Deep Search

What This Skill Does

Deep Search is a sophisticated, multi-tier intelligence gathering engine designed for the OpenClaw agent ecosystem. Powered by Perplexity AI, it functions as a highly optimized router that intelligently selects the most appropriate model tier based on the complexity of the user's request. By utilizing three distinct tiers—Quick (sonar), Pro (sonar-pro), and Deep (sonar-reasoning-pro)—the skill balances speed and computational depth. It also integrates specialized focus modes, allowing the agent to restrict data retrieval to academic, news, YouTube, or Reddit sources to ensure the output is relevant to the user's intent. The skill features automatic Langfuse observability, allowing users to track performance, cost, and usage patterns in real-time.

Installation

To integrate Deep Search into your OpenClaw environment, execute the following command in your terminal: clawhub install openclaw/skills/skills/aiwithabidi/deep-search. Before running the tool, ensure you have a valid Perplexity API key. Add this key to your environment variables as PERPLEXITY_API_KEY. Optionally, if you wish to track model costs via Langfuse, provide an OPENROUTER_API_KEY. Once installed, the deep_search.py script becomes available within your skill directory for seamless command-line or programmatic invocation.

Use Cases

This skill is ideal for professionals and developers who require reliable, fact-checked information without hallucinations. Use it for:

• Fact-Checking: Quickly verifying specific claims against live internet data.
• Market Research: Utilizing the 'deep' tier to conduct competitive analysis and synthesize long-form reports.
• Academic Inquiry: Leveraging the 'academic' focus mode to aggregate research papers and technical documentation.
• Content Monitoring: Tracking breaking news or sentiment trends on platforms like Reddit.
• Comparative Analysis: Assessing product features or technology stacks side-by-side using the 'pro' tier.

Example Prompts

1. "Deep Search, use the pro tier to compare the current architectural advantages of using Postgres versus MongoDB for a scalable microservices backend."
2. "Perform a deep analysis on the recent advancements in Large Language Model reasoning capabilities, citing academic sources and news releases from the last month."
3. "Search Reddit for user discussions regarding OpenClaw agent deployment issues and summarize the top three commonly reported solutions."

Tips & Limitations

To maximize the efficiency of Deep Search, always match your tier to the complexity of the prompt; using the 'deep' tier for simple factual queries increases latency and cost unnecessarily. Be aware that the 'deep' mode can take up to 20 seconds to process, so ensure your agent's timeout settings allow for this. Always provide your API key securely via environment variables rather than hardcoding it into your scripts. If you encounter connectivity errors, confirm your network allows access to the Perplexity API endpoints.