OpenRouter Web Search

A Python application that combines OpenRouter AI models with web search capabilities to provide intelligent responses enriched with current web information.

Features

• 🤖 Multiple AI Models: Access 100+ AI models through OpenRouter
• 🔍 Web Search Integration: Search DuckDuckGo, Wikipedia, and Google
• 💬 Interactive Chat Mode: Real-time conversation with AI
• 📦 CLI Interface: Single query processing with various options
• 📊 JSON Output: Structured data output for integrations
• 🔄 Model Switching: Easy switching between different AI models

Installation

1. Clone the repository:

   git clone https://github.com/nvrunx/openrouter-web-search.git
   cd openrouter-web-search

2. Install dependencies:

   pip install -r requirements.txt

3. Set up environment variables:

   cp .env.example .env

   Edit .env and add your OpenRouter API key:

   OPENROUTER_API_KEY=your_openrouter_api_key_here
   

Getting API Keys

OpenRouter API Key (Required)

1. Visit OpenRouter.ai
2. Sign up for an account
3. Go to Keys section
4. Create a new API key
5. Add it to your .env file

Cost: OpenRouter offers competitive pricing with models starting from $0.0005 per 1K tokens.

Usage

Interactive Mode

Run the application in interactive chat mode:

python main.py

This starts an interactive session where you can:

• Ask questions and get AI responses with web search
• Type models to see and switch between available AI models
• Type nosearch <query> to query without web search
• Type quit, exit, or q to end the session

Single Query Mode

Process a single query:

# Basic query
python main.py -q "What is the latest in AI technology?"

# Use specific model
python main.py -q "Explain quantum computing" -m "anthropic/claude-3-haiku"

# Skip web search
python main.py -q "What is Python?" --no-search

# JSON output
python main.py -q "Latest AI news" --json

Command Line Options

Options:
  -q, --query QUERY        Single query to process
  -m, --model MODEL        AI model to use (default: openai/gpt-3.5-turbo)
  --no-search              Skip web search and use AI only
  --json                   Output results in JSON format
  --list-models            List available models and exit
  -h, --help               Show help message

Available Models

List all available models:

python main.py --list-models

Popular models include:

• openai/gpt-3.5-turbo (Default - Fast and cost-effective)
• openai/gpt-4 (Most capable OpenAI model)
• anthropic/claude-3-haiku (Fast Claude model)
• anthropic/claude-3-sonnet (Balanced Claude model)
• google/gemini-pro (Google's flagship model)
• meta-llama/llama-3-70b-instruct (Open source alternative)

Code Structure

openrouter_client.py

Main OpenRouter API client with features:

• Model listing and selection
• Chat completions with streaming support
• Token usage tracking
• Error handling

web_search.py

Web search tool supporting:

• DuckDuckGo API: Instant answers and related topics
• Wikipedia API: Encyclopedia search results
• Google Search: Web scraping (educational purposes)
• Comprehensive Search: Combines multiple sources

main.py

Main application featuring:

• Interactive chat mode
• CLI argument processing
• Query enhancement with search results
• Multiple output formats

Examples

Basic Usage

from openrouter_client import OpenRouterClient
from web_search import WebSearchTool

# Initialize clients
ai_client = OpenRouterClient()
search_tool = WebSearchTool()

# Simple AI query
response = ai_client.simple_query("Explain machine learning")
print(response)

# Web search
results = search_tool.comprehensive_search("Python tutorials", 5)
print(search_tool.format_results(results))

Advanced Usage

from main import AIWebSearchApp

# Initialize the combined app
app = AIWebSearchApp()

# Query with web search
result = app.search_and_ask(
    "What are the latest developments in AI?",
    model="anthropic/claude-3-sonnet",
    use_search=True,
    max_search_results=10
)

print(result['ai_response'])

Cost Optimization

Since you're cost-conscious about API usage:

1. Use efficient models: Start with gpt-3.5-turbo for most queries
2. Limit search results: Reduce max_search_results to save on context tokens
3. Skip search when not needed: Use --no-search for general knowledge questions
4. Monitor usage: Check token usage in responses

Estimated costs (with gpt-3.5-turbo):

• Simple query: ~$0.001-0.002
• Query with web search: ~$0.003-0.005
• Interactive session (10 queries): ~$0.02-0.05

Troubleshooting

Common Issues

1. Missing API Key:

   ValueError: OpenRouter API key is required
   

   Solution: Set OPENROUTER_API_KEY in your .env file

2. Network Errors:

   Error: Request timeout
   

   Solution: Check internet connection, try different search sources

3. Rate Limiting:

   Error: Too Many Requests
   

   Solution: Wait a moment before retrying, consider upgrading API plan

---

Built by nvrunx 🚀

If you find this project useful, please give it a star! ⭐