Project introduction and structure

The "Leads-Qualifier-Agent" is an autonomous agent that receives, analyzes, enriches, and prioritizes new business leads from a website contact form. Built using LangGraph and LangChain, and powered by a locally-run open-source LLM through Ollama, this agent can automate manual administrative tasks and increase sales efficiency.

Visual flowchart of real-time usage:

A brief description of each file in this directory:

• main.py: The core application. Contains the FastAPI server, the LangGraph agent definition (state, nodes, edges), and the API endpoint logic.
• tools.py: A module containing the standalone functions (tools) that the AI agent uses, such as the web scraper and database writer.
• streamlit_app.py: A simple web-based user interface for demonstrating the system. It acts as a client that sends requests to the FastAPI backend.
• requirements.txt: A list of all the Python dependencies required to run the project.
• leads.db: The SQLite database file that is automatically created to store the processed lead data.
• .env: (To be created by user) A file to store secret API keys for services like LangSmith.
• test_tools.py: A script used during development to test the functions in tools.py independently. (Run once to create the sqlite database if not existing)

Setup & Installation

Follow these steps to set up the project environment.

1. Prerequisites

• Ensure you have Python 3.10 installed.
• Install Ollama on your machine.
• Then, open the terminal on your device.
• Create a virtual environment for the project to ensure installed dependencies stay local to the project. Click here for a guide on how to create and activate one.

3. Project Setup cd /path/to/this/folder

4. Install Dependencies Install all required packages using the requirements.txt file: pip install -r requirements.txt

5. Set Up Local LLM: Pull the llama3:8b model, which will be used by the agent: ollama pull llama3:8b

6. Configure LangSmith: To visualize the agent's workflow, you can use LangSmith. Follow the below steps:

• Sign up for a free account at smith.langchain.com.
• Create a new project and an API key.
• Create a file named .env in the project root and add your keys: Then, inside this projects .env file:

1. LANGCHAIN_TRACING_V2="true"
2. LANGCHAIN_API_KEY="YOUR_LANGSMITH_API_KEY"
3. LANGCHAIN_PROJECT="YOUR_PROJECT_NAME"

Running the End-to-End System

The system requires two separate processes to be running in two different terminal windows. Make sure you activate the virtual environment (source .venv/bin/activate) in each new terminal.

Start the Ollama Server

First, ensure the Ollama application is running. On macOS, this is typically done by launching the Ollama.app. On other systems, you might run ollama serve in the terminal. The Ollama icon should be visible in your menu bar/system tray.

Terminal 1: run the FastAPI Backend

uvicorn main:app --reload

Wait until you see the message Application startup complete.

Terminal 2: Run the Streamlit Frontend

streamlit run streamlit_app.py

A new browser tab should automatically open to the Streamlit app.

How to Use the Demo

With both terminals running, use the Streamlit web interface that opened in your browser.

• Fill in the "Name", "Email", and "Message" fields.
• Click the "Qualify Lead" button. Observe the results displayed on the Streamlit page. You can also monitor the FastAPI server logs in Terminal 1 to see the agent's real-time thought process.

Access the stored leads:

Finally, open the leads.db sqlite3 database using a database viewer or by installing the SQLite Viewer extension on VSCode. You can also visualise the agents processes on langsmith under this project.