Tasker

A spatial task cockpit — tasks live on a 2D board, arranged by computed dimensions.

• Dev: npm install && npm run dev — server + Vite with HMR.
• Architecture: docs/architecture.md.
• Run on another machine: see below, or the full guide in docs/operations/install.md.

Run with npx / npm

The fastest way to try it — no clone, no build:

npx tasker-board                      # download + run  ->  http://localhost:5173
# or install the CLI globally:
npm i -g tasker-board && tasker

Prerequisites: Node ≥ 22 and a C toolchain for the native better-sqlite3 build (python3 + make + a C++ compiler). Configure with TASKER_* env vars — port, DB path, etc. (see docs/operations/configuration.md); the SQLite DB defaults to ./tasker.sqlite in the working directory.

Upgrade: npm i -g tasker-board@latest — your DB file is untouched.

Run with Docker

The default Dockerfile builds an image that installs the published tasker-board from npm — no source checkout needed. Data persists in the tasker-data volume.

docker compose up -d            # -> http://localhost:5173
docker compose logs -f
# or without compose:
docker build -t tasker-board .
docker run -d --name tasker -p 5173:5173 -v tasker-data:/data tasker-board

Upgrade — pull the latest release; the volume DB is kept, schema migrates on start:

docker compose build --pull && docker compose up -d
# without compose — rebuild (installs the latest) + recreate the container:
docker build -t tasker-board . && docker rm -f tasker \
  && docker run -d --name tasker -p 5173:5173 -v tasker-data:/data tasker-board

Build the image from this source tree instead (contributors / unreleased changes — last resort): docker build -f Dockerfile.source -t tasker .

Run from source (tarball — last resort)

No registry, no Docker — copy a source tarball plus two scripts.

Prerequisites (target machine): Node ≥ 22 and a C toolchain for the native better-sqlite3 build (python3 + make + a C++ compiler; preinstalled on most dev boxes and in the node:22 image).

1. On the source machine — build a versioned tarball of the current tree:

   bash scripts/package.sh        # -> tasker-<version>.tar.gz

2. Copy tasker-<version>.tar.gz to the target machine and unpack it:

   tar xzf tasker-<version>.tar.gz && cd tasker

3. Install (installs deps, rebuilds the native module, builds the web bundle):

   bash scripts/install.sh

4. Start it:

   npm start                      # serves web + API on one port (PORT, default 5173)

Update an existing machine

One line from your dev machine — push the current source to a running remote:

npm run deploy -- <ssh-target> [remote-dir]   # e.g. npm run deploy -- user@host /opt/tasker

It packages → copies → unpacks on top → runs install.sh (which backs up the DB first — server/tasker.sqlite.bak-<timestamp> — and never deletes it; the schema migrates itself on the next start) → restart.sh (portable managed restart, Linux + macOS, no systemd/pm2). No SSH? Unpack the tarball on top and run bash scripts/install.sh && npm run restart.

See docs/operations/install.md for the deploy/restart model, the situation matrix, the version marker, native-build notes, and known limitations.

Drive it from an AI client (MCP)

Tasker is an MCP control surface — Claude Code / Claude Desktop / Kiro can read + write your board. Quickest (server running): point the client at http://localhost:5173/mcp. Or a stdio command: tasker-mcp (npm install) or npm run mcp (from source). Full setup per install method: docs/operations/install.md.

License

MIT © 2026 Mikhail Chernogorsky.