Acceleracers RAG
Jun 2026PersonalJun 202614 technologies12 features
About this project
A RAG-powered chatbot for the Hot Wheels Acceleracers universe. Ask questions about characters, realms, accelechargers, drivers, and teams.
The system scrapes the Acceleracers Fandom Wiki, indexes it into a local vector store using Ollama and Vectra, and answers with context retrieved from the corpus plus an LLM.
Built with Clean Architecture in TypeScript — Hono server, React client, hybrid search (dense vectors + BM25) with MMR reranking, SSE streaming, and Docker support.
Features
Hybrid search (dense vectors + BM25) with MMR diversity reranking
SSE streaming responses with token-by-token output
Chat history sidebar with localStorage persistence
Conversation context across turns
Markdown rendering for responses
Guardrails (rate limiting, max length validation)
Connection status indicator with periodic health checks
Toast notifications for transient errors
Inline error bubbles for server/network failures
Source preview popover on hover
CLI REPL mode for terminal interaction
Docker Compose setup for one-command deployment
Tech Stack
TypeScriptReactHonoTailwind CSSViteOllamaVectraMiniSearchCheerioreact-markdownlucide-reactTanStack QueryVercel AI SDKDocker