RAG Pipeline Builder

使用场景

• Building production RAG over internal documentation
• Choosing chunking and embedding strategies for a corpus
• Adding hybrid search and re-ranking to improve recall
• Designing citation-grounded answer prompts
• Creating retrieval and faithfulness evaluation suites

Design a RAG pipeline for a customer support knowledge base of ~50,000 markdown articles.

Requirements:
- Sub-second retrieval latency
- Answers must cite source articles
- Multilingual content (English, Spanish, German)

Provide:
1. Chunking strategy with rationale
2. Embedding model recommendation and vector store choice
3. Hybrid retrieval + re-ranking design
4. Prompt template enforcing grounded citations
5. Evaluation plan (retrieval recall, answer faithfulness)
6. Example implementation code

推荐模型

兼容工具

模态

相关 Skills