To implement a RAG powered chatbot in jac-lang, different methodologies can be adopted. In this tutorial we are using a separate .jac file including all functionalities required for RAG as methods of an object.
Main Object : rag_engine
-
This object handles information storing and retrieval.
Document Loader :
load_documents- This function will load all source documents (PDF files)
Document Splitter :
split_documents- The PDF documents are broken down into smaller chunks.
Embedding Function :
get_embedding_function- Embeddings are generated for each chunk using an embedding model.
- In this implementation we are using OllamaEmbeddings from langchain using the model nomic-embed-text.
Chunk ID Modifier :
add_chunk_id- This will generate a unique ID for each chunk which will allow better management of the existing data.
- When adding new documents, these IDs can be checked before adding the new document which indicated whether it is a new chunk or not.
- Having the unique ID will also create a direct source link to where the information was extracted.
Adding embeddings for Vector DB :
add_to_chroma- Embeddings are saved in the vector database created using ChromaDB.
- Documents are added to the V-DB only if they are not added to ChromaDB previously.
Context Retrieval :
get_from_chroma- Using the query passed into the query engine, 5 most relevent chunks are retrived from the vector DB which will be forwarded as the context for the LLM to generate a relevent response.
Add a RAG powered response generating feature for the chatbot and complete the full application implementation.
- A RAG engine as described above.
- A full fledged chatbot in any domain (do not need to restrict to medical domain).
- Can have multiple chats.