Reference

Every toolbar action at a glance, plus the full CLI and the honest caveats.

The toolbar is contextual: the full set shows in the notes view; the graph and chat show only a Back to notes button to keep the chrome clean. Each icon shows its name as a label.

Icon	Does	More
➕ Add	Import a course PDF, or add one PDF as a single note	Using → Add
🔗 Link	Cross-link related notes domain-wide	Connect → Link
🔄 Refresh	Reload domains & notes from the vault	below
🧠 Concepts	Build the multi-layer concept graph	Connect → Concepts
🕸️ Graph	Open the interactive graph (drill Areas → Themes)	Connect → Graph
💬 Chat	Ask your notes, grounded & cited	Connect → Chat
🌗 Appearance	Light / dark / system	below
⚙️ Settings	Models, filters, API key	below

Appearance

The appearance icon cycles system → light → dark and persists your choice. It sets the app's appearance directly, so the whole UI — including rendered note summaries and chat — follows it. Leave it on system to track macOS.

Refresh

Reloads domains and notes from the vault on disk — useful if you changed files outside the app (e.g. in Obsidian) or ran the CLI in a terminal.

Settings

The gear opens model dropdowns (transcription · concept/cross-link · chat), the content filters, your Claude Code login status, and an optional Claude API key. The concept-map, cross-link, and chat reasoning run through Claude Code by default — no API key needed.

uv run python -m engram config set transcribe claude-code
uv run python -m engram config set-key sk-ant-…   # optional

CLI reference

# domains
engram domain create | list | remove | status "Data Science"
engram domain groups "Data Science"            # section groups + sections

# ingest
engram add "Data Science" "CSE 234.pdf" [--group "CSE 234"]
engram add-page "Data Science" page.pdf --section "Lectures" [--title "…"]

# edit notes
engram note delete | move | duplicate <note.md>
engram retranscribe <note.md>

# connect
engram link "Data Science"
engram concepts "Data Science" ["Mathematics"] [--layers 3]

# chat (needs a concept graph)
engram chat query "Data Science" "…" (--concept "…" | --course "…")
engram chat synthesis "Mathematics" ["Data Science"] -q "…"

# inspect / config
engram inspect "CSE 234.pdf"
engram status
engram config set transcribe claude-code

Run as uv run python -m engram … from the repo root.

Honest limitations

Bulk auto-split is OneNote-tuned — other layouts may mis-split; use add-page for those.
Structure parsing is heuristic & locale-dependent (the 分区 … 的第 N 页 footer); an English fallback is future work.
Re-exports aren't pixel-stable, so cross-export incremental determinism isn't fully solved.
The graph rebuilds from scratch — cross-links & concepts recompute each run; incremental refresh is next.
Chat has no internet / external tools yet — the model's own knowledge over your notes only.
One-way (notes → vault); never written back.