grok-search

Prerequisites

Set these environment variables before use:

• GROK_API_BASE — Your Grok-compatible API endpoint (e.g., http://localhost:8100/v1)
• GROK_API_KEY — Your API key for the Grok endpoint
• TAVILY_API_KEY — Your Tavily API key (get one free at tavily.com)

Instructions

Search for: $ARGUMENTS

Step 1: Do I Need to Search?

Yes — search when:

• User explicitly asks for external info
• Real-time data needed (latest versions, recent events, current prices)
• Need to verify knowledge accuracy
• Specific URLs, projects, or products' current status

No — skip when:

• Pure code writing/debugging with sufficient context
• User explicitly says not to search

Step 2: Choose Engine

Step 3: Execute

Grok Search (AI-driven, deep analysis)

curl -s ${GROK_API_BASE}/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ${GROK_API_KEY}" \
  -d '{
    "model": "grok-3",
    "messages": [
      {"role": "system", "content": "You are a search assistant. Search the web and provide accurate, well-sourced answers."},
      {"role": "user", "content": "SEARCH_QUERY"}
    ],
    "stream": false
  }' | jq -r '.choices[0].message.content'

Tavily Search (structured results)

curl -s "https://api.tavily.com/search" \
  -H "Content-Type: application/json" \
  -d '{
    "api_key": "'${TAVILY_API_KEY}'",
    "query": "SEARCH_QUERY",
    "search_depth": "basic",
    "max_results": 5,
    "include_answer": true
  }' | jq '{answer: .answer, results: [.results[] | {title, url, content, score}]}'

Options:

• search_depth: basic (fast) or advanced (thorough)
• topic: general (default), news, finance
• max_results: 1-20
• time_range: day, week, month, year

Tavily Extract (scrape a URL)

curl -s "https://api.tavily.com/extract" \
  -H "Content-Type: application/json" \
  -d '{
    "api_key": "'${TAVILY_API_KEY}'",
    "urls": ["https://example.com/page"],
    "extract_depth": "basic"
  }' | jq '.results[] | {url, raw_content}'

Dual Search (cross-verify)

Run both in parallel for important queries, then compare results.

Evidence Standards

• All factual conclusions need ≥2 independent sources for cross-verification
• If relying on a single source, explicitly state this limitation
• Prefer: official docs, Wikipedia, academic databases, authoritative media
• When sources conflict: present both sides, assess credibility and recency
• Confidence levels: High (multiple sources agree) · Medium (few sources) · Low (single source)

Search Complexity