RAG

up: LLMs ⚐ Retrieval Augmented Generation. This is best to provide a knowledge base.

• LlamaCloud
• relevance.ai
• Vector databases

Set up RAG

1. Vectorise the source data
   • I think LlamaIndex can do that with the CSV loader
2. Do a similarity search
   • Get the relevant docs
3. Pass docs to LLM and Prompts
   • Prompt example: “Here is a message I received: {{message}}. These are best practices how we would normally respond: {{retrievedDocs}}. Write the best response that I should send.”
4. Execute RAG