Web Search Notes

Modelworks Chatbot with Web Search

This wiki explains how to set up and use the Modelworks Gradio chat application, which integrates OllamaLLM with live web search via SerpAPI.

---

Prerequisites

• Python 3.8+
• Virtual environment (recommended)
• Ollama server running locally with a model loaded (e.g., deepseek-r1:1.5b).
• SerpAPI account and API key.

---

Installation

1. Clone your repo (or copy main.py into your project directory):

   git clone <your-repo-url>
   cd <your-project>

2. Create and activate a Python virtual environment: python3 -m venv venv source venv/bin/activate # macOS/Linux venv\Scripts\activate # Windows
3. Install dependencies

   pip install --upgrade \
   gradio \
   langchain[serpapi] \
   langchain-community \
   langchain-ollama

Configuration

1. Start Ollama with your chosen model

   ollama serve --model gemma3:1b

2. Set your SerpAPI key in the environment

   # macOS / Linux
   export SERPAPI_API_KEY="your_serpapi_key"
   
   # Windows PowerShell
   setx SERPAPI_API_KEY "your_serpapi_key"
   

File Structure

project/
├── main.py          # Gradio app with web search agent
├── venv/            # Python virtual environment
└── requirements.txt # optionally freeze your deps

Running the App

With your venv activated and Ollama running:

python main.py

By default, the Gradio UI will launch at http://localhost:7860.

Usage

1. Open your browser at http://localhost:7860.
2. Type a query in the chatbox (e.g., “What’s the latest news on renewable energy?”).
3. Send — the agent will:
   • Optionally call SerpAPI to fetch real-time web results.
   • Combine its reasoning with those results.
   • Return a final answer in the chat UI.

Example: User: What’s the population of Australia? Bot: (🔍 Web Search) “population of Australia” → 26 million (est.) Final Answer: As of 2025, Australia’s population is approximately 26 million.