AI chat
Ask your archive anything — it will cite its sources.
A retrieval-augmented chat that never steps outside your collection. Every answer links to the item, page, or timestamp where it came from. No hallucinations, no content drift.
pgvector retrievalOpenAI embeddings
- Grounding
- Your records only
- Citations
- Every claim
- Scope
- Per-tenant
archively.ai/chat
Grounded in your holdings. Nothing else.
Semantic retrieval over every published record, transcript, and OCR text. The model writes an answer only from what it retrieved — and cites every claim to a source.
- Cites item, page, and (for media) exact timestamp
- Filter chat scope to a fonds, a series, or a date range
- No external web calls in answer generation
archively.ai/chat/c-0014
Researcher-grade controls.
Switch between strict (grounded-only) and exploratory modes. Pin sources. Export the whole conversation with inline citations for a research note.
- Strict / exploratory modes
- Pin and quote specific records
- Export Q&A with bibliography to Markdown
Safer than general chat.
Because retrieval is scoped to your published layer, the model can't answer from drafts, embargoed material, or public-web noise. Privacy and accuracy by construction.
- Respects embargo and access-restriction flags
- Curator-configurable sensitive-record masking
- Every Q&A logged for audit
Ready to see it on your collection?
Load a few records. Run the AI. Review and publish — before lunch.