My name is Teb, I'm an elephant bot and I live in a docker container :)
My name is Teb, I'm an elephant bot and I live in a docker container :)
This personal project,has been created just for fun. It was born after seeing Chris Mair's talk about Self-Hosted RAG and it is based on stuartbot project by Noi-techpark of Bolzano.
The idea was to create a chatbot whose knowledge was based on the official PostgreSQL documentation.
The chatbot has been containerized using docker and can be installed locally or on a Linux server.
The chatbot uses:
- PostgreSQL with the pgvector extension
- The reduced 3GB model mistral-7b-instruct-v0.2.Q2_K.gguf
- The sentence embedding model is bge-m3 (license: MIT). The model is quite large (2.1 GiB) for sentence embedding models, but performs very well, can embed a variety of text sizes from short sentences to longer documents (8192 tokens) and has been trained on many languages. The model is instantiated in rag/librag.py.
Here at my personal gitlab page you can find the entire project https://gitlab.com/sscotty71/teb
and now ask to teb whatever you need about PostgreSQL .... ENJOY !!