🌌 LaunchConstellation Suite

LaunchConstellation is a full-stack launch autopilot: a production workflow orchestrator plus every companion system needed to plan, QA, translate, narrate, and visualize a product launch. This suite repo makes the entire ecosystem available from one clone so the AI frontier (Claude, MCP, LangGraph, DeepSeek, Diffusers, RAG, TTS) can plug in instantly.

🚀 Modules at a glance

Directory	Frontier-friendly Focus
orchestrator/	Multi-agent workflow engine (TypeScript, pnpm, BullMQ, Redis Streams, Prisma, pgvector) driving RootAgent DAGs, guardrails, approvals, launch DAG execution.
mission-control/	MCP Ground Crew server exposing telemetry to Claude/Cursor and any Model Context Protocol client. Real-time intents/tasks via Redis → MCP resources.
reasoning-copilot/	DeepSeek-R1 guardrail brain. FastAPI service that classifies artifacts (pass / needs_revision / block) with reasoning traces you can drop into guard pipelines.
qa-observatory/	Eval & safety lab with Giskard test suites, TruLens feedback, and Ragas translation checks; emits HTML/Markdown reports for marketing/legal copy.
creative-bay/	Diffusers-powered creative studio turning artifact copy into FLUX/SDXL prompts and marketing hero art, feature cards, social teasers.
mission-autopilot/	Auto-playbooks + Memory Forge + Radio Mode. Scripted launch simulations, sentence-transformer knowledge base, pyttsx3 audio briefings.

Each project preserves its own README, .env.example, and install scripts. Start anywhere, or wire them together for an end-to-end launch autopilot.

🔗 Quickstart cheatsheet

# 1) clone once
# git clone https://github.com/joshuaworth/launch-constellation-suite.git

# orchestrator – multi-agent DAG engine
cd orchestrator
pnpm install
pnpm dev

# mission control – MCP server for Claude/Cursor
cd ../mission-control
npm install
npm run dev

# reasoning copilot – DeepSeek guardrails
cd ../reasoning-copilot
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
uvicorn copilot.main:app --reload --port 9000

# QA observatory – eval dashboards
cd ../qa-observatory
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
python scripts/run_observatory.py --config configs/sample.yaml

# Creative Bay – prompt → image
cd ../creative-bay
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
python scripts/generate.py --artifact samples/marketing_copy.json --dry-run

# Mission Autopilot – playbooks, memory, radio briefings
cd ../mission-autopilot
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
python scripts/autopilot.py --help

🔥 Why the frontier cares

• Claude-ready via MCP – Mission Control lets Claude desktop or Cursor ask for live telemetry (launch-constellation://summary, …/approvals, etc.).
• Reasoning-native guardrails – DeepSeek-R1 (or any open reasoning LLM) scores artifacts with JSON verdicts for orchestrator guardrails.
• Evaluation stack – Giskard, TruLens, Ragas, LangChain-ready output. Plug in OpenAI/Anthropic or keep it all OSS.
• Launch art from copy – Creative Bay feeds FLUX/Stable Diffusion/SDXL to produce assets in seconds.
• Knowledge autopilot – Memory Forge builds sentence-transformer embeddings + summaries for reusable launches.
• Audio briefings – Radio Mode narrates launch status for leadership stand-ups or podcasts.

🧭 Repository layout

launch-constellation-suite/
 ├─ orchestrator/
 ├─ mission-control/
 ├─ reasoning-copilot/
 ├─ qa-observatory/
 ├─ creative-bay/
 └─ mission-autopilot/

MIT © LaunchConstellation contributors — contributions, forks, and frontier experiments welcome.