Package Exports
- @remnic/connector-bee
Readme
@remnic/connector-bee
Bee wearable connector for Remnic. Pulls your Bee bracelet conversations into Remnic's wearable-transcript pipeline: cleaned, speaker-labeled, redacted, searchable day transcripts — and, under per-source trust gates, memories. Optionally imports Bee's own extracted "facts" into the review queue.
This is an à-la-carte optional companion of @remnic/core:
npm install -g @remnic/connector-beeRemnic discovers it at runtime. No further registration is needed.
Setup
Heads up: Bee's original public API (
api.bee.computerwithx-api-key) was retired after the Amazon acquisition. This connector targets the current developer surface.
Option A — local proxy (recommended, zero config)
- Enable Developer Mode in the Bee app (Settings → tap Version
5 times), then pair the CLI:
npm install -g @beeai/cli bee login bee proxy # serves the developer API on http://127.0.0.1:8787
- Enable the source — no token needed:
{
"wearables": {
"enabled": true,
"sources": {
"bee": {
"enabled": true,
"memoryMode": "smart", // smart (default) | off | review | auto
"importNativeMemories": "smart" // Bee facts through the same trust pipeline
}
}
}
}Option B — direct API access
Set baseUrl to the direct host and provide the bearer token from
bee login (~/.bee/token-prod) via BEE_API_TOKEN (or
REMNIC_BEE_API_TOKEN, or apiKey in config). The direct host uses
Bee's private CA — export NODE_EXTRA_CA_CERTS pointing at it.
Usage
remnic wearables check bee
remnic wearables sync --source bee --days 7
remnic wearables transcript --date 2026-06-10 --source beeSpeaker labels
Bee exposes opaque diarization labels ("0", "1", ...) with no wearer marker. Map them once and every stored transcript uses the names:
remnic wearables speakers set bee 0 "Your Name" --self
remnic wearables speakers set bee 1 "Jane Doe"Notes
- Bee's list API has no date filter; the connector paginates newest-first and filters to the requested local day, so backfills stay bounded.
- Conversations still in the
CAPTURINGstate are skipped until they settle; the next sync picks them up. - Default
memoryMode: "smart": the LLM judge + per-source trust prior- cross-device corroboration write high-trust facts active, queue borderline ones, and drop the rest. Bee-native facts run through the same pipeline with a reduced prior.
Full documentation: docs/wearables.md.
License
MIT