Search Components¶

This directory contains various search service components that provide intelligent search, information retrieval, and content discovery capabilities.

📋 Component List¶

1. ModelstudioSearch - DashScope Search Component¶

Core intelligent search service that supports multiple search strategies and information sources.

Prerequisites:

Valid DashScope API key, This component is currently in beta testing phase, please contact developers and provide DASHSCOPE_API_KEY
Configured search service strategy
Stable network connection

Input Parameters (SearchInput):

messages (List): Search-related conversation messages
search_options (Dict): Search option configurations
- search_strategy: Search strategy (web, news, academic, etc.)
- max_results: Maximum number of search results
- time_range: Time range limitation
- language: Search language
- region: Geographic region limitation
search_output_rules (Dict): Output format rules
search_timeout (int): Search timeout duration
type (str): Search type

Output Parameters (SearchOutput):

search_result (str): Search result summary
search_info (Dict): Detailed search information
- sources: List of information sources
- relevance_score: Relevance scoring
- search_time: Search duration
- result_count: Number of results

Core Features:

Intelligent Search: Semantic understanding-based intelligent search
Multi-source Integration: Integrates search results from multiple information sources
Real-time Search: Retrieves latest real-time information
Result Filtering: Filters results based on relevance and quality
Search Optimization: Automatically optimizes search queries and strategies

2. ModelstudioSearchLite - DashScope Search Lite Version¶

Provides lightweight search functionality, suitable for quick queries and resource-constrained scenarios.

Prerequisites:

Visit DashScope Search to enable the service
Low network latency requirements

Key Features:

Faster response speed
Lower resource consumption
Simplified search options
Suitable for mobile applications

🔧 Environment Variable Configuration¶

Environment Variable	Required	Default	Description
`DASHSCOPE_API_KEY`	✅	-	DashScope API key
`SEARCH_DEFAULT_STRATEGY`	❌	web	Default search strategy
`SEARCH_MAX_RESULTS`	❌	10	Default maximum search results
`SEARCH_TIMEOUT`	❌	30	Search timeout (seconds)
`SEARCH_ENABLE_CACHE`	❌	true	Enable search cache
`SEARCH_REGION`	❌	global	Default search region

🚀 Usage Examples¶

Basic Search Example¶

from agentscope_runtime.tools.searches import ModelstudioSearch
import asyncio

# Initialize search component
search = ModelstudioSearch()


async def basic_search_example():
    result = await search.arun({
        "messages": [
            {"role": "user", "content": "Latest artificial intelligence development trends"}
        ],
        "search_options": {
            "search_strategy": "news",
            "max_results": 5,
            "time_range": "last_month",
            "language": "en-US"
        },
        "search_timeout": 20
    })

    print("Search result summary:", result.search_result)
    print("Information sources:", result.search_info["sources"])


asyncio.run(basic_search_example())

Multi-strategy Search Example¶

async def multi_strategy_search_example():
    # Academic search
    academic_result = await search.arun({
        "messages": [
            {"role": "user", "content": "Deep learning applications in medical diagnosis"}
        ],
        "search_options": {
            "search_strategy": "academic",
            "max_results": 10,
            "language": "en"
        }
    })

    # News search
    news_result = await search.arun({
        "messages": [
            {"role": "user", "content": "Latest AI policy updates"}
        ],
        "search_options": {
            "search_strategy": "news",
            "time_range": "last_week",
            "region": "global"
        }
    })

    print("Academic search results:", academic_result.search_result)
    print("News search results:", news_result.search_result)

asyncio.run(multi_strategy_search_example())

Advanced Search Configuration Example¶

async def advanced_search_example():
    result = await search.arun({
        "messages": [
            {"role": "user", "content": "Compare performance of different machine learning algorithms"},
            {"role": "assistant", "content": "I'll search for comparative information for you"},
            {"role": "user", "content": "Focus on accuracy and efficiency"}
        ],
        "search_options": {
            "search_strategy": "comprehensive",
            "max_results": 15,
            "filters": {
                "content_type": ["article", "paper", "report"],
                "quality_threshold": 0.8,
                "exclude_domains": ["low-quality-site.com"]
            },
            "ranking_criteria": ["relevance", "authority", "freshness"]
        },
        "search_output_rules": {
            "include_citations": True,
            "summarize_results": True,
            "highlight_key_points": True
        }
    })

    print("Comprehensive search results:", result.search_result)
    print("Search statistics:", result.search_info)

asyncio.run(advanced_search_example())

🔍 Supported Search Strategies¶

Web Search¶

General Web Search: Searches web content across the entire internet
Real-time Search: Retrieves latest published web information
Deep Search: Multi-level content mining and analysis

Professional Search¶

Academic Search: Searches academic papers and research reports
News Search: Searches news reports and current affairs
Technical Search: Searches technical documentation and API references
Product Search: Searches product information and user reviews

Multimedia Search¶

Image Search: Searches related images based on descriptions
Video Search: Searches related video content
Document Search: Searches PDF, Word and other documents

🏗️ Search Architecture¶

Query Processing¶

Query Understanding: Analyzes user query intent and key information
Query Expansion: Adds synonyms and related vocabulary
Query Optimization: Optimizes search queries to improve accuracy
Multi-strategy Routing: Selects optimal search strategy based on query type

Result Processing¶

Result Aggregation: Integrates results from multiple search sources
Deduplication: Removes duplicate and similar results
Quality Assessment: Evaluates result quality and credibility
Relevance Ranking: Ranks results by relevance
Content Summarization: Generates result summaries and key points

Caching Mechanism¶

Query Cache: Caches results for common queries
Result Cache: Caches high-quality search results
Smart Updates: Automatically updates cache based on content timeliness

📊 Search Optimization¶

Performance Optimization¶

Parallel Search: Simultaneously queries multiple information sources
Result Prefetching: Prefetches potentially relevant search results
Smart Caching: Intelligent caching strategy based on user behavior
Load Balancing: Distributes search requests to different service nodes

Quality Control¶

Source Credibility Assessment: Evaluates information source credibility
Content Quality Check: Checks content accuracy and completeness
Timeliness Verification: Verifies information timeliness
Bias Detection: Detects and flags potentially biased content

📦 Dependencies¶

aiohttp: Async HTTP client
dashscope: DashScope SDK
beautifulsoup4: HTML parsing
lxml: XML/HTML processing
nltk: Natural language processing (optional)
elasticsearch: Search engine (optional)

⚠️ Usage Considerations¶

Search Strategy Selection¶

Choose appropriate search strategy based on query type
Consider timeliness requirements of results
Balance search depth and response speed
Adjust search parameters based on user scenarios

Result Quality Management¶

Set appropriate relevance thresholds
Verify accuracy of search results
Handle cases with insufficient search results
Establish user feedback mechanism to improve search quality

API Usage Limitations¶

Follow search service call frequency limits
Set reasonable timeout durations to avoid long waits
Implement error handling and retry mechanisms
Monitor API usage and costs