tooluniverse.compose_scripts package¶
Submodules¶
tooluniverse.compose_scripts.biomarker_discovery module¶
Biomarker Discovery Workflow Discover and validate biomarkers for a specific disease condition using compose tools
tooluniverse.compose_scripts.comprehensive_drug_discovery module¶
Comprehensive Drug Discovery Pipeline Complete end-to-end drug discovery workflow from disease to optimized candidates
tooluniverse.compose_scripts.drug_safety_analyzer module¶
Drug Safety Analysis Pipeline Comprehensive drug safety analysis combining adverse event data, literature review, and molecular information
- tooluniverse.compose_scripts.drug_safety_analyzer.compose(arguments, tooluniverse, call_tool)[source]¶
Main composition function for DrugSafetyAnalyzer
- Parameters:
arguments (
dict) – Input arguments containing drug_name, patient_sex, serious_events_onlytooluniverse – ToolUniverse instance
call_tool – Function to call other tools
- Returns
dict: Comprehensive drug safety analysis result
tooluniverse.compose_scripts.enhanced_multi_agent_literature_search module¶
Enhanced Multi-Agent Literature Search System Uses MemoryManager for session-based memory management
tooluniverse.compose_scripts.literature_tool module¶
Literature Search & Summary Tool Minimal compose tool perfect for paper screenshots
tooluniverse.compose_scripts.multi_agent_literature_search module¶
Multi-Agent Literature Search Compose Function¶
This module implements the compose function for the multi-agent literature search system.
- tooluniverse.compose_scripts.multi_agent_literature_search.compose(arguments, tooluniverse, call_tool, stream_callback=None)[source]¶
Multi-agent literature search compose function
- Parameters:
arguments (
dict) – Input parameters from the tool calltooluniverse (
ToolUniverse) – Reference to the ToolUniverse instancecall_tool (
function) – Function to call other toolsstream_callback (
callable, optional) – Callback function for streaming output
- Returns
dict: The result of the multi-agent search
tooluniverse.compose_scripts.output_summarizer module¶
Output Summarization Composer Script
This script handles the intelligent summarization of tool outputs by: 1. Chunking large outputs into manageable pieces 2. Processing each chunk with AI-powered summarization 3. Merging the summarized chunks into a coherent final summary
The script leverages ToolUniverse’s AgenticTool infrastructure to provide intelligent, context-aware summarization that focuses on information relevant to the original query.
- tooluniverse.compose_scripts.output_summarizer.compose(arguments, tooluniverse, call_tool)[source]¶
Main composition function for output summarization.
This function orchestrates the complete summarization workflow: - Chunks the input text into manageable pieces - Summarizes each chunk using AI - Merges the summaries into a final coherent result
- Parameters:
arguments (
Dict[str, Any]) –Dictionary containing: - tool_output (str): The original tool output to be summarized - query_context (str): Context about the original query - tool_name (str): Name of the tool that generated the output - chunk_size (int, optional): Size of each chunk for processing - focus_areas (str, optional): Areas to focus on in summarization - max_summary_length (int, optional): Maximum length of final
summary
tooluniverse – ToolUniverse instance for tool execution
call_tool – Function to call other tools within the composition
- Returns
- Dict[str, Any]: Dictionary containing:
success (bool): Whether summarization was successful
original_length (int): Length of original output
summary_length (int): Length of final summary
chunks_processed (int): Number of chunks processed
summary (str): The summarized output
tool_name (str): Name of the original tool
error (str, optional): Error message if summarization failed
tooluniverse.compose_scripts.tool_description_optimizer module¶
tooluniverse.compose_scripts.tool_discover module¶
- tooluniverse.compose_scripts.tool_discover.iterative_comprehensive_optimization(tool_config, call_tool, max_iterations=5, target_score=8.5, temp_dir=None)[source]¶
Comprehensive optimization with guaranteed minimum iterations and multi-agent improvement strategy
tooluniverse.compose_scripts.tool_graph_composer module¶
Simplified and fixed tool_graph_composer.py This version includes better error handling and avoids the ‘unhashable type’ issue.
- tooluniverse.compose_scripts.tool_graph_composer.compose(arguments, tooluniverse, call_tool)[source]¶
Compose function for building tool compatibility graphs.
- Parameters:
arguments – Dictionary with composition parameters
tooluniverse – ToolUniverse instance
call_tool – Function to call other tools
- Returns
Dictionary with results and file paths
tooluniverse.compose_scripts.tool_graph_generation module¶
Tool Graph Generation Compose Script
Efficiently evaluates directional data-flow relationships between all unique pairs of provided tool configs using one agentic tool:
ToolRelationshipDetector
Outputs a graph structure with edges representing valid directional relationships. Each edge stores: source, target, rationale.
- Performance considerations:
Iterates i<j once (O(N^2/2) pairs)
Lightweight JSON serialization of minimal fields
Optional batching hook (currently sequential because call_tool likely remote)
- param tool_configs:
- type tool_configs:
list[dict]- param max_tools:
- type max_tools:
- param output_path:
- type output_path:
str) path to write resulting graph JSON (default './tool_relationship_graph.json'- param save_intermediate_every:
- type save_intermediate_every:
int) checkpoint frequency (default 5000 pairs processed- returns:
nodes, edges, stats
- rtype:
dict with keys
tooluniverse.compose_scripts.tool_metadata_generator module¶
Tool Metadata Generation Pipeline Generates comprehensive metadata for a list of tools by extracting details from their configuration files
- tooluniverse.compose_scripts.tool_metadata_generator.compose(arguments, tooluniverse, call_tool)[source]¶
Main composition function for Tool Metadata Generation
- Parameters:
arguments (
dict) – Input arguments containing a list of tool config JSONs as well as a tool_type_mappings dict for non-API tools (e.g., {‘Databases’: [‘XMLTool’]})tooluniverse – ToolUniverse instance
call_tool – Function to call other tools
- Returns
list: List of tool metadata dictionaries (JSON-compatible)