SearchOnline Skill

A powerful, multi-source web search skill designed for AI agents like Claude Code, OpenCode, or Codex. This tool serves as a superior replacement or enhancement for built-in WebSearch tools by aggregating results from multiple premium sources and synthesizing them using Gemini AI.

🚀 Key Features

• Multi-Source Aggregation: Concurrently fetches search results from Tavily, WebSearchAPI, and Gemini Google Search Grounding.
• AI-Powered Summarization: Uses Gemini models (Flash/Pro) to synthesize aggregated data into structured, actionable Markdown.
• Optimized for Agents: Output is specifically formatted for AI consumption—concise for simple facts, structured for research.
• Robustness:
  • Concurrent execution using ThreadPoolExecutor for speed.
  • Automatic retry logic with exponential backoff.
  • Graceful fallback: If AI summarization fails, it returns formatted raw results.
• Highly Configurable: Control result depth per source (1-10) and manage API rate limits with automatic model switching.

🛠 Installation

1. Requirements

• Python 3.8+
• requests library

pip install requests

2. Environment Variables

Create a .env file in the project root or set these variables in your environment:

Variable	Required	Purpose	Get API Key
GEMINI_API_KEY	Yes	AI summarization and Google Search Grounding	Google AI Studio
TAVILY_API_KEY	Optional	Premium search results via Tavily	Tavily
WEBSEARCHAPI_KEY	Optional	Additional search coverage via WebSearchAPI	WebSearchAPI

📖 Usage

Command Line Interface (CLI)

python3 SearchOnline.py "<your search query>" [max_results]

• max_results: (Optional) Results per source (1-10, default: 5).

Example:

python3 SearchOnline.py "latest features in Python 3.13" 3

As a Claude Code Skill

This project is structured as a Claude Code skill. When integrated, the agent will prioritize SearchOnline over its built-in search tool for:

• Current events and news.
• Real-time data (weather, stocks, trending repos).
• Information beyond the model's knowledge cutoff.

⚙️ How it Works

1. Parallel Search: The script launches concurrent requests to all configured search providers.
2. Context Synthesis: It aggregates all successful responses, removes duplicates, and filters for relevance.
3. Smart Summarization: Gemini processes the data based on the query type (e.g., tables for lists, paragraphs for research).
4. Adaptive Rate Limiting: If one Gemini model hits a rate limit, the script automatically tries the next available model (gemini-3-flash-preview, gemini-flash-latest, etc.).

📂 Project Structure

• SearchOnline.py: The core implementation logic.
• SKILL.md: Configuration and instructions for Claude Code integration.
• .env.example: Template for environment variables.
• .gitignore: Pre-configured to keep your API keys safe.

📄 License

MIT License - feel free to use and modify!

---

Created with Claude Code