LunaGuide : Instructional AI Co-Pilot LunaGuide is a Multimodal AI Orchestrator designed for "hands-free, eyes-up" safety guidance in complex environments. By combining Computer Vision and Retrieval-Augmented Generation (RAG), it allows field engineers to identify hardware and retrieve critical NASA safety protocols using only a camera and their voice.

🛠 The Tech Stack Orchestration: LangGraph (Stateful multi-agent workflows) Vision Model: Llama 3.2-Vision (Running locally via Ollama) Vector Database: Pinecone (Cloud-scale semantic search) API Framework: FastAPI (Asynchronous web service) Speech Synthesis: gTTS (Real-time audio generation) Embeddings: nomic-embed-text

🧠 System Architecture The core logic is built as a 3-node state machine using LangGraph to ensure reliable, reasoning-based outputs: Vision Node: Analyzes the hardware image to identify components (e.g., Raspberry Pi 4, Compute Module 4). RAG Node: Queries a Pinecone index containing NASA technical manuals to find high-relevance safety protocols. Synthesis Node: Merges visual data with retrieved technical facts to generate a concise safety instruction.

🚀 Key Features Multimodal API: Supports both JSON text responses and direct MP3 audio streaming. Hands-Free Safety: Specifically designed for emergency protocol retrieval where the user cannot stop to read a manual. Local-First Intelligence: Uses Ollama to process images locally, ensuring low latency and data privacy.

🏁 Technical Checkpoints Logic Validation (graph.py) Vision Success: Accurately identifies hardware versions and states. Retrieval Success: Successfully matches hardware to specific NASA space-grade thermal and connection protocols. Reasoning Success: Warns users about specific hardware pitfalls (e.g., interposer requirements).

API Deployment (main.py) Production Handshake: Live orchestration between local LLMs and cloud Vector DBs. Voice Integration: Successful real-time generation of instruction.mp3 for audible guidance.

📦 Installation & Setup Clone the repo: git clone https://github.com/chanz1996/LunaGuide-Copilot-.git Install dependencies: pip install -r requirements.txt Run the API: fastapi dev main.py