Miao Assistant is an implementation of a Retrieval-Augmented Generation (RAG) system with LangChain that can read, understand, and summarize vast amounts of data in real time. It is currently being developed to function as a local assistant on a personal machine.
Python · LangChain · watsonx.ai (Mistral 8x7B) · Chroma DB · Gradio
Running the code setup in the localapp folder and you will meet your Miao assistant, Miao can be shown as the following:
First, load your document and ask any question as you want :), see:
At the moment, for small-size files, it works well 💯 as the following:
But not yet always perfectly for the bigger ones, for example:
-
Designed and implemented an intelligent assistant capable of reading, understanding, and summarizing complex documents in real time using a Retrieval-Augmented Generation (RAG) architecture.
-
Built an end-to-end RAG pipeline: document loading, text chunking, embedding generation, and vector storage with Chroma for semantic search.
-
Developed a retriever–LLM QA bot with a Gradio interface, integrating watsonx.ai API (Mistral 8x7B) for contextual question answering and summarization.
-
Achieved high factual reliability, Grounded Answer Rate = 87.1 %, and Hallucination < 4 % on the test dataset.
Currently optimizing for local deployment and large-file processing, after identifying performance bottlenecks on multi-MB PDF inputs and of course expect to be the best local assistant on a personal machine ;)