Pantograph is a local-first desktop app that turns sketches and prompts into editable Svelte UI.

1. Clone the repository.
2. Install dependencies:

   npm install

3. Run the desktop app:

   npm run dev:desktop

• Node.js (for npm)
• Rust toolchain (cargo, rustc)
• Tauri system libraries for your OS

Install Rust:

curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
source "$HOME/.cargo/env"

Install Tauri system dependencies:

# Debian/Ubuntu
sudo apt install pkg-config libsoup2.4-dev libjavascriptcoregtk-4.0-dev

# Fedora
sudo dnf install pkgconf-pkg-config libsoup-devel javascriptcoregtk4.0-devel

# Arch
sudo pacman -S pkgconf libsoup2 webkit2gtk

Install project dependencies:

npm install

npm run dev:desktop

npm run dev

npm run build:desktop

./launcher.sh --help

• External OpenAI-compatible server (for example LM Studio)
• Bundled llama.cpp sidecar with local model files

• Node.js + npm
• Rust toolchain
• Tauri system dependencies (above)

# Lint (configured scope)
npm run lint

# Full lint scan
npm run lint:full

# Critical anti-pattern gate (src/ + packages/)
npm run lint:critical

# Type check
npm run typecheck

# Tests
npm test

# Runtime separation guard (no compile-time Python linkage)
npm run test:runtime-separation

# All quality gates
npm run check

Python-backed model execution is intentionally out-of-process and externally provisioned. See docs/python-runtime-separation.md for configuration and migration details. For a local diffusion worker smoke path, run:

./.venv/bin/python scripts/diffusion_cli_smoketest.py --model-path /path/to/tiny-sd-turbo

Pantograph exposes a Rust-first headless workflow API for host integrations through crates/pantograph-workflow-service:

• workflow_run
• workflow_get_capabilities
• workflow_get_io
• workflow_preflight
• create_workflow_session
• run_workflow_session
• close_workflow_session
• workflow_get_session_status
• workflow_list_session_queue
• workflow_cancel_session_queue_item
• workflow_reprioritize_session_queue_item
• workflow_set_session_keep_alive

Integration boundary:

• Headless hosts should integrate with the core API/service crate directly.
• src-tauri commands are desktop app transport adapters, not the headless API.
• HTTP binding exports are opt-in frontend adapters for modular standalone GUI hosting (frontend-http in UniFFI and Rustler).
• Recommended headless flow: workflow_get_io -> workflow_preflight -> workflow_run.

Reference docs:

• Contract: docs/headless-embedding-api-v1.md
• Migration guide: docs/headless-embedding-migration.md
• Service boundary ADR: docs/adr/ADR-001-headless-embedding-service-boundary.md
• Implementation notes: docs/headless-embedding-implementation-notes.md

1. Create a focused branch for one logical change.
2. Follow coding, tooling, accessibility, and documentation standards.
3. Run npm run check and relevant targeted Rust tests before opening a PR.
4. Use Conventional Commits for all commits.

Workspace crates declare MIT OR Apache-2.0 in Cargo metadata. Review individual package metadata for any exceptions.