Technical
Research Chat
The conversational research surface for asking CaveauAI questions, choosing routing, using corpus context, and saving useful answers back into the workspace.
Public and internal interfaces
These are the main routes, handlers, and integration surfaces used by this product area.
POST /api/portal/chat
POST /api/portal/feedback
GET /api/portal/providers/models
POST /api/portal/rag/ingest-text
Architecture notes
- The chat handler streams server-sent events from /api/portal/chat.
- Default routes use platform model lanes; BYOK models bypass corpus grounding.
- Feedback is non-blocking and tied to query IDs when the answer includes one.
Access and permissions
App pages are authenticated through the shared bnl_session cookie unless explicitly marked public. Public docs should describe integration behavior without including secrets, private tokens, private IPs, or tenant-specific records.
Troubleshooting
- HTTP 401 means the session cookie is missing or expired.
- No matching sources can be a valid result when the corpus filter is too narrow.
- Streaming failures usually surface as an HTTP status in the chat error panel.