data/processed/RQ1/SQLAssistant-H_A2A/README.md

SQLAssistant-H_A2A Evaluation Scripts

This directory contains evaluation scripts for the SQLAssistant-H_A2A project, which analyzes the execution trajectories and performance metrics of different AI models.

Files

Configuration

• reference_trajectory.yaml - Reference trajectory configuration defining the expected execution path for the SQLAssistant workflow

Evaluation Scripts

• evaluate_trajectory-Filter_Tools-mix.py - Main trajectory evaluation script that computes 6 trajectory metrics
• evaluate_success.py - Success rate analysis based on required tool execution
• evaluate_scores.py - Score statistics collection from execution logs
• analyze_retry_patterns.py - RETRY pattern analysis including business retries and orchestrator retries

Trajectory Metrics

The evaluation framework computes the following 6 metrics:

1. Exact Match - Predicted trajectory must exactly match the reference trajectory
2. In-Order Match - Reference trajectory must be a subsequence of predicted trajectory
3. Any-Order Match - Predicted trajectory must contain all necessary actions (order-agnostic)
4. Precision - Ratio of correct actions in predicted trajectory
5. Recall - Ratio of reference actions covered by predicted trajectory
6. Single-Tool Use - Usage rate of specific tools

Additional diversity metrics:

• Unique Path Ratio - Number of unique trajectories / total samples
• Path Entropy - Shannon entropy of trajectory distribution

Usage

Trajectory Evaluation

python3 evaluate_trajectory-Filter_Tools-mix.py --config reference_trajectory.yaml

Options:

• --config - Path to reference trajectory configuration file (YAML format)
• --base-dir - RESULTS directory path (defaults to two levels up from script location)
• --output - Output CSV file path
• --diagnose-failures - Diagnose any_order_match failure reasons and generate report

Success Rate Analysis

python3 evaluate_success.py

Success criterion: Whether the get_database_schema tool was executed (required for understanding database structure).

Score Statistics

python3 evaluate_scores.py

Collects and analyzes scores from execution logs.

Retry Pattern Analysis

python3 analyze_retry_patterns.py

Analyzes two types of retries:

1. Orchestrator RETRY - Framework-level retries (e.g., [SPAN] crew_execution (retry N))
2. BUSINESS-RETRY - Business logic retries (e.g., [SPAN] generate_sql (business_retry N))

Models Evaluated

• GPT-5
• GPT-4o-mini
• DeepSeek-V3-1
• DeepSeek-R1
• Gemini-2.5-flash
• Gemini-2.5-flash-nothinking
• Qwen3-235b

Project Structure

The SQLAssistant-H_A2A project uses a hybrid architecture combining:

• CrewAI for SQL generation stage
• LangGraph for compliance checking stage
• AutoGen for result interpretation stage

Workflow Stages

1. Generate SQL - CrewAI-based SQL generation with two agents:

   • Expert SQL Query Generator
   • Senior SQL Code Reviewer

2. Check Compliance - LangGraph orchestration for compliance validation

3. Interpret Results - AutoGen-based result interpretation and SQL execution

Output Files

• evaluation_results.csv - Trajectory evaluation results for all models
• success_rate.csv - Success rate summary
• success_detailed_results.json - Detailed success analysis
• score_summary.csv - Score statistics summary
• score_analysis.json - Detailed score analysis
• retry_summary.csv - Retry pattern summary
• retry_analysis.json - Detailed retry analysis
• error_by_agent.csv - Error statistics by agent
• business_retry_chapters.csv - Business retry location statistics
• any_order_match_failures.csv - Failure diagnosis (if enabled)

Requirements

• Python 3.x
• pandas
• PyYAML
• pathlib
• collections
• json
• csv
• re
• math