Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
Embedchain: Simple LLM Bots
Embedchain(3.8K GitHub stars, 750 forks) is a simple framework to create LLM powered bots.
Problem: Developers want to create ChatGPT like bots over their continuously updating data sets, but most of them are struggling to achieve good results. Developers face following problems:
- Given a data set, how to chunk, which embedding model and vector database to use
- How to retrieve relevant documents and what retrieval strategy to apply (semantic search or lexical or both, metadata filtering, re-ranking, HYDE)
- How to create and manage embeddings of dataset which is updating?
Langchain and LLamaIndex provide extensive tools to interface with an LLM but they are hard for most software developers to understand and use.
Embedchain solution: Python and JS packages to add a dataset and query/chat over it. Built on top of Langchain for quick GTM, Embedchain abstracts out:
โข Loading, chunking, creating vectors and storing in a vector database
โข Query formation and document retrieval strategies
โข Creating and managing embeddings.