Kenhuang Academy
  • English

      • Resources

      Vector-store and embedding APIs still move. Prefer each project’s official docs. This course only pins the principles: chunk, embed, store, retrieve / rerank, evaluate.

      Official

      ResourceURL
      Chroma docshttps://docs.trychroma.com
      Chroma: Clientshttps://docs.trychroma.com/docs/run-chroma/clients
      Chroma: Add datahttps://docs.trychroma.com/docs/collections/add-data
      pgvector (extension)https://github.com/pgvector/pgvector
      pgvector-pythonhttps://github.com/pgvector/pgvector-python
      Qdrant docshttps://qdrant.tech/documentation/
      Qdrant Quickstarthttps://qdrant.tech/documentation/quickstart/
      OpenAI Embeddingshttps://platform.openai.com/docs/guides/embeddings
      Ollama Embeddingshttps://docs.ollama.com/capabilities/embeddings
      BGE / BAAI (Hub)https://huggingface.co/BAAI

      Field names in chromadb, psycopg + pgvector, and qdrant-client follow your installed minor version. This tutorial tracks the mainstream APIs on those pages (PersistentClient, <=>, upsert / query_points).

      Do not append .html / .md to these paths:

      CourseLinkWhen to open it
      This course (principles + stores)/en/rag/Chunking, embeddings, Chroma / pgvector / Qdrant, eval
      LangChain RAG/en/langchain/ragLoaders, retrievers, agentic RAG
      LlamaIndex/en/llamaindex/Index- and query-engine-centric framework
      Ollama Embeddings/en/ollama/embeddingsLocal /api/embed, bge-m3
      Hugging Face/en/huggingface/BAAI model cards, sentence-transformers
      Dify Knowledge Base/en/dify/knowledge-baseConsole chunking, hybrid search, rerank
      PostgreSQL/en/postgresql/SQL, tables, indexes—the floor under pgvector
      Docker/en/docker/Running the qdrant/qdrant image

      Chinese pages use the same paths without /en.

      A useful map: this course explains where chunks go; LangChain / LlamaIndex / Dify wrap the same pipeline; Ollama / Hugging Face supply embedding models; Postgres hosts pgvector.

      Tutorial index

      1. RAG tutorial
      2. Introduction
      3. Pipeline
      4. Chunking
      5. Embeddings
      6. Chroma
      7. pgvector
      8. Qdrant
      9. Retrieval
      10. Evaluation
      11. Practical Examples

      Suggested exercises

      1. Write 3 Markdown files with invented clause IDs, ask 10 questions via Chroma, compute hit@4
      2. Embed the same corpus with text-embedding-3-small and Ollama bge-m3—never mix vectors
      3. ALTER TABLE ... ADD COLUMN embedding on an existing Postgres table; filter by tenant
      4. Add a Qdrant dept payload and confirm engineering docs never appear in HR questions
      5. Run the Evaluation chapter’s token scan on model answers; note paraphrases (“a week” vs “7 days”)

      Contribute

      This tutorial lives in Kenhuang Academy. When APIs change, open an issue or PR and update both docs/zh/rag/ and docs/en/rag/.

      Next steps

      评论