## Conclusion

### What You Built

- A text RAG pipeline with vector similarity search
- An image search system using multimodal embeddings
- A hybrid search system combining vector + BM25 retrieval
- A GraphRAG-powered Q&A system

### Key Takeaways

- TiDB Cloud Starter serves as a unified storage layer for vectors, text, and graph data
- `pytidb` simplifies embedding storage and retrieval with a few lines of code
- Hybrid search improves recall by combining multiple retrieval channels

### Next Steps

- [Build an Agentic Branching Workflow](/labs/demo_411) — full GraphRAG pipeline
- [TiDB Vector Search documentation](https://docs.pingcap.com/ai/vector-search-overview/)
- [pytidb GitHub repo](https://github.com/pingcap/pytidb) — explore the SDK further