Notes → Chatbot knowledge base
This is the feature the mobile app is for — the reason Hilal made one app instead of two. Today, mobile users capture great content in AI Notes — the warehouse photo of a defect, the voice memo from the vendor call, the typed-up FAQ entry from a Slack thread. Tomorrow, that content can land directly in your chatbot’s knowledge base, where it answers customer questions automatically. No web round-trip, no copy-paste, no “I’ll do that later.”
What this will let you do
- Tap “Add to chatbot KB” on any note to ingest it as a knowledge chunk.
- Pick which chatbot receives the note (if you have several).
- Approve the chunk preview — see what the bot will see.
- Track ingested notes — every note that’s been added to a KB shows a small “in KB” badge.
- Remove from KB — pull a note out of the KB if it stops being relevant; the note itself stays in your Notes.
The mobile app already has the schema in place — every note carries an embedding_id field, currently empty. When you opt in, the embedding pipeline runs server-side, fills the field, and the note becomes a searchable chunk in your chatbot’s vector index.
How it’ll work (preview)
- You write a note. Type, dictate, attach files. Same as today.
- You tap “Add to chatbot knowledge” in the note’s actions menu.
- Pick a chatbot (or it defaults to your active one).
- Hilal generates a chunk preview. What text gets indexed, what gets skipped (e.g., file metadata), what tags carry over.
- You approve. The chunk is embedded, stored in your bot’s vector index, and the note flips to “in KB.”
- The bot can now answer customer questions using that content.
When you edit the note later, the chunk re-syncs automatically.
Why this matters
The hardest part of running a chatbot is keeping its knowledge base fresh. Every operator we’ve talked to has the same flow: Slack thread happens → “we should add this to the bot” → no one ever does. The mobile-note → KB pipeline kills that gap. The capture surface is already on the phone you have on you. Hilal makes the ingest one tap.
This is the feature that makes “one app for both products” earn its name.
Tags as ingestion signals (suggested pattern)
Once this ships, the natural pattern is to use a tag like for-bot to mark notes you intend to ingest. You can then bulk-ingest from the Notes list filtered to that tag, or set up an auto-ingest rule (also coming soon) so any note you tag for-bot is queued automatically.
Status
- Schema: ✅ in place. The
embedding_idfield exists on every note record. - Frontend: UI design in progress.
- Backend: the embedding service is shared with the existing chatbot ingest pipeline; the integration glue is the missing piece.
- Beta: planned next quarter; we’ll DM the TestFlight/internal-build users when it lands.
What you can do today to prepare
- Adopt the
for-bottag pattern. Tag any note you’d ingest if the feature were live. - Categorize notes intentionally. When ingest goes live, you’ll likely want to filter by category before bulk-ingesting.
- Keep your Hilal Chatbot’s existing knowledge base tidy. The mobile pipeline adds chunks — it doesn’t replace anything. If your KB is messy today, it’ll be messy tomorrow.