Settings

Health, activity, and advanced config. You can usually ignore this page.

What's in the knowledge base right now.

Documents

Chunks

Q&A

What needs attention. Crons clear most of this automatically.

Ingests, edits, re-embeds, extractions.

Last 8

Service area — zip codes we cover
Extracted 8 service-area zips
system
3d ago
Booking SOP — how to schedule a new appointment
7 chunks embedded with text-embedding-3-small
system
3d ago
Cancellation SOP — trip fees and rescheduling windows
Manual re-embed after edit
Dave Morrison
3d ago
qa_created
Created Q&A: Can I reschedule my appointment?
Dave Morrison
4d ago
Reschedule policy — 4-hour window
Manual paste, 1 chunk
Dave Morrison
4d ago
Supplier list and reorder cadence
Drive API quota exceeded — retry scheduled
system
4d ago

Embedding + extraction LLM cost, scoped to knowledge-base work only.

last 30 days · 0 calls

Last 24h

$0.0000

Last 30d

$0.0000

By purpose

No KB LLM calls yet.

What we use to turn text into vectors. Changing this re-embeds every chunk in the background.

Current model

OpenAI text-embedding-3-small is the locked default for all tenants during beta.

Changing the model re-embeds every chunk. Estimated cost for this tenant: $0.08. Runs in the background and may take minutes.

Applies to NEW ingests. Existing documents stay on whatever chunking was used at ingest time until resynced.

Chunk size (tokens)

Target size for each chunk. 128 to 2048. Default 512.

Overlap (tokens)

Tokens carried between adjacent chunks. 0 to 256. Default 64.

How long we keep soft-deleted documents and audit events.

Undo window

Soft-deleted documents recoverable for this many days.

Audit log retention

Event log (ingests, edits, re-embeds) kept for this many days.

Test a query and see what the bot would get back. Admin only.

Test query

Runs structured-fact lookup, Q&A match, and chunk retrieval in order.

Who can edit documents, chunks, and Q&A pairs. Everyone else on the tenant is read-only.