aggregation_config.toml

# Multi-Signal Responsibility Classification Configuration
#
# This configuration file defines the weights and strategies used for
# aggregating multiple signals to classify function responsibilities.

[weights]
# Signal weights must sum to 1.0
#
# I/O Detection (35%): Primary signal for identifying I/O operations
io_detection = 0.35

# Call Graph Analysis (25%): Identifies orchestration and coordination patterns
call_graph = 0.25

# Type Signatures (15%): Infers responsibility from parameter and return types
type_signatures = 0.15

# Name Heuristics (15%): Uses function names to infer responsibility
name_heuristics = 0.15

# Purity Analysis (5%): Identifies pure computation functions
purity_side_effects = 0.05

# Framework Patterns (5%): Detects framework-specific patterns
framework_patterns = 0.05

[conflict_resolution]
# Strategy to use when signals disagree
# Options: "weighted_voting", "framework_first", "io_first", "highest_confidence"
strategy = "weighted_voting"

# Minimum confidence threshold for classification (0.0 - 1.0)
minimum_confidence = 0.30

# Threshold for framework pattern override (0.0 - 1.0)
# When a framework signal has confidence >= this threshold, it overrides other signals
framework_override_threshold = 0.70

[accuracy]
# Target accuracy goal
target_accuracy = 0.85

# Baseline accuracy (name-based classification alone)
baseline_accuracy = 0.50

[tuning]
# Enable automatic weight tuning based on ground truth corpus
auto_tune = false

# Path to ground truth corpus for validation
ground_truth_path = "tests/responsibility_ground_truth.json"

# Optimization metric: "accuracy", "precision", "recall", "f1"
optimization_metric = "accuracy"

[explainability]
# Include evidence from all signals in classification results
include_evidence = true

# Show alternative classifications with their scores
show_alternatives = true

# Maximum number of alternative classifications to show
max_alternatives = 2

[performance]
# Enable parallel processing for multi-signal analysis
enable_parallel = true

# Maximum performance overhead target (as percentage)
max_overhead_percent = 3.0

[categories]
# Map of responsibility categories and their descriptions
# These are used for documentation and output formatting

[categories.io]
FileIO = "File system I/O operations"
NetworkIO = "Network communication and HTTP requests"
DatabaseIO = "Database queries and transactions"
ConfigurationIO = "Configuration file loading and parsing"
MixedIO = "Multiple types of I/O operations"

[categories.handlers]
HttpRequestHandler = "HTTP request handler (web framework)"
WebSocketHandler = "WebSocket connection handler"
CliHandler = "Command-line interface handler"
DatabaseHandler = "Database connection handler"

[categories.computation]
PureComputation = "Pure mathematical or logical computation"
Validation = "Input validation and verification"
Transformation = "Data transformation between formats"
Parsing = "Parsing text or structured data"
Formatting = "Formatting data for output"

[categories.coordination]
Orchestration = "High-level workflow orchestration"
Coordination = "Coordination between components"
ErrorHandling = "Error handling and recovery"

[categories.testing]
TestFunction = "Unit or integration test function"

[categories.other]
SideEffects = "Functions with side effects"
Unknown = "Unable to classify responsibility"