tavily-search

What This Skill Does

The tavily-search skill is a specialized AI agent tool designed to perform real-time web searches, specifically engineered for the needs of LLMs and Retrieval-Augmented Generation (RAG) workflows. Unlike traditional search engines that return raw, cluttered HTML, Tavily processes search results into clean, AI-ready text. It handles search-depth adjustments, domain filtering, and time-sensitive queries, allowing agents to retrieve current data on news, research papers, and niche topics that are beyond their static training cut-off.

Installation

To integrate this skill into your OpenClaw environment, ensure you have your Tavily API key ready. You can install the skill by executing the following command in your terminal:

clawhub install openclaw/skills/skills/chasehl/search-with-tavily

Once installed, authenticate by setting your environment variable: export TAVILY_API_KEY="your_key_here". This allows the agent to communicate directly with the Tavily API, ensuring all queries are authorized and tracked according to your plan.

Use Cases

• Market Research: Quickly gather real-time data on industry trends or financial news.
• Academic & Technical Writing: Use the get_search_context function to pull relevant excerpts for citations or technical literature reviews.
• Content Creation: Fetch current events and facts to improve the accuracy and relevance of AI-generated content.
• Fact-Checking: Verify claims or specific data points against the latest indexed web content.

Example Prompts

1. "Perform a comprehensive search on the latest advancements in solid-state battery technology and summarize the top three breakthroughs from the last month."
2. "Research the recent 2024 legislative changes regarding privacy policies in the EU and provide a detailed summary suitable for a business briefing."
3. "Search for the current market sentiment regarding renewable energy stocks and extract only the relevant statistics for my report."

Tips & Limitations

• Token Efficiency: For large RAG applications, always use get_search_context with a specified max_tokens parameter to prevent overflowing your context window.
• Search Depth: Use basic search for quick facts and comprehensive for deep-dive research tasks. Note that comprehensive searches may take slightly longer to process.
• Cost Awareness: Each search consumes credits. Avoid high-frequency loops in your agent logic unless necessary.
• Limitations: The skill relies on the Tavily index; it cannot access behind-the-login-wall private content or live-stream media.