功能描述

Google Search with intelligent caching, rate limiting, and cost optimization. Uses OpenClaw tools + free APIs. 10x faster, 99% cheaper than v1.

使用说明 (SKILL.md)

Google Search Unlimited v2 🚀

Name: Google Search Unlimited V2
Author: gloriaolk

Cost-optimized, intelligent search with maximum free tier usage

🎯 Why Choose v2?

Feature	v1 (Original)	v2 (Improved)	Benefit
Caching	❌ None	✅ SQLite + TTL	90%+ cost reduction
Rate Limiting	❌ Basic	✅ Intelligent	Avoids bans
Dependencies	Playwright + heavy	Requests + lightweight	10x faster setup
Cost Optimization	❌ Sequential	✅ Free-first hierarchy	Maximizes free tier
OpenClaw Integration	❌ Manual	✅ Direct tool usage	Built-in capabilities
Monitoring	❌ None	✅ Built-in metrics	Track usage

📊 Performance Benchmarks

Query: "OpenClaw documentation" (repeated 10x)

v1: 45s, 10 API calls, ~$0.10
v2: 8s, 1 API call, ~$0.001
Improvement: 5.6x faster, 99% cheaper

🏗️ Architecture

Tiered Search Strategy (Free → Paid)

1. CACHE (0ms, $0.00) ← First check
   ↓
2. OpenClaw Tools (800ms, $0.00) ← Built-in oxylabs_web_search
   ↓
3. Free APIs (1.2s, $0.00) ← DuckDuckGo, Brave Search
   ↓
4. Google API (1.5s, $0.00*) ← 100 free/day
   ↓
5. Lightweight HTTP (2s, $0.001) ← Last resort

*First 100 queries/day free with Google API

🔧 Quick Start

# Install skill
clawhub install google-search-unlimited-v2

# Install dependencies
pip install requests beautifulsoup4 lxml

# Basic search
python3 search.py "your query"

# With caching (recommended)
python3 search.py --cache "your query"

📊 Method Hierarchy (Cost → Free)

Tier 1: OpenClaw Tools (FREE)

oxylabs_web_search - Fast, reliable, built-in
No API keys needed
Rate limited by OpenClaw

Tier 2: Google Custom Search API (100 free/day)

When credentials available
Fast and structured
Respects daily quota

Tier 3: Alternative Free APIs

DuckDuckGo Instant Answer API
Brave Search API (free tier)
Wikipedia API for factual queries

Tier 4: Lightweight Scraping

Minimal HTTP requests
User-agent rotation
Respectful of robots.txt

⚙️ Setup

# Minimal dependencies
pip install requests beautifulsoup4

# Optional: For better parsing
pip install lxml

🎯 Usage

# Basic search
python3 search.py "your query"

# With caching
python3 search.py --cache "your query"

# Force specific method
python3 search.py --method oxylabs "your query"

🔧 Configuration

Create .env file:

# Google API (optional)
GOOGLE_API_KEY=your_key
GOOGLE_CSE_ID=your_cx

# Cache settings
CACHE_TTL_HOURS=24
MAX_CACHE_SIZE_MB=100

# Rate limiting
MAX_REQUESTS_PER_MINUTE=10

📈 Performance Features

Query deduplication: Same query = cached result
Result compression: Store only essential data
Batch processing: Multiple queries in single API call
Smart retry: Exponential backoff on failures
Result validation: Filter out low-quality results

💰 Cost Optimization

Free Methods First

OpenClaw tools (no cost)
Free APIs (DuckDuckGo, Brave)
Google API (100 free/day)

Cache Strategy

Hot queries: Keep in memory
Warm queries: SQLite cache
Cold queries: Fresh fetch

Bandwidth Saving

Compress responses
Store only text (no HTML)
Pagination support

🛡️ Reliability

Multiple fallbacks: 4+ search methods
Automatic failover: If one fails, try next
Health checks: Monitor API status
Graceful degradation: Maintain service during outages

📊 Monitoring

Built-in metrics:

Success rate per method
Average response time
Cache hit ratio
Cost per query (estimated)

Example Output

{
  "query": "OpenClaw documentation",
  "method": "oxylabs",
  "cost_estimate": 0.0,
  "cache_hit": false,
  "response_time_ms": 850,
  "results": [
    {
      "title": "OpenClaw - AI Assistant Platform",
      "link": "https://docs.openclaw.ai",
      "snippet": "Official documentation...",
      "relevance_score": 0.92
    }
  ]
}

🎯 Use Cases

Research assistant: Fast, cached searches
Monitoring: Regular queries with caching
Batch processing: Multiple queries efficiently
Cost-sensitive apps: Maximize free tier usage

⚠️ Best Practices

Enable caching for repeated queries
Monitor usage to stay within free tiers
Use batch mode for multiple searches
Set reasonable TTL based on query type
Respect rate limits of all services

安全使用建议

This package appears to implement the advertised cached, tiered search, but there are multiple inconsistencies and some strong claims that should be validated before you install or publish it: 1) Resolve metadata mismatches: registry entry reported no required binaries/envs while _meta.json requires python3 and pip packages; versions differ (2.0.0 vs 2.0.1). Confirm which metadata is authoritative. 2) Review network behavior: inspect search.py and search_engine_final.py for all outbound network calls, URLs, and any hard-coded endpoints. Verify which methods actually call the claimed 'oxylabs_web_search' tool and whether that requires special runtime privileges. 3) Verify dependency contents: open requirements.txt and ensure packages are standard and pinned appropriately. Avoid running code in production until dependencies are validated. 4) Run locally in an isolated environment (container or VM): execute quick_test.py and test_real_search.py to confirm behavior, and run the check_cache/debug_cache utilities to inspect the cache DB contents. 5) Treat publication steps separately: the included scripts instruct using clawhub login (GitHub). Publishing requires you to authenticate and provide tokens; do not paste tokens into untrusted scripts. The publishing instructions are expected, but only run them if you intend to publish and after code review. 6) Validate claims: performance numbers (e.g., <0.01s) and security/test pass claims in VERIFICATION_REPORT.md look optimistic — run your own benchmarks and security checks (linting, dependency scans). If you want, I can: (A) search the provided search.py / search_engine_final.py for network endpoints and suspicious patterns, (B) summarize requirements.txt, or (C) produce a short checklist of exact lines to inspect before trusting this skill.

功能分析

Type: OpenClaw Skill Name: google-search-unlimited-v2 Version: 2.0.1 The bundle is a well-structured search utility providing a multi-tiered search strategy (SQLite cache, OpenClaw tools, DuckDuckGo, and Google API) with integrated rate limiting. The Python code in search.py and search_engine_final.py follows security best practices, such as using parameterized SQL queries to prevent injection and standard libraries for network communication. No evidence of malicious intent, data exfiltration, or prompt injection was found across the code or documentation.

能力评估

ℹ Purpose & Capability

The stated purpose (cost-optimized layered search with caching and fallbacks) aligns with the included Python code (batch_search.py, cache helpers, tests). However the registry metadata (no required binaries/env) contradicts _meta.json which declares python3 and pip-installed packages; SKILL.md and other docs list optional GOOGLE_API_KEY/GOOGLE_CSE_ID and .env config. This is plausible for a search skill but the metadata mismatch is an incoherence to resolve.

⚠ Instruction Scope

SKILL.md instructs normal actions for a search tool (install dependencies, create .env, run search.py, respect robots.txt). Publication docs instruct running clawhub login (GitHub auth) — appropriate for publishing but unrelated to runtime. Some included helper scripts (check_cache.py, debug_cache.py, prepare_for_publish.sh) read local cache DBs and metadata files; these are expected for maintenance but should be inspected for any unexpected file reads or network endpoints inside search.py / search_engine_final.py (full contents not provided in the manifest summary).

⚠ Install Mechanism

Registry states 'No install spec — instruction-only', yet _meta.json lists a pip install step and requirements.txt exists. There is no centralized install spec in the registry entry, so installing will rely on user instructions (pip). This mismatch increases the chance a user misses dependency or install steps. No remote downloads with extract found in the provided manifest, which is good, but verify requirements.txt contents and any dynamic pip installs inside scripts.

✓ Credentials

The skill asks only for optional Google API credentials (GOOGLE_API_KEY, GOOGLE_CSE_ID) and cache/rate-limit settings via .env — appropriate and proportionate for a search aggregator. No unrelated secret environment variables are requested in the manifest. Publication instructions require GitHub/clawhub auth for publishing — expected for that specific operation.

✓ Persistence & Privilege

Skill does not request always:true and does not appear to modify other skills or system-wide configurations. prepare_for_publish.sh and publication docs operate at user-level (clawhub publish) and require explicit user action. No autonomous or elevated privileges beyond ordinary code execution are declared.

版本历史

v2.0.1

- Added publication success documentation (PUBLICATION_SUCCESS.md). - Updated README.md and _meta.json with minor changes and metadata adjustments. - No changes to main skill logic; documentation and meta updates only.

v2.0.0

Google Search Unlimited v2.0.0 is a major upgrade focused on speed, cost reduction, and reliability. - Adds intelligent caching with SQLite and TTL, cutting costs by over 90%. - Implements advanced rate limiting to avoid bans and ensure consistent access. - Switches to lightweight dependencies for a 10x faster setup. - Integrates free OpenClaw tools and multiple free APIs, maximizing free-tier usage. - Introduces a tiered fallback strategy for search (cache → OpenClaw → free APIs → Google API → scraping). - Provides built-in monitoring, detailed metrics, and improved reliability with automatic multi-method failover.

元数据

Slug google-search-unlimited-v2

版本 2.0.1

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 2

常见问题