rbrooks/TeaLeaves
1
0
Fork 0
Code Issues 17 Pull requests Projects Releases 10 Packages Wiki Activity Actions
10 releases 10 tags
  • v5.0.0 7b7bf4e6ad

    v5.0.0 Stable

    rbrooks released this 2026-06-19 16:28:17 +00:00 | 143 commits to main since this release

    Tea Leaves v5.0.0

    Focused on capturing more kinds of source material and finding what you already have faster: audio transcription, AI analysis of specific image regions, a more expressive search syntax, saved/alerting searches, and semantic "more like this" recommendations.

    Highlights

    • Audio capture & transcription — record voice notes directly in the capture flow, or upload audio files; transcribed automatically via a Whisper-compatible endpoint.
    • Annotation region AI analysis — select a specific region of an image and get an AI description + tag suggestions for just that region, not the whole image.
    • Advanced search syntax — inline field:value filters (e.g. tag:concert source:bluesky), quoted phrases, and boolean operators, with an in-app syntax reference.
    • Saved searches & alerts — save any search by name for one-click reuse; optionally get a webhook notification when a new entry matches.
    • "More Like This" — every entry and search result can surface semantically similar entries using existing embeddings, no new infrastructure required.

    Everything in this release

    • P1 — Audio Capture & Transcription: voice recording in the quick-capture flow, audio file uploads, async transcription worker, inline transcript with "use as entry content."
    • P2 — Annotation Region AI Analysis: crop-and-analyze mode in the annotation canvas; reuses the existing image description/tagging pipeline scoped to a selected region.
    • P3 — Advanced Search Query Language: websearch_to_tsquery-based parsing, inline field filters, phrase and exclusion syntax, in-app help panel.
    • P4 — Saved Searches & Search Alerts: named saved searches with one-click re-run, optional webhook alerts on new matching entries.
    • P5 — "More Like This": semantic similarity tab on entry detail pages and search results, powered by existing text embeddings.

    No breaking changes to existing data — all new columns are additive with safe defaults.

    Downloads