🚀 AI Web Navigator (Local LLM)

This version runs fully offline using llama3 via Ollama and automates product navigation on Bunnings using Playwright.

📦 Requirements

• Ollama (to run llama3 model)

  apt-get update && sudo apt-get install -y curl gpg
  curl -fsSL https://ollama.com/install.sh | sh
  ollama serve
  ollama pull llama3

🧠 Stack

• Python 3.10
• Streamlit (for UI)
• Playwright (browser automation)
• LangChain + langchain_community.llms.Ollama

🚀 Run Locally

ollama serve In new terminal

• streamlit run main.py

Build Image

docker build --network=host -t bunnings-navigator-local .

Run App

docker run -p 8501:8501 bunnings-navigator-local

Make sure Ollama is running on the host.

✅ Features

• Entity prediction for User query, which helps in facet search. And query rephrasing
• Accepts dynamic product name
• LangChain agent (ZERO_SHOT_REACT_DESCRIPTION)
• Local reasoning via llama3
• Trajectory recovery if navigation fails