extract → normalize → validate (13 rules) → store with provenance → update embeddings → track corroboration.