Avocet is a TUI for browsing your Raindrop.io bookmarks. It is written in Python with the Textual framework. It syncs your collections and bookmarks into a local SQLite cache, generates a concise summary of a bookmark — using either Anthropic Claude or OpenAI — the first time you open it (cached thereafter), and opens links in your default browser.

• Python 3.12+
• uv

1. Clone the repo.
2. uv sync (or just install) to install dependencies.
3. Provide the required credentials, either as environment variables or in a .env file (copy .env.example to .env and fill it in — exported variables take precedence over .env):
   • RAINDROP — your Raindrop.io API token (from app.raindrop.io/settings/integrations).
   • A summary provider key — set one of ANTHROPIC_API_KEY or OPENAI_API_KEY. Avocet auto-detects which to use from the key you provide (Anthropic wins if both are set). Override with AVOCET_SUMMARY_PROVIDER=anthropic|openai, and the model with AVOCET_SUMMARY_MODEL (defaults: claude-haiku-4-5 / gpt-5-mini).
4. just run (or uv run textual run --dev avocet/app.py).
5. Navigate with the arrow keys. Move focus between the collections sidebar and the bookmarks table; press Enter on a bookmark to view its details — a Claude summary is generated on first view and cached. Useful keys: o open in browser, / search, a/e/d add/edit/delete, f filter by tag, r refresh from Raindrop, and ctrl+p for the command palette (including runtime theme switching). ctrl+c quits.

Summaries are generated lazily the first time you open each bookmark, so startup is instant.

Common tasks (see the justfile for all of them):

just test            # uv run pytest
just lint            # uv run ruff check .
just typecheck       # uv run ty check
just snapshot-update # regenerate visual snapshot baselines after a UI change

The test suite has three layers: unit tests (Raindrop API client and database manager), interaction tests driven through Textual's run_test/Pilot, and visual-regression snapshot tests via pytest-textual-snapshot.