Chat Assistant

• Building a Basic Webpage Opener and EXCEL SHEET (XLSX OR CSV) Question Answering Agent

Screenshot - Conversational Agent:

Requirements

Make sure to have following dependencies listed in your requirements.txt file:

• streamlit
• langchain
• langchain-community
• python-dot
• google

Installation

Git clone the repo

python3 -m venv ./ollama-langchain-venv

source ./ollama-langchain-venv/bin/activate

docker build . -t yourdocker/ollama-langchain:0.5

ollama-langchain docker images

docker-compose up

docker exec -it ollama-langchain-ollama-container-1 ollama run phi

you should be able to see the downloaded model files and manifests in your local folder ./data/ollama (which is internally mapped to /root/.ollama for the container, which is where Ollama looks for the downloaded models to serve)

Screenshot - Opening Website through natural prompt

Run

python app_term.py 

To converse from terminal To converse from streamlit

Closing website

Streamlit has a sandboxed environment hence it can't redirect to another site. So we are giving the sites to open explicitly. Couldn't implement because of limitations in webbrowser module of python, tried selenium which was also unsuccessful.

RAG training was done in this notebook Colab

.env setup

LANGCHAIN_API_KEY=""
LANGCHAIN_PROJECT="Task"
Please comment the part if not required. Use your own API key and set up a project name to see the analytics in Langsmith.