readwise-reader

Fetch, analyze, and triage Readwise Reader documents.

Usage

# 1. Fetch all Inbox + Later documents from Readwise Reader API
readwise-reader fetch

# 2. Analyze with LLM (default: Claude Opus 4-6 via Vertex AI)
readwise-reader analyze

# Or use a different model:
readwise-reader analyze -m sonnet    # Claude Sonnet 4 (faster, cheaper)
readwise-reader analyze -m gemini    # Gemini 3 Pro Preview (fastest)
readwise-reader analyze -m gemini25  # Gemini 2.5 Pro

# Resume interrupted analysis (checkpoints automatically):
readwise-reader analyze

# Start fresh:
readwise-reader analyze --reset

# 3. Generate interactive HTML triage report
readwise-reader report
readwise-reader report --no-open    # Don't auto-open browser

Profile

User profile is loaded from $XDG_DATA_HOME/readwise/profile.toml (default: ~/.local/share/readwise/profile.toml).

The profile controls:

• User interests — what topics are relevant to you
• Favorite authors/sites — auto-boosted to 4★+ regardless of topic
• Summary style — brief or detailed

See the example in this directory or create your own.

Data

All data stored in $XDG_DATA_HOME/readwise/:

• reader-latest.json — symlink to latest fetch
• reader-analyzed.json — enriched with LLM analysis
• analysis-checkpoint.json — resume point for interrupted analysis
• triage-report.html — generated report
• profile.toml — user profile

Requirements

• passage — for Readwise API token (readwise/key)
• gcloud — for Vertex AI auth (Opus/Sonnet models)
• GEMINI_API_KEY env var or passage — for Gemini models
• requests Python package

Report Features

• Documents grouped by action (Must Read, Finish, Archive, Delete, etc.)
• Within each group, clustered by topic (Nix, Go, Emacs, AI, etc.)
• Relevance scores (1-5★) with LLM-generated summaries and reasons
• Click summaries to expand full analysis
• Filter by title, relevance, or tag
• Collapsible sections
• Dark theme