SpectraForge · Optional backend services
BACKEND UPGRADEUnlock full AI potential
SPECTRAFORGE works beautifully standalone — no setup needed. Connecting the optional local backend services unlocks the full AI preset pool, session analysis, and real-time musical intelligence. Everything runs on your own machine.
All services are local. No cloud. No account. No data leaves your machine.
01 Backend services
THREE SERVICES
Each service is independent — you can run any combination. Start with Ollama for AI preset selection, add the Conductor for the full 148k library, and optionally wire in the MilkDropLM bridge for AI-generated presets.
port 8765
CONDUCTOR SERVER
The preset library manager. Hosts the full 148k+ preset pool with lazy streaming so the browser never loads everything at once. Handles library seeding, audio preprocessing, BPM caching, and session brief analysis. This is the service that makes the complete preset library navigable.
148k preset pool
Lazy streaming
BPM cache
Session analysis
Python · Flask
port 8766
MILKDROPLM BRIDGE
AI-powered preset generation and recommendation. Connects to your local Ollama instance to generate new .milk presets on demand, run a real-time debate chamber between visual candidates, accept user voting and scoring, and make intelligent recommendations based on musical context.
AI preset generation
Debate chamber
User voting
Smart recommendations
Python · Ollama
OLLAMA INTEGRATION
Local LLM runtime — Mistral, Llama3.2, Qwen, or any compatible model. Provides the musical intelligence layer: reading audio feature vectors, classifying mood and energy, understanding track sections, and dynamically adjusting visual parameters in response to what's playing.
Mistral · Llama3.2 · Qwen
Mood classification
Section awareness
Parameter control
ollama.ai
02 Setup
GETTING STARTED
All services start from a single batch file on Windows. The browser connects automatically once the services are running on their default ports.
> launch.bat
[CONDUCTOR] Starting on port 8765...
[MILKDROPLM] Starting on port 8766...
[OLLAMA] Detected — connecting model...
[OK] All services running. Open SPECTRAFORGE in browser.
Requirements: Python 3.10+, Ollama installed and running a compatible model, Windows 10/11. Mac/Linux paths differ — see README.
