mirror of
https://github.com/mims-harvard/ToolUniverse.git
synced 2026-09-19 07:31:47 +08:00
Shanghua
270 lines
9.4 KiB
Python
270 lines
9.4 KiB
Python
#!/usr/bin/env python3
|
|
"""
|
|
Tool Finder Example
|
|
|
|
Demonstrates ToolUniverse's tool finding capabilities for discovering relevant tools
|
|
"""
|
|
|
|
from tooluniverse import ToolUniverse
|
|
import time
|
|
|
|
# =============================================================================
|
|
# Tool Initialization
|
|
# =============================================================================
|
|
# Description: Initialize ToolUniverse and load all available tools
|
|
# Syntax: tu = ToolUniverse(); tu.load_tools()
|
|
tu = ToolUniverse()
|
|
tu.load_tools()
|
|
|
|
# =============================================================================
|
|
# Method 1: Semantic Tool Discovery
|
|
# =============================================================================
|
|
# Description: Use Tool_Finder for semantic-based tool discovery
|
|
# Syntax: tu.run({"name": "Tool_Finder", "arguments": {"description": "tool description", "limit": 5, "return_call_result": False}})
|
|
result1 = tu.run({
|
|
"name": "Tool_Finder",
|
|
"arguments": {
|
|
"description": "a tool for finding tools related to diseases",
|
|
"limit": 5,
|
|
"return_call_result": False
|
|
}
|
|
})
|
|
|
|
# =============================================================================
|
|
# Method 2: Keyword-Based Tool Discovery
|
|
# =============================================================================
|
|
# Description: Use Tool_Finder_Keyword for keyword-based tool discovery
|
|
# Syntax: tu.run({"name": "Tool_Finder_Keyword", "arguments": {"description": "keyword", "limit": 3}})
|
|
result2 = tu.run({
|
|
"name": "Tool_Finder_Keyword",
|
|
"arguments": {
|
|
"description": "disease",
|
|
"limit": 3
|
|
}
|
|
})
|
|
|
|
# =============================================================================
|
|
# Method 3: Performance Timing
|
|
# =============================================================================
|
|
# Description: Measure execution time for tool discovery operations
|
|
# Syntax: start_time = time.time(); result = tu.run(...); end_time = time.time()
|
|
start_time = time.time()
|
|
result3 = tu.run({
|
|
"name": "Tool_Finder",
|
|
"arguments": {
|
|
"description": "machine learning tools",
|
|
"limit": 3,
|
|
"return_call_result": False
|
|
}
|
|
})
|
|
end_time = time.time()
|
|
execution_time = end_time - start_time
|
|
|
|
# =============================================================================
|
|
# Method 4: Error Handling and Timeout Management
|
|
# =============================================================================
|
|
# Description: Handle potential errors and timeouts in tool discovery
|
|
# Syntax: try/except blocks around discovery calls
|
|
try:
|
|
result4 = tu.run({
|
|
"name": "Tool_Finder",
|
|
"arguments": {
|
|
"description": "complex query that might timeout",
|
|
"limit": 100 # Large limit that might cause timeout
|
|
}
|
|
})
|
|
except Exception as e:
|
|
# Handle timeout or other errors
|
|
if "timeout" in str(e).lower():
|
|
# Specific handling for timeout errors
|
|
pass
|
|
else:
|
|
# Handle other types of errors
|
|
pass
|
|
|
|
# =============================================================================
|
|
# Method 5: Result Processing
|
|
# =============================================================================
|
|
# Description: Process and analyze tool discovery results
|
|
# Syntax: Check result structure and extract tool information
|
|
|
|
# Process Tool_Finder results
|
|
if isinstance(result1, dict) and 'tools' in result1:
|
|
tools = result1['tools']
|
|
# Access tool information: tools[0]['name'], tools[0]['description'], etc.
|
|
pass
|
|
|
|
# Process Tool_Finder_Keyword results
|
|
if isinstance(result2, dict) and 'tools' in result2:
|
|
tools = result2['tools']
|
|
# Access tool information: tools[0]['name'], tools[0]['description'], etc.
|
|
pass
|
|
|
|
# =============================================================================
|
|
# Method 6: Batch Tool Discovery
|
|
# =============================================================================
|
|
# Description: Perform multiple tool discovery queries in sequence
|
|
# Syntax: Loop through multiple discovery queries
|
|
discovery_queries = [
|
|
{
|
|
"name": "Tool_Finder",
|
|
"arguments": {
|
|
"description": "data analysis tools",
|
|
"limit": 3,
|
|
"return_call_result": False
|
|
}
|
|
},
|
|
{
|
|
"name": "Tool_Finder_Keyword",
|
|
"arguments": {
|
|
"description": "protein",
|
|
"limit": 2
|
|
}
|
|
},
|
|
{
|
|
"name": "Tool_Finder",
|
|
"arguments": {
|
|
"description": "visualization tools",
|
|
"limit": 2,
|
|
"return_call_result": False
|
|
}
|
|
}
|
|
]
|
|
|
|
batch_results = []
|
|
for query in discovery_queries:
|
|
try:
|
|
result = tu.run(query)
|
|
batch_results.append(result)
|
|
except Exception as e:
|
|
# Handle individual query failures
|
|
batch_results.append({"error": str(e)})
|
|
|
|
# =============================================================================
|
|
# Method 7: Discovery Parameter Optimization
|
|
# =============================================================================
|
|
# Description: Optimize discovery parameters for better results
|
|
# Syntax: Adjust limit and other parameters
|
|
|
|
# Small limit for quick testing
|
|
quick_result = tu.run({
|
|
"name": "Tool_Finder",
|
|
"arguments": {
|
|
"description": "bioinformatics tools",
|
|
"limit": 1,
|
|
"return_call_result": False
|
|
}
|
|
})
|
|
|
|
# Larger limit for comprehensive results
|
|
comprehensive_result = tu.run({
|
|
"name": "Tool_Finder",
|
|
"arguments": {
|
|
"description": "bioinformatics tools",
|
|
"limit": 10,
|
|
"return_call_result": False
|
|
}
|
|
})
|
|
|
|
# =============================================================================
|
|
# Method 8: Result Validation
|
|
# =============================================================================
|
|
# Description: Validate discovery results and check for errors
|
|
# Syntax: Check result structure and error conditions
|
|
|
|
def validate_discovery_result(result, tool_name):
|
|
"""Validate tool discovery result structure and content"""
|
|
if isinstance(result, dict):
|
|
if "error" in result:
|
|
# Handle error response
|
|
return False, f"Error in {tool_name}: {result['error']}"
|
|
elif "tools" in result:
|
|
# Valid discovery result
|
|
tools = result['tools']
|
|
return True, f"Found {len(tools)} tools"
|
|
else:
|
|
# Unexpected result structure
|
|
return False, f"Unexpected result structure from {tool_name}"
|
|
else:
|
|
# Non-dictionary result
|
|
return False, f"Unexpected result type from {tool_name}"
|
|
|
|
# Validate results
|
|
is_valid, message = validate_discovery_result(result1, "Tool_Finder")
|
|
is_valid, message = validate_discovery_result(result2, "Tool_Finder_Keyword")
|
|
|
|
# =============================================================================
|
|
# Method 9: Tool Information Extraction
|
|
# =============================================================================
|
|
# Description: Extract specific information from discovered tools
|
|
# Syntax: Access tool properties and metadata
|
|
|
|
def extract_tool_info(tools_result):
|
|
"""Extract key information from tool discovery results"""
|
|
if isinstance(tools_result, dict) and 'tools' in tools_result:
|
|
tools = tools_result['tools']
|
|
tool_info = []
|
|
for tool in tools[:3]: # First 3 tools
|
|
info = {
|
|
'name': tool.get('name', 'Unknown'),
|
|
'description': tool.get('description', 'No description')[:100] + '...',
|
|
'type': tool.get('type', 'Unknown')
|
|
}
|
|
tool_info.append(info)
|
|
return tool_info
|
|
return []
|
|
|
|
# Extract tool information
|
|
tool_info1 = extract_tool_info(result1)
|
|
tool_info2 = extract_tool_info(result2)
|
|
|
|
|
|
# =============================================================================
|
|
# Tool_Rag Test (simple, like Tool_Finder examples)
|
|
# =============================================================================
|
|
|
|
# Try: simple Tool_Rag query
|
|
result_rag = tu.run({
|
|
"name": "Tool_RAG",
|
|
"arguments": {
|
|
"description": "Find tools for gene prediction.",
|
|
"limit": 2
|
|
}
|
|
})
|
|
print("\n[Tool_RAG] Result:")
|
|
print(result_rag)
|
|
|
|
|
|
# =============================================================================
|
|
# Summary of Tool Discovery Tools
|
|
# =============================================================================
|
|
# Available tool discovery tools provide intelligent tool finding capabilities:
|
|
# - Tool_Finder: Semantic-based tool discovery using natural language descriptions
|
|
# - Tool_Finder_Keyword: Keyword-based tool discovery for specific terms
|
|
#
|
|
# Common parameters:
|
|
# - description: Natural language description or keyword for tool search
|
|
# - limit: Maximum number of tools to return
|
|
# - return_call_result: Whether to return actual tool execution results
|
|
#
|
|
# Result structures:
|
|
# - Both tools return "tools" array with tool information
|
|
# - Each tool entry contains: name, description, type, and other metadata
|
|
#
|
|
# Error handling:
|
|
# - Check for "error" key in dictionary responses
|
|
# - Handle timeout exceptions for complex queries
|
|
# - Validate result structure before processing
|
|
# - Use appropriate limits to avoid timeouts
|
|
#
|
|
# Performance considerations:
|
|
# - Start with small limits for testing
|
|
# - Use timing to measure discovery performance
|
|
# - Consider batch operations for multiple queries
|
|
# - Handle individual query failures gracefully
|
|
#
|
|
# Use cases:
|
|
# - Finding tools for specific research tasks
|
|
# - Discovering available functionality
|
|
# - Tool recommendation systems
|
|
# - Automated tool selection workflows |