Engram
GitHub ↗
Engram turns your handwritten notes — a OneNote export, or any PDF — into an Obsidian-native vault. Not a tutor: an organizer for what you've already learned. Each page becomes a searchable summary of its key ideas and formulas (original below), notes link into a concept graph, and you can chat with it — grounded and cited — to see what you know and what's still missing.
You already write notes by hand — on an iPad, a tablet, or paper you scan. Export them to PDF (from OneNote one section group at a time, e.g. one course, or bring any PDF of notes) and Engram groups several such PDFs into a single domain — treated as one combined knowledge base — and turns it into something you can actually think with. You can also drop in a single page as one note.
The folder tree, plus a multi-layer concept map — Concepts → Themes → Areas. Drill from an Area into its themes; click any node for the notes it covers.
GraphEngine (repulsion · springs · gravity · clustering). Click a node.
A searchable summary above, the original page one scroll below. Read → Connect → Use.
Step downhill along the negative gradient; the learning rate η sets the step size.
![[Gradient Descent.pdf]] # the raw leaf, embeddedChat with your notes — grounded and cited, with anything beyond your notes clearly marked. Query a course or concept, or Synthesise a whole field.
How does value iteration converge, and why?
Thinking…
Value iteration converges because the Bellman operator is a γ-contraction in the sup-norm:
‖ℬV − ℬV′‖∞ ≤ γ‖V − V′‖∞, γ < 1
so the value estimates form a Cauchy sequence with a unique fixed point V* — geometric convergence from any start.
▸ Beyond your notes — next: asynchronous / prioritized sweeping, sample-based VI.
[[sources]].An organizer for what you've already learned — local, faithful, and reproducible. It shows you the shape of your knowledge, not a feed of new material.
Engram doesn't try to teach you the material. It takes what you've already written and makes it legible to you again — a clear map of what you know, where ideas connect, and where the gaps are. You learn from your own notes; Engram just organizes them.
Each note becomes a searchable digest of its key ideas and all formulas — find and reconnect, then jump to the original page (the source of truth, embedded in every note). It filters out noise — papers, homeworks, long code — and never invents; illegible parts are flagged.
Pick the reader in Settings: Claude Code (vision, via your login — default) or a local Qwen2.5-VL 7B/32B, fully on-device. Concept-map & cross-link reasoning runs through Claude Code too — all from a dropdown.
A holistic pass builds a hierarchy of concept graphs — Concepts → Themes → Areas — that span section groups. Click an Area to drill into its themes; nodes cluster by parent, edges are shared notes (solid) + knowledge (dashed).
Ask your knowledge base. Query a course or concept, or synthesise a field. Context is pulled from the graph + your notes; answers are grounded and cited, with anything beyond your notes clearly marked.
Python 3.11+ and uv on macOS 14+ (Apple Silicon). The first run downloads the local model (~5 GB).
# 1 · set up the environment (add --extra local for the on-device model)
uv sync --extra dev --extra local
# 2 · create a domain, then add a section-group PDF (one per course)
uv run python -m engram domain create "Data Science"
uv run python -m engram add "Data Science" "CSE 234.pdf"
# 3 · cross-link the whole domain (links span section groups)
uv run python -m engram link "Data Science"
# 4 · build the multi-layer concept map → the graph
uv run python -m engram concepts "Data Science" --layers 3
uv run python -m engram statusOutput lands in ~/Engram/<Domain>/<SectionGroup>/<Section>/<Title>.{md,pdf} — open ~/Engram in Obsidian.
Prefer a GUI? Build the native app:
cd macapp && ./build_app.sh && open Engram.appEngram's whole ethos is honesty over polish. This is a working v0.2, run end-to-end on real multi-course domains (and v0.1 validated on a known notebook of 18 notes, 73 links) — but it's early, and several things are heuristic or not yet field-validated.