What it is Runs locally, lets you experiment with: • LLM providers (Ollama is the default, but you can switch to OpenAI, Anthropic, Microsoft, Google, and more.). • RAG workflows: upload documents, chunk, embed, search with scoring, and filter metadata. • A visual MCP (Model Context Protocol, a spec for connecting models to external tools) Playground to set up tool integrations (HTTP, STDIO, SSE), inspect tool metadata, and call them from a chat interface.

I wanted a place to play with RAG workflows and external tool calls without wiring up a full app or handling API boilerplate.

GitHub: https://github.com/JM-Lab/spring-ai-playground

Why it’s different • Built with Spring AI—no new language or framework if you’re already in the Spring ecosystem. • Zero API key setup (thanks to Ollama support), though you can switch to OpenAI easily. • Vector DB agnostic—plug in providers like Pinecone, Milvus, PGVector, Weaviate, Elasticsearch, Redis, and many more. • Live MCP debugging—inspect tools, tweak arguments, see execution history in one place. • Everything runs locally—your data stays on your machine (Docker or native). • Built with Spring Boot DevTools—fast application restarts when you modify configurations or experiment with different settings.

Background I built this after repeating the same setup tasks for Spring AI. I wanted a simple sandbox to prototype ideas faster. Currently lacks authentication, user management, and production-ready features, but it already cuts my prototyping time in half.

Feedback welcome If you try it, I’d love to know: • Was anything confusing or hard to use? • Any rough edges or gotchas you’d like fixed?

Jemin Huh