RAG enabled Chatbots using LangChain and Databutton
- For the front-end :
app.py - PDF parsing and indexing :
brain.py - API keys are maintained over databutton secret management
- Indexed are stored over session state
Oversimplified explanation : (Retrieval) Fetch the top N similar contexts via similarity search from the indexed PDF files -> concatanate those to the prompt (Prompt Augumentation) -> Pass it to the LLM -> which further generates response (Generation) like any LLM does. More in the blog!
Blog Post - Here
YouTube video - Here
To get started quickly, you can use the "Chat with PDF" template within Databutton π
Alternatively, you can use Streamlit to build and deploy! In that case few changes such as
st.secretsneeds to be implemented!
Memory implementation can also be an interesting feature in this current RAG enabled Chatbot.
Repo - MemoryBot
The live demo app is hosted over here
Blog - here
Video - here
Similar to Chat with PDF approach, with enabled memory.
Demo App - here
Video - here
Blog - here
