# Project export: Magentic This document was generated by HackStack to give an AI agent context about a hackathon project. Sections are labeled with their provenance; content marked as truncated was cut to keep this document small. ## Project metadata - Hackathon: TreeHacks 2026 - Tagline: Your AI-powered music production co-pilot for REAPER. - Devpost: https://devpost.com/software/magentic-yqihfw - GitHub: https://github.com/adrian-pan/Magentic - Video: https://www.youtube.com/embed/nGH5mVFKC1U?enablejsapi=1&hl=en_US&rel=0&start=&version=3&wmode=transparent - Team: 5 GitHub contributor(s) — adrian-pan (22 commits), Cursor (16 commits), kaisunwang (15 commits), eddyjiang (9 commits), aaronwniu (2 commits) ## Devpost submission (written by the team) ### Overview Your AI-powered music production co-pilot for REAPER (slightly chopped DAW) ### Inspiration Music production has a steep learning curve, especially in powerful DAWs like REAPER. Beginners often know what they want to hear — something like a “house beat” — but don’t yet know what that means: a four-on-the-floor kick, a clap on 2 and 4, and off-beat hi-hats. And even if they do, they may not know how to build it inside the DAW. That gap between musical instinct and technical execution is where most people get stuck. AI coding assistants have transformed software development — not by replacing engineers, but by helping them move faster and learn as they build. Music production hasn’t had that kind of support. We built Magentic to change that. ### What it does Magentic is a web-based AI agent with direct, real-time control over REAPER. It combines a conversational chat interface with deep DAW integration to help producers: Control REAPER directly — create tracks, add MIDI patterns, apply FX chains, adjust mixing parameters, and more — all from natural-language commands executed instantly in your session AI Stem Separation + MIDI Conversion — Upload any track, split it into drums, bass, vocals, and more using AI tools, then convert those stems to MIDI so you can reuse and reinterpret the original musical ideas directly inside REAPER. Manage FX presets — browse, switch, and apply FX presets on any track through chat Stay project-aware — Magentic reads your live REAPER session (tracks, FX, items, BPM, duration) and gives context-specific responses referencing your actual project state ### How we built it Frontend: React + Vite with a custom dark magenta UI, sleek styling, smooth animations, and an intuitive, split-panel layout (file upload, project details, and main chat panel). Backend: Node.js + Express API server using OpenAI’s GPT-4o and Anthropic’s Claude 3.7 with structured function-calling tools — not just prompt engineering, but a full agentic loop. The AI can call 40+ production tools (create tracks, generate MIDI, apply FX, separate stems, convert audio to MIDI, etc.) and chain them together into multi-step workflows. Python Bridge: A Flask-based bridge server connected to REAPER via the reapy library. Tool calls are translated into live DAW modifications, giving Magentic real-time read/write access for the full session. Stem Separation + Audio-to-MIDI: AI-powered source separation using Demucs (open-source, originally by Meta) splits tracks into drums, bass, vocals, and more. We also convert audio to MIDI using Basic Pitch (open-source by Spotify), allowing users to study and reinterpret musical ideas from their favorite artists directly inside REAPER. Live Editing: Magentic isn’t just offline automation — it works during performance. You can ask the chatbot to modify FX, adjust arrangements, or trigger changes in real time while you’re recording, listening, and producing. ### Challenges we ran into Some of the biggest challenges were technical and architectural. We had to correctly interface with REAPER’s API in a way that was stable, safe, and responsive in real time. Getting GPU inference running reliably required moving from RunPod to Modal (>>>>), where we were finally able to deploy and scale the models properly. Integrating open-source ML models like Demucs and Basic Pitch into the existing agentic tool-calling loop also required restructuring parts of the backend so model outputs could flow cleanly into REAPER actions without breaking the conversational workflow. ### Accomplishments we're proud of (as 1st time hackers) We’re proud of: A working, tool-calling AI agent that can directly modify a live REAPER session — not just suggest ideas, but execute real production actions inside the DAW. Seamless integration of open-source ML models (like Demucs and Basic Pitch) into our agentic architecture, enabling stem separation and audio-to-MIDI inside a unified creative workflow. A fully functional live demo — real-time editing, model inference, and DAW control working together in one system. A creator-first philosophy — empowering bedroom producers to develop and share their ideas, not replacing their creativity, but helping them learn, build confidence, and move faster. ### What we learned We learned that building an AI agent is hard. We also dove deep into REAPER's internals via reapy, learning how DAWs represent tracks, items, takes, and MIDI at the API level. On the music side, building the Educational Engine taught us how much implicit knowledge exists in production — terms like "Backbeat" and "Sidechain Compression" are second nature to pros but completely opaque to beginners. ### What's next Stronger contextual memory — so Magentic can recognize patterns in a user’s style and give more consistent, personalized guidance over time. Surround sound support — tapping into REAPER’s multichannel capabilities to help producers experiment with spatial and immersive audio. Expanded music knowledge — adding more coverage of mixing, arrangement, and sound design concepts so users can learn as they build. Ongoing improvements — refining live editing, model integration, and tools that empower producers without replacing their creative control. ## README (from the GitHub repository)

Magentic Logo

Magentic

AI-powered music production assistant for REAPER


MagenticUI


--- ## Why Magentic? Music production has a steep learning curve, especially in powerful DAWs like REAPER. Beginners often know what they want to hear — something like a "house beat" — but don't yet know what that means: a four-on-the-floor kick, a clap on 2 and 4, and off-beat hi-hats. And even if they do, they may not know how to build it inside the DAW. That gap between musical instinct and technical execution is where most people get stuck. AI coding assistants have transformed software development — not by replacing engineers, but by helping them move faster and learn as they build. Music production hasn't had that kind of support. **We built Magentic to change that.** ### What it does - **Control REAPER directly** — Create tracks, add MIDI patterns, apply FX chains, adjust mixing parameters, and more — all from natural-language commands executed instantly in your session. - **AI Stem Separation + MIDI Conversion** — Upload any track, split it into drums, bass, vocals, and more using **Demucs** (open-source, originally by Meta), then convert those stems to MIDI using **Basic Pitch** (open-source, by Spotify) so you can reuse and reinterpret the original musical ideas directly inside REAPER. - **Manage FX presets** — Browse, switch, and apply FX presets on any track through chat. - **Stay project-aware** — Magentic reads your live REAPER session (tracks, FX, items, BPM, duration) and gives context-specific responses referencing your actual project state. --- ## Architecture & Workflow

Magentic Architecture

**Frontend** — React + Vite with a custom dark magenta UI, sleek styling, smooth animations, and an intuitive split-panel layout (file upload, project details, and main chat panel). **Backend** — Node.js + Express API server using OpenAI's **GPT-4o** and Anthropic's **Claude 3.7** with structured function-calling tools — not just prompt engineering, but a full agentic loop. The AI can call **40+ production tools** (create tracks, generate MIDI, apply FX, separate stems, convert audio to MIDI, etc.) and chain them together into multi-step workflows. **Python Bridge** — A FastAPI-based bridge server connected to REAPER via the **reapy** library. Tool calls are translated into live DAW modifications, giving Magentic real-time read/write access to the full session. **ML Models (via Modal)** — Two open-source ML models are integrated into the workflow and run on GPU via [Modal](https://modal.com): | Model | Source | Purpose | |-------|--------|---------| | **Demucs** (htdemucs) | Meta / Facebook Research | AI-powered source separation — splits any audio track into drums, bass, vocals, and other stems | | **Basic Pitch** (ONNX) | Spotify | Audio-to-MIDI transcription — converts audio stems into MIDI note data so users can study and reinterpret musical ideas | **Reasoning** — An optional reasoning layer using **Qwen 2.5** (served via Modal on A100 GPUs) provides a dedicated planning step for complex, multi-tool production requests before execution. ``` User ──► React Frontend ──► Express Backend ──► Python Bridge ──► REAPER │ │ │ GPT-4o / Claude 3.7 │ (40+ tool schemas) │ │ │ Modal GPU Endpoints │ ├─ Demucs (stem separation) │ ├─ Basic Pitch (audio → MIDI) │ └─ Qwen 2.5 (reasoning) │ │ └────────────────────┘ responses + tool results ``` --- ## Project Structure ``` Magentic/ ├── bridge/ Python FastAPI bridge (REAPER via reapy) — port 5001 │ └── main.py ├── backend/ Express + OpenAI/Anthropic API — port 3001 │ ├── server.js │ ├── agent/ AI system prompt, tools, Anthropic client │ ├── orchestrator/ Intent routing (direct vs. plan) │ ├── musicPlan/ Reasoning model planner client │ ├── routes/ API routes (chat, files, functions) │ └── lib/ Supabase client, file store, music theory ├── planner/ Modal serverless functions │ ├── modal_planner.py Qwen 2.5 reasoning model │ └── modal_functions.py Demucs + Basic Pitch GPU endpoints └── frontend/ React (Vite) — port 5173 └── src/ ``` --- ## Installation ### Prerequisites - [REAPER](https://www.reaper.fm/) installed and open - [reapy](https://github.com/RomeoDespworktrees/reapy) configured (Python REAPER API) - Node.js 18+ - Python 3.10+ - An OpenAI API key (required) - An Anthropic API key (optional, used as fallback when OpenAI rate-limits) - A Supabase project with a `magentic-files` storage bucket (for persistent file/stem/MIDI storage) ### 1. Bridge (REAPER control) The bridge is a FastAPI server that connects to REAPER via reapy. It must be running for any DAW interaction to work. ```bash cd bridge pip install fastapi uvicorn python-reapy pydantic python main.py ``` > The bridge starts on port **5001**. Make sure REAPER is open before starting it. If you see `reapy` connection errors, run `python -c "import reapy; reapy.configure_reaper()"` first to enable the REAPER API. ### 2. Backend ```bash cd backend cp .env.example .env npm install npm run dev ``` Edit `.env` with your keys: ```env OPENAI_API_KEY=sk-your-key ANTHROPIC_API_KEY=sk-ant-your-key # optional — Claude fallback SUPABASE_URL=https://your-project.supabase.co SUPABASE_SERVICE_KEY=your-service-role-key # ML Functions (see Limitations below) FUNCTIONS_PROVIDER=modal FUNCTIONS_MODAL_URL=https://your-modal-url.modal.run # Reasoning model REASONING_PROVIDER=openai # or 'modal' for Qwen 2.5 ``` ### 3. Frontend ```bash cd frontend npm install npm run dev ``` ### Start order ``` 1. Open REAPER 2. bridge/ → python main.py 3. backend/ → npm run dev 4. frontend/ → npm run dev ``` Open **http://localhost:5173** — chatbot on the right, import panel and project state on the left. --- ## Current Limitations 1. **Reasoning model** — The Qwen 2.5 reasoning model (originally served via Modal) does not yet produce reliable enough plans for complex multi-step production tasks. In most cases, the system defaults to gpt-4o for both direct execution and planning. 2. **Modal GPU endpoints disabled** — As of February 16, 2026, the Modal GPU endpoints for Demucs (stem separation), Basic Pitch (audio-to-MIDI), and QWEN (planning) have been disabled. Users will **not** be able to run stem separation or audio-to-MIDI conversion until the endpoints are re-deployed. To re-enable, deploy the Modal functions: ```bash cd planner modal deploy modal_functions.py ``` Then set `FUNCTIONS_MODAL_URL` in `backend/.env` to the deployed URL. --- ## Detected evidence (automated analysis) Indexed codebase: 2501 recognized source files, 27684 KB. - Anthropic (technology) — detected in the code - CSS (language) — detected in the code - Express (technology) — detected in the code - FastAPI (technology) — detected in the code - HTML (language) — detected in the code - JavaScript (language) — detected in the code - OpenAI (technology) — detected in the code - Python (language) — detected in the code - PyTorch (technology) — detected in the code - React (technology) — detected in the code - Supabase (technology) — detected in the code - Node.js (technology) — claimed on Devpost, not found in the code - AI coding agent: Claude Code — evidence: commit authorship or trailers - AI coding agent: Cursor — evidence: commit authorship or trailers ## Codebase structure (from repository index) ### Files (120 of 2725) ``` .gitignore agents/__init__.py agents/midi_agent.py agents/mix_agent.py agents/orchestrator.py agents/sound_agent.py agents/tools.py backend/.env.example backend/.gitignore backend/agent/anthropicClient.js backend/agent/systemPrompt.js backend/agent/tools.js backend/backend.log backend/frontend.log backend/functions/__init__.py backend/functions/.dockerignore backend/functions/api.py backend/functions/build.sh backend/functions/DOCKER_README.md backend/functions/Dockerfile backend/functions/handler.py backend/functions/README.md backend/functions/requirements-runpod.txt backend/functions/requirements.txt backend/functions/run_cli.py backend/functions/setup_repos.sh backend/lib/fileStore.js backend/lib/musicTheory.js backend/lib/supabase.js backend/musicPlan/benchmark.js backend/musicPlan/compileNotes.js backend/musicPlan/evaluateNeeds.js backend/musicPlan/executor.js backend/musicPlan/index.js backend/musicPlan/normalizeProjectState.js backend/musicPlan/plannerClient.js backend/musicPlan/planWithRepair.js backend/musicPlan/schema.json backend/musicPlan/validatePlan.js backend/musicPlan/validateSchema.js backend/orchestrator/directExecute.js backend/orchestrator/planExecute.js backend/orchestrator/projectCache.js backend/orchestrator/routeIntent.js backend/package.json backend/routes/chat.js backend/routes/execute.js backend/routes/files.js backend/routes/functions.js backend/scripts/extract-stems.js backend/scripts/test-stem-chat.js backend/server.js backend/services/orchestrator.js bridge/apply_variations.py bridge/bridge.log bridge/check_hidden_envelope.py bridge/cleanup_items.py bridge/configure_reaper.py bridge/debug_midi_insert.py bridge/debug_ppq.py bridge/debug_reapy_stdout.py bridge/debug_reapy.py bridge/debug_server.py bridge/debug_silence.py bridge/fix_eddy_envelope.py bridge/fix_relative_fade.py bridge/force_fade_final.py bridge/inspect_envelopes.py bridge/main.py bridge/melody_notes.json bridge/reaper_server.py bridge/run_request.py bridge/test_copy_paste.py bridge/test_envelope_creation.py bridge/test_force_show.py bridge/test_nudge.py bridge/test_remove_envelope.py bridge/verify_extension.py bridge/verify_volume_fade.py frontend/.gitignore frontend/eslint.config.js frontend/index.html frontend/package.json frontend/README.md frontend/src/App.css frontend/src/App.jsx frontend/src/components/ChatPanel.jsx frontend/src/components/ImportPanel.jsx frontend/src/components/Logo.jsx frontend/src/components/WelcomePage.jsx frontend/src/index.css frontend/src/main.jsx frontend/vite.config.js package.json planner/.gitignore planner/.runpod/resources.pkl planner/#/bin/activate planner/#/bin/activate.csh planner/#/bin/activate.fish planner/#/bin/Activate.ps1 planner/#/bin/pip planner/#/bin/pip3 planner/#/bin/pip3.12 planner/#/bin/python planner/#/bin/python3 planner/#/bin/python3.12 planner/#/lib/python3.12/site-packages/pip-25.0.1.dist-info/AUTHORS.txt planner/#/lib/python3.12/site-packages/pip-25.0.1.dist-info/entry_points.txt planner/#/lib/python3.12/site-packages/pip-25.0.1.dist-info/INSTALLER planner/#/lib/python3.12/site-packages/pip-25.0.1.dist-info/LICENSE.txt planner/#/lib/python3.12/site-packages/pip-25.0.1.dist-info/METADATA planner/#/lib/python3.12/site-packages/pip-25.0.1.dist-info/RECORD planner/#/lib/python3.12/site-packages/pip-25.0.1.dist-info/REQUESTED planner/#/lib/python3.12/site-packages/pip-25.0.1.dist-info/top_level.txt planner/#/lib/python3.12/site-packages/pip-25.0.1.dist-info/WHEEL planner/#/lib/python3.12/site-packages/pip/__init__.py planner/#/lib/python3.12/site-packages/pip/__main__.py planner/#/lib/python3.12/site-packages/pip/__pip-runner__.py planner/#/lib/python3.12/site-packages/pip/_internal/__init__.py planner/#/lib/python3.12/site-packages/pip/_internal/build_env.py [2605 more files omitted for size] ``` ### Dependencies - backend/functions/requirements.txt: dora-search, einops, julius@>=0.2.3, lameenc@>=1.2, librosa@>=0.8.0, mir_eval@>=0.6.0, numpy@>=1.18, onnxruntime@>=1.10.0, openunmix, pretty_midi@>=0.2.9, pyyaml, resampy@>=0.2.2,<0.4.3, scikit-learn, scipy@>=1.4.1, torch@>=1.8.1, torchaudio@>=0.8, tqdm, typing_extensions - backend/package.json: @anthropic-ai/sdk@^0.74.0, @supabase/supabase-js@^2.95.3, ajv@^8.18.0, ajv-formats@^3.0.1, cors@^2.8.5, dotenv@^16.4.7, express@^4.21.2, multer@^1.4.5-lts.1, openai@^4.79.0 - frontend/package.json: @eslint/js@^9.39.1, @types/react@^19.2.7, @types/react-dom@^19.2.3, @vitejs/plugin-react@^5.1.1, eslint@^9.39.1, eslint-plugin-react-hooks@^7.0.1, eslint-plugin-react-refresh@^0.4.24, globals@^16.5.0, react@^19.2.0, react-dom@^19.2.0, react-router-dom@^7.13.0, vite@^7.3.1 - requirements.txt: fastapi, openai, python-dotenv, python-reapy, requests, uvicorn ### Recent commits (newest first) - Update README.md - Update README.md - Add files via upload - changed ui name - Update README.md - added ui - Update README.md - style: improve README header spacing and UI demo sizing - Update README.md - added ui ss - fix: Supabase URL routing, Modal float16 crash, and updated README - Merge branch 'main' of https://github.com/adrian-pan/Magentic - feat: migrate ML functions to Modal, add timeout fallback + vocal MIDI fix - fix(agent): harden add_fx plugin resolution - musicTheory update - Stop button! - feat(agent): support moving media items by beat offsets - Update package-lock.json - Merge main into reasoning — combine routing architecture with all main tools - feat: add deterministic intent router, split direct/plan execution paths ## Key source files (fetched from GitHub, selected and truncated for size) ### backend/functions/DOCKER_README.md ```markdown # Docker / RunPod Deployment Package for deploying stem separation and audio-to-MIDI to RunPod serverless GPU. ## Build (from backend/functions/) ```bash ./build.sh # or docker build --platform linux/amd64 -t magentic-ml:latest . ``` ## Push to Docker Hub ```bash docker tag magentic-ml:latest YOUR_USERNAME/magentic-ml:latest docker push YOUR_USERNAME/magentic-ml:latest ``` ## RunPod Input/Output **Input:** ```json { "action": "separate_stems", "input_url": "https://example.com/track.mp3" } ``` or ```json { "action": "transcribe_to_midi", "input_url": "https://example.com/bass.mp3" } ``` **Output (separate_stems):** `{ "stems": { "drums": "", "bass": "", ... } }` **Output (transcribe_to_midi):** `{ "midi_base64": "", "filename": "..." }` The backend decodes base64 and uploads to Supabase when calling RunPod. ## Deploy on RunPod 1. Push image to Docker Hub 2. RunPod Console → Serverless → New Endpoint 3. Import from Docker Registry → enter image URL 4. Select GPU (e.g. RTX 4090) 5. Set handler timeout (stem separation: 1–5 min per track) ``` ### requirements.txt ``` fastapi uvicorn python-reapy requests openai python-dotenv ``` ### package.json ``` { "name": "magentic", "private": true, "scripts": { "benchmark": "cd backend && npm run benchmark" } } ``` ### backend/package.json ``` { "name": "magentic-backend", "version": "1.0.0", "description": "Magentic AI Agent Backend", "main": "server.js", "scripts": { "start": "node server.js", "dev": "node --watch server.js", "benchmark": "node musicPlan/benchmark.js" }, "dependencies": { "@anthropic-ai/sdk": "^0.74.0", "@supabase/supabase-js": "^2.95.3", "ajv": "^8.18.0", "ajv-formats": "^3.0.1", "cors": "^2.8.5", "dotenv": "^16.4.7", "express": "^4.21.2", "multer": "^1.4.5-lts.1", "openai": "^4.79.0" } } ``` ### frontend/package.json ``` { "name": "frontend", "private": true, "version": "0.0.0", "type": "module", "scripts": { "dev": "vite", "build": "vite build", "lint": "eslint .", "preview": "vite preview" }, "dependencies": { "react": "^19.2.0", "react-dom": "^19.2.0", "react-router-dom": "^7.13.0" }, "devDependencies": { "@eslint/js": "^9.39.1", "@types/react": "^19.2.7", "@types/react-dom": "^19.2.3", "@vitejs/plugin-react": "^5.1.1", "eslint": "^9.39.1", "eslint-plugin-react-hooks": "^7.0.1", "eslint-plugin-react-refresh": "^0.4.24", "globals": "^16.5.0", "vite": "^7.3.1" } } ``` ### backend/functions/requirements.txt ``` # =========================================== # Demucs - Stem separation # =========================================== dora-search einops julius>=0.2.3 lameenc>=1.2 openunmix pyyaml torch>=1.8.1 torchaudio>=0.8 tqdm # =========================================== # Basic Pitch - Audio-to-MIDI # =========================================== librosa>=0.8.0 mir_eval>=0.6.0 numpy>=1.18 pretty_midi>=0.2.9 resampy>=0.2.2,<0.4.3 scikit-learn scipy>=1.4.1 typing_extensions # Model runtimes (install at least one for basic-pitch) # macOS: coremltools or onnxruntime # Linux/Windows: onnxruntime onnxruntime>=1.10.0 # coremltools; platform_system == 'Darwin' ``` ### backend/functions/Dockerfile ``` # Magentic ML functions: stem separation (Demucs) + audio-to-MIDI (Basic Pitch) # Target: linux/amd64 for RunPod serverless GPU # Build: docker build --platform linux/amd64 -t magentic-ml:latest . FROM nvidia/cuda:12.1.0-cudnn8-runtime-ubuntu22.04 ENV DEBIAN_FRONTEND=noninteractive ENV PYTHONUNBUFFERED=1 # Install Python 3.10 (PyTorch bundles its own CUDA libs via pip, no system CUDA libs needed) RUN apt-get update && apt-get install -y --no-install-recommends \ python3.10 python3.10-venv python3-pip \ ffmpeg \ && rm -rf /var/lib/apt/lists/* RUN ln -sf /usr/bin/python3.10 /usr/bin/python WORKDIR /app # Install PyTorch with CUDA (for Demucs) RUN pip install --no-cache-dir torch torchaudio --index-url https://download.pytorch.org/whl/cu121 # Install Demucs deps (excluding torch/torchaudio which are already installed above) # Demucs pins torchaudio<2.1 which would downgrade our cu121 install, so we use --no-deps COPY demucs/ /app/demucs/ RUN pip install --no-cache-dir dora-search einops "julius>=0.2.3" "lameenc>=1.2" openunmix pyyaml tqdm RUN pip install --no-cache-dir --no-deps -e /app/demucs/ # Pre-download the htdemucs model so it's baked into the image RUN python -c "from demucs.pretrained import get_model; get_model('htdemucs')" # Pin NumPy <2 to stay compatible with PyTorch/torchaudio compiled against NumPy 1.x RUN pip install --no-cache-dir "numpy<2" # Install Basic Pitch with ONNX only (no TensorFlow) COPY basic-pitch/ /app/basic-pitch/ RUN pip install --no-cache-dir onnxruntime librosa mir_eval pretty_midi "resampy>=0.2.2,<0.4.3" scikit-learn scipy typing_extensions RUN pip install --no-cache-dir --no-deps -e /app/basic-pitch/ # Install RunPod handler deps COPY requirements-runpod.txt /app/requirements-runpod.txt RUN pip install --no-cache-dir -r /app/requirements-runpod.txt # Copy API and handler COPY api.py /app/api.py COPY handler.py /app/handler.py CMD ["python", "-u", "/app/handler.py"] ``` ### backend/server.js ```javascript require('dotenv').config(); const express = require('express'); const cors = require('cors'); const path = require('path'); const chatRoutes = require('./routes/chat'); const fileRoutes = require('./routes/files'); const executeRoutes = require('./routes/execute'); const functionRoutes = require('./routes/functions'); const app = express(); const PORT = process.env.PORT || 3001; // Middleware app.use(cors()); app.use(express.json({ limit: '10mb' })); // API Routes app.use('/api/chat', chatRoutes); app.use('/api/files', fileRoutes); app.use('/api/functions', functionRoutes); app.use('/api/execute', executeRoutes); app.use('/api/reaper', executeRoutes); // Health check app.get('/api/health', (req, res) => { res.json({ status: 'ok', timestamp: new Date().toISOString() }); }); const BRIDGE_URL = process.env.BRIDGE_URL || 'http://localhost:5001'; // Start server app.listen(PORT, async () => { console.log(`\n🧲 Magentic Backend running on http://localhost:${PORT}`); console.log(` API endpoints:`); console.log(` POST /api/chat — Chat with the AI agent`); console.log(` POST /api/files/upload — Upload file (Supabase)`); console.log(` GET /api/files — List uploaded files`); console.log(` POST /api/functions/separate-stems — Stem separation`); console.log(` POST /api/functions/transcribe-to-midi — Audio to MIDI`); console.log(` POST /api/execute — Execute code in REAPER`); console.log(` GET /api/reaper/status — REAPER connection status`); console.log(` GET /api/reaper/analyze — Analyze REAPER project`); console.log(` GET /api/reaper/analyze/instruments — List installed instruments`); console.log(` GET /api/health — Health check\n`); if (!process.env.OPENAI_API_KEY || process.env.OPENAI_API_KEY.startsWith('sk-your')) { console.log(` ⚠️ No valid OPENAI_API_KEY found. Create backend/.env with your key.\n`); } try { const bridgeRes = await fetch(`${BRIDGE_URL}/status`); const bridgeData = await bridgeRes.json(); if (bridgeData.reaper_connected) { console.log(` ✓ Bridge at ${BRIDGE_URL} — REAPER connected`); } else { console.log(` ⚠ Bridge at ${BRIDGE_URL} — REAPER not connected. Start bridge: cd bridge && python main.py`); } } catch { console.log(` ⚠ Bridge at ${BRIDGE_URL} unreachable. Start it with: cd bridge && python main.py`); } try { const { checkPlannerHealth } = require('./musicPlan/plannerClient'); const reasoningHealth = await checkPlannerHealth(); if (reasoningHealth.ok) { console.log(` ✓ Reasoning provider (${reasoningHealth.provider}) — ready`); } else { console.log(` ⚠ Reasoning provider (${reasoningHealth.provider}) — ${reasoningHealth.error || 'unreachable'}`); } } catch (e) { console.log(` ⚠ Reasoning provider health check failed: ${e.message}`); } console.log(''); }); ``` ### frontend/src/main.jsx ```javascript import { StrictMode } from 'react' import { createRoot } from 'react-dom/client' import { BrowserRouter } from 'react-router-dom' import './index.css' import App from './App.jsx' createRoot(document.getElementById('root')).render( , ) ``` ### backend/musicPlan/index.js ```javascript /** * musicPlan — Planner → Validator → Needs Gate → Executor pipeline. * Modular, vendor-agnostic, no Python orchestrator dependency. */ const { validateSchema } = require('./validateSchema'); const { validatePlan } = require('./validatePlan'); const { normalizeProjectState } = require('./normalizeProjectState'); const { evaluateNeeds } = require('./evaluateNeeds'); const { planMusic } = require('./plannerClient'); const { planWithRepair } = require('./planWithRepair'); const { executePlan } = require('./executor'); const { compileClip, compileEvent, parseChordSymbol } = require('./compileNotes'); module.exports = { validateSchema, validatePlan, normalizeProjectState, evaluateNeeds, planMusic, planWithRepair, executePlan, compileClip, compileEvent, parseChordSymbol, }; ``` [2341 more indexed source files omitted to keep this export small. The full file list is in the Codebase structure section above.]