Env setup:

Go to https://ollama.com/library/llama3:instruct and download ollama and then both the llama3 model and the nomic-embed-text model.

Using conda, setup the python env using

conda env create -f environment.yml

---

Overview

This script implements a simple Retrieval-Augmented Generation (RAG) loop backed by:

1. A PostgreSQL table of past conversations
2. A ChromaDB vector database of embeddings
3. Calling out to an Ollama LLM for both embeddings and chat

You'll be able to:

• Store and fetch chat history from Postgres
• Build/update a vector store of those conversations
• Generate “needle-in-the-haystack” queries via the LLM
• Retrieve the most relevant past exchanges
• Stream a new LLM response, remembering it for next time
• Delete the last chat entry on command (/f)

---

Imports and Initialization

• ollama: your local LLM interface (chat + embeddings)
• chromadb: the vector database client
• psycopg: PostgreSQL driver
• colorama: for colored console output
• tqdm: progress bars
• ast, re: parsing and regexp utilities

You also define your PostgreSQL credentials in DB_PARAMS.

---