It gathers the scrolls from the shelves in the hall, it thinks up the links that connect them all!
It takes a messy math-scribble and turns it to speed, then watches the agents to ensure they succeed.
Logic-Loom is a desktop app for research workflows. It turns a Zotero library into a searchable Obsidian vault with zotero-summarizer, and adds optional knowledge-graph tooling from AKMS, paper discovery, a computational-mechanics compiler from MechDSL, and read-only views of agent memory.
Documentation: https://cemm2.github.io/Logic-Loom/
| Mode | What it is for |
|---|---|
| zsum | Run the zotero-summarizer pipeline and browse the published vault |
| AKMS | Batch node generation, node review, the read-only graph explorer, Learn |
| Assistant | Paper discovery, source batches, workflow runs, Cmd-K workspace search |
| MechDSL | Compile % mechanics LaTeX into FEM solver code |
| Failure Memory | Inspect .failure-memory lesson registries, with an opt-in operator mode |
| Records | Browse akms-records stores and their record graph |
Each mode is described in the user guide.
Download the installer for your platform from the installer releases page, then follow Install and First run.
Requires Python 3.12, uv, Node.js 20+, pnpm 11 and Rust stable.
uv sync # core (zsum, Assistant); add --all-packages --all-extras for every mode
pnpm install
pnpm logic-loom:dev # starts the sidecar, Vite and the Tauri shell togetherTests:
uv run pytest services/logic_loom_api/tests -q
pnpm --filter logic-loom-desktop testDeveloper setup, Architecture and Build from source have the details.
apps/desktop/ # Tauri app (Rust shell + React frontend)
services/logic_loom_api/ # Python FastAPI sidecar
packaging/ # sidecar freezer entrypoints
scripts/ # build, release and docs helpers
docs/ # the documentation site (MkDocs)
The site is built from docs/ with MkDocs Material and published on every release tag.
uv sync --all-packages --all-extras --group docs
pnpm docs:serve # live preview
pnpm docs:build # regenerate the reference pages and build strictlyMIT
