Full-Stack•2025
GunnerGPT
Retrieval-Augmented Generation system that answers questions about Arsenal's modern era using a curated knowledge base.
PROBLEM
Large language models hallucinate when answering niche domain questions.
APPROACH
Implemented a RAG pipeline that retrieves relevant documents before generating responses.
ARCHITECTURE
User query
│
▼
Embedding generation
│
▼
Vector search (ChromaDB)
│
▼
Context retrieval
│
▼
LLM response (Gemini)TECHNICAL DECISIONS
ChromaDB
Lightweight vector storage for fast semantic search.
- ‣Efficient embedding indexing
- ‣Low latency retrieval
LangChain
Orchestrates the RAG pipeline segments.
- ‣Modular component management
- ‣Easy context injection
FastAPI
Async inference API for real-time interaction.
- ‣Handles concurrent LLM calls efficiently
OUTCOME
Ensures accurate, grounded responses for Arsenal-specific inquiries.
STACK
Frontend
Next.jsBackend
FastAPIAI/Intelligence
LangChainChromaDB
Google GeminiUP NEXT
Mindplex.ai
Infrastructure and backend development for a decentralized, AI-focused social platform.
→