PromptLooper

The one who loops prompts — a universal LLM pipeline tuning workbench.

PromptLooper is a self-hosted tool for systematically optimizing LLM prompts, model selection, and inference parameters. It runs experiments across prompt x model x parameter combinations, caches every response, scores results against pluggable evaluation functions, and surfaces the best configurations through a real-time observability dashboard with human-in-the-loop steering.

It ships as a single Docker container (SQLite mode) for zero-config quickstart, or a Docker Compose stack (Postgres + Redis) for production use. An MCP server enables any AI agent to drive PromptLooper programmatically — creating experiments, running sweeps, and reading results without human intervention.

Quick Start

Single Container (zero dependencies)

docker run -p 8000:8000 -v promptlooper-data:/data ghcr.io/xpltdco/promptlooper

Open http://localhost:8000 — you'll be prompted to create an admin account on first boot.

In single-container mode, the API serves the built frontend as static files at the root. Database migrations run automatically on startup.

Production (Docker Compose)

git clone git@git.xpltd.co:xpltdco/promptlooper.git
cd promptlooper
cp .env.example .env
# Edit .env — set JWT_SECRET at minimum
docker compose up -d

Open http://localhost:8400 — nginx proxies the frontend (port 80 → 8400) and API (/api/ → port 8000).

Services started:

• promptlooper-db — PostgreSQL 16 on port 5434
• promptlooper-redis — Redis 7
• promptlooper-api — FastAPI + Alembic migrations (auto-runs on startup)
• promptlooper-worker — Celery worker for experiment execution
• promptlooper-web — Nginx reverse proxy on port 8400

First boot: Navigate to http://localhost:8400/setup to create the admin account.

Features

• Systematic experimentation — grid, random, and guided sweeps across prompt x model x parameter space
• Response caching — SHA-256 deduplication means re-runs cost zero tokens
• Pluggable scoring — embedding similarity, format compliance, keyword presence, LLM-as-judge, human rating, custom webhooks
• Real-time dashboard — live progress, leaderboard, side-by-side comparison, steering controls
• MCP server — AI agents can create experiments, run sweeps, and export results programmatically
• Single-container mode — SQLite + in-process queue when no external dependencies are configured

Development

# Start backing services
docker compose up -d promptlooper-db promptlooper-redis

# Backend
cd backend && pip install -r requirements.txt
alembic upgrade head
uvicorn main:app --reload --host 0.0.0.0 --port 8000

# Frontend (separate terminal)
cd frontend && npm install && npm run dev

Testing

cd backend && pytest
cd frontend && npm test

License

AGPL-3.0