Skip to main content

Visual graph editor, inspector, and debugger for gen-dsp signal graphs

Project description

dsp-graph

Visual graph editor, inspector, and debugger for gen-dsp signal-processing graphs.

Overview

dsp-graph provides a web UI (React + React Flow) backed by a Python server (FastAPI) for loading, visualizing, simulating, optimizing, building, and inspecting DSP signal graphs defined with gen-dsp's graph DSL.

Features:

  • Load graphs from JSON or .gdsp DSL source (toolbar or drag-and-drop), with localStorage auto-save/restore
  • .gdsp editor with syntax highlighting, autocomplete, snippets, go-to-definition, inline error markers, and debounced live preview
  • Interactive React Flow canvas (pan, zoom, minimap) with ELK auto-layout and auto fit-to-view
  • Canvas editing: add/delete/duplicate nodes and draw/replace/delete edges via per-input handles, with undo/redo
  • Validation overlays: per-node error/warning borders and feedback-cycle highlighting
  • Node inspector showing all properties
  • Stateful, streaming simulation: live parameter sliders, "step N more samples", peek-value overlays, oscilloscope waveform and FFT spectrum views, with a configurable sample rate
  • Multi-pass and per-pass graph optimization with before/after comparison
  • C++ code generation preview with copy/download
  • Binary plugin build for multiple platforms (CLAP, VST3, AU, LV2, etc.), with batch build + zip download
  • Content-addressed build cache (avoids recompilation for unchanged graphs)
  • Graphviz DOT and SVG export; round-trip export back to graph JSON and .gdsp

Installation

pip install dsp-graph

Or from source:

git clone https://github.com/shakfu/dsp-graph.git
cd dsp-graph
make install-dev
make frontend-build

Quick Start

dsp-graph serve

Starts the server at http://127.0.0.1:8765. Open that URL in a browser, then load a gen-dsp graph JSON or .gdsp file via the toolbar or by dragging it onto the window. Use --open to auto-open the browser on start.

CLI Options

dsp-graph serve [--host HOST] [--port PORT] [--reload] [--open] [--experimental] [--disable-build]
  • --host: Bind address (default: 127.0.0.1)
  • --port: Port (default: 8765)
  • --reload: Auto-reload on code changes
  • --open: Open browser on start
  • --experimental: Enable experimental features (the gen~/GenExpr transpiler tab and its Max .maxpat test-patch export). Off by default; when omitted the GenExpr tab is hidden and POST /api/genexpr and POST /api/graph/export/maxpat return 404.
  • --disable-build: Disable the native build endpoints (/api/build*), which compile graphs to plugin binaries on the host. Enabled by default; when set the Plugin Target build panel is hidden and every /api/build* route returns 404. Source generation (/api/generate*) is unaffected either way. Use this for a hardened, inspect-only deployment.

Security model

dsp-graph is designed for single-user, localhost use and binds to 127.0.0.1 by default. The build endpoints generate and compile native code on the host, so the server should not be exposed to untrusted networks. For a hardened, inspect-only deployment they can be turned off entirely with --disable-build. As a safeguard against cross-origin requests from other browser tabs, all state-changing (POST/PUT/PATCH/DELETE) /api requests require a per-process session token (GET /api/session), which the bundled UI fetches automatically.

Single-worker deployment. The server runs a single worker, and this is assumed throughout: simulation sessions, the batch-build cache, and the build-cache singleton are all in-process, per-worker state. Blocking work (native builds, simulation, cache disk IO) is offloaded to a worker-thread pool so it does not stall the async event loop, but the shared dictionaries are only ever touched on the event loop, so they need no locking. Running multiple workers (uvicorn --workers N) would break session and batch-download continuity -- a request can land on a worker that does not hold its state -- and requires moving that state to a shared store first.

Development

# Install dev dependencies
make install-dev

# Run backend tests
make test

# Install frontend dependencies
make frontend-install

# Run frontend dev server (hot reload, proxies /api to backend)
make frontend-dev

# Build frontend (outputs to src/dsp_graph/static/)
make frontend-build

# Run frontend tests (Vitest)
make frontend-test

# Run full backend QA (test + lint + typecheck + format)
make qa

# Run full frontend QA (ESLint + tsc + Vitest)
make frontend-qa

# Start server with auto-reload
make dev

CI (.github/workflows/ci.yml) runs the backend lint/format/type-check/tests (Python 3.10 and 3.13) and the frontend tests + build on every push and pull request.

Architecture

Backend (Python)

  • convert.py -- Core Graph <-> ReactFlow conversion layer
  • server.py -- FastAPI app with SPA static serving
  • security.py -- Session-token (anti-CSRF) and body-size middleware
  • cli.py -- CLI entry point (dsp-graph serve)
  • cache.py -- Content-addressed disk cache for build artifacts
  • api/graph.py -- Graph load/validate/export/catalog endpoints
  • api/simulate.py -- Per-sample simulation endpoint
  • api/optimize.py -- Multi-pass optimization endpoint
  • api/compile.py -- C++ code generation endpoint
  • api/genexpr.py -- gen~ codebox (GenExpr) transpile endpoint
  • api/maxpat.py -- Max .maxpat test-patch export endpoint (wraps maxpat.py)
  • api/generate.py -- Project generation (source files, zip download, platform listing)
  • api/build.py -- Binary build, batch build, and build cache management
  • api/layout.py -- Auto-layout endpoint

Frontend (React + TypeScript)

  • React Flow (xyflow) for graph visualization
  • Zustand for state management
  • CodeMirror editor with custom .gdsp language support
  • Custom node types: input, output, param, dsp_node
  • Vite build, outputs to src/dsp_graph/static/

API Endpoints

FastAPI also serves interactive docs at /docs (Swagger UI) and /redoc.

Endpoint Method Description
/api/session GET Per-process session token (required header for mutating requests)
/api/graph/load/json POST Load Graph JSON -> ReactFlow
/api/graph/load/gdsp POST Parse .gdsp source -> ReactFlow
/api/graph/validate POST Validate a Graph (structured errors)
/api/graph/export/json POST ReactFlow -> Graph JSON
/api/graph/export/gdsp POST ReactFlow -> .gdsp source
/api/graph/node-types GET Node type catalog + colors
/api/graph/dot POST Graph -> Graphviz DOT
/api/simulate POST Run per-sample simulation (starts a session)
/api/simulate/continue POST Step N more samples on an existing session
/api/simulate/param POST Set a parameter on a session
/api/simulate/peek POST Read peek-node values from a session
/api/simulate/reset POST Reset session state to initial
/api/simulate/buffer/get POST Read a buffer's contents
/api/simulate/buffer/set POST Write a buffer's contents
/api/optimize POST Optimize graph, all passes (before/after)
/api/optimize/pass POST Apply a single named optimization pass
/api/compile POST Graph -> C++ source
/api/config GET Runtime feature flags (experimental, build_enabled)
/api/genexpr POST Graph -> gen~ codebox (GenExpr) source (experimental; 404 unless --experimental)
/api/graph/export/maxpat POST Graph -> Max .maxpat test patch wrapping the gen~ codebox (experimental; 404 unless --experimental)
/api/layout POST Auto-layout ReactFlow nodes
/api/generate POST Generate project source files
/api/generate/zip POST Generate project as zip download
/api/generate/platforms GET List supported target platforms (host-filtered)
/api/build POST Compile graph to binary plugin
/api/build/binary POST Compile and download binary
/api/build/batch POST Build for multiple platforms
/api/build/batch/{id}/zip GET Download batch build as zip
/api/build/cache DELETE Clear build cache
/api/build/cache/info GET Build cache statistics

All /api/build* routes are enabled by default but return 404 when the server is started with --disable-build.

License

MIT

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

dsp_graph-0.2.1.tar.gz (805.8 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

dsp_graph-0.2.1-py3-none-any.whl (814.7 kB view details)

Uploaded Python 3

File details

Details for the file dsp_graph-0.2.1.tar.gz.

File metadata

  • Download URL: dsp_graph-0.2.1.tar.gz
  • Upload date:
  • Size: 805.8 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.2

File hashes

Hashes for dsp_graph-0.2.1.tar.gz
Algorithm Hash digest
SHA256 cceb3bd2392214837c7166ee728ed70dc9e3415701900ec0a05f2ac8ccefa785
MD5 2cb1dab6af0d72a638a98c3ca6edcca0
BLAKE2b-256 e056dffebaee15fc54d50c2fb8499165cfd3f887d4b6265ebfd11d226d98787d

See more details on using hashes here.

File details

Details for the file dsp_graph-0.2.1-py3-none-any.whl.

File metadata

  • Download URL: dsp_graph-0.2.1-py3-none-any.whl
  • Upload date:
  • Size: 814.7 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.2

File hashes

Hashes for dsp_graph-0.2.1-py3-none-any.whl
Algorithm Hash digest
SHA256 207bda46339c9ac8ed6492b9d1a56227e38044aecca65c70942e430e63fbf406
MD5 e950c0a74cb63c1d23a9ea5171314d40
BLAKE2b-256 b5a8ddeb1cf35abee53f8a8df2568de3ee74c6b9e84d5abfb9870252f98ae27b

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page