sujee.dev
RAG
RAG Architecture and Dataflow
Code
Here is the step by step lab that walks you through the RAG process
Components of RAG
Vector database is MongoDB Atlas.
Here is a combination of embedding models and LLMs. We try various LLMs. We access LLMs using API and local.
| Embedding model | LLM | Code |
|---|---|---|
| OpenAI embed (access via API) | OpenAI (access via API) | TODO |
| Open source model (runs locally) | OpenAI (access via API) | rag-10k-b-query-open-embeddings-openAI-llm.ipynb |
| Mistral embed (access via API) | Mistral (access via API) | rag-10k-b-query-mistral-embeddings-mistral-llm.ipynb |
| Open source model (runs locally) | Mistral (access via API) | rag-10k-b-query-open-embeddings-mistral-llm.ipynb |
| Open source model (runs locally) | Mistral-Instruct-7B-v02 (runs locally) | rag-10k-c-query-local-llm-mistral-instruct-1.ipynb |