File size: 6,752 Bytes

8c10cf2

# Full reference trajectory configuration - SQLAssistant-A2A project
#
# This configuration is ported from the SQLAssistant-MCP version with the core kept consistent:
# - It evaluates the same Agent / LLM / Tool structure
# - It uses the same 6 trajectory metrics + path diversity metrics
# Only the trajectory hierarchy is adapted to the actual A2A execution path:
# - Top-level SPAN: sql_assistant_workflow_a2a
# - Second-level SPAN: a2a_execution
# - Then three stage SPANs: generate_sql → check_compliance → interpret_results
# - Intermediate framework SPANs such as a2a_call_xxx / xxx_server_execution are also required steps in the reference trajectory

# Project name (determines where to collect traces from RESULTS/<model>/<project_name>/test_results)
project_name: "SQLAssistant-A2A"

# Trajectory extraction types: same as the MCP version; extracts a multi-layer structure
extract_types:
  - "SPAN" # Multi-level SPAN structure (including *_a2a, a2a_execution, generate_sql, etc.)
  - "Chain" # Kickoff chain for each Crew
  - "Agent" # Agent for each Crew
  - "LLM" # LLM calls
  - "Tool" # Tool calls

# The A2A version also has no chapter-style repeatable patterns
repeatable_patterns: []

# ==========================================================================
# Reference trajectory (Ground Truth) - A2A architecture, based on code design + trace examples
# ==========================================================================
#
# Ideal execution path hierarchy (no retries, no errors):
#
# [SPAN] sql_assistant_workflow_a2a
# └─ [SPAN] a2a_execution
#    ├─ [SPAN] generate_sql
#    │  └─ [SPAN] a2a_call_sql_generation
#    │     └─ [SPAN] sql_generation_server_execution
#    │        └─ [Chain] Crew***.kickoff
#    │           ├─ [AGENT] Expert SQL Query Generator
#    │           │  ├─ [LLM] *
#    │           │  ├─ [Tool] get_database_schema
#    │           │  └─ [LLM] *
#    │           └─ [AGENT] Senior SQL Code Reviewer
#    │              ├─ [LLM] *
#    │              ├─ [Tool] validate_sql_syntax
#    │              ├─ [LLM] *
#    │              ├─ [Tool] get_database_schema
#    │              └─ [LLM] *
#    ├─ [SPAN] check_compliance
#    │  └─ [SPAN] a2a_call_compliance_checker
#    │     └─ [SPAN] compliance_checker_server_execution
#    │        └─ [Chain] Crew***.kickoff
#    │           └─ [AGENT] Data Security and Compliance Auditor
#    │              ├─ [LLM] *
#    │              ├─ [Tool] get_database_schema
#    │              └─ [LLM] *
#    └─ [SPAN] interpret_results
#       └─ [SPAN] a2a_call_result_interpreter
#          └─ [SPAN] result_interpreter_server_execution
#             └─ [Chain] Crew***.kickoff
#                └─ [AGENT] Business Intelligence Analyst
#                   ├─ [LLM] *
#                   ├─ [Tool] run_sql_query
#                   └─ [LLM] *
#
# Notes:
# - In real A2A traces, framework-level SPANs such as [SPAN] a2a_call_sql_generation and
#   [SPAN] sql_generation_server_execution are inserted between stages;
#   these intermediate SPANs are now required matching steps.
# - Business retries (business_retry N) and orchestrator-level retries (retry N) are not written into the reference trajectory;
#   during evaluation, extra steps and retries are allowed by the in_order / any_order metrics.

reference_trajectory:
  # ===== Top-level SPAN (A2A top-level workflow) =====
  - "SPAN: sql_assistant_workflow_a2a"

  # ===== Second-level SPAN (A2A execution wrapper) =====
  - "SPAN: a2a_execution"

  # ===== Stage 1: Generate SQL (SQLGenerationCrew) =====
  - "SPAN: generate_sql"
  - "SPAN: a2a_call_sql_generation"
  - "SPAN: sql_generation_server_execution"
  - "Chain: Crew***.kickoff"
  # Agent 1: SQL Generator (uses key schema exploration tools)
  - "Agent: Expert SQL Query Generator"
  - "LLM: *" # Decide overall generation strategy
  - "Tool: get_database_schema" # Read full schema
  - "LLM: *" # Generate candidate SQL based on schema
  # Other exploration tools (list_tables / get_table_sample / get_column_stats) are considered optional

  # Agent 2: SQL Reviewer (syntax and structure checks)
  - "Agent: Senior SQL Code Reviewer"
  - "LLM: *" # Read and understand the generated SQL
  - "Tool: validate_sql_syntax" # Required: syntax validation
  - "LLM: *" # Analyze syntax validation results
  - "Tool: get_database_schema" # Required: verify tables/columns exist based on schema
  - "LLM: *" # Revise SQL based on schema checks

  # ===== Stage 2: Check Compliance (ComplianceCheckerCrew) =====
  - "SPAN: check_compliance"
  - "SPAN: a2a_call_compliance_checker"
  - "SPAN: compliance_checker_server_execution"
  - "Chain: Crew***.kickoff"
  - "Agent: Data Security and Compliance Auditor"
  - "LLM: *" # Initial read of SQL and identify potential risks
  - "Tool: get_database_schema" # Required: assess sensitivity with schema context
  - "LLM: *" # Provide PASS/FAIL decision

  # ===== Stage 3: Interpret Results (ResultInterpreterCrew) =====
  # Note: only entered when compliance passes
  - "SPAN: interpret_results"
  - "SPAN: a2a_call_result_interpreter"
  - "SPAN: result_interpreter_server_execution"
  - "Chain: Crew***.kickoff"
  - "Agent: Business Intelligence Analyst"
  - "LLM: *" # Understand the business question and SQL
  - "Tool: run_sql_query" # Required: execute SQL to get real data
  - "LLM: *" # Produce business interpretation based on results (count_rows is an optional stats tool)

# Target tool list (used by the single_tool_use metric)
# Consistent with the MCP version; covers all core tools (including some optional tools)
target_tools:
  # SQL Generation stage - Generator
  - "Tool: get_database_schema"
  - "Tool: get_table_sample"
  - "Tool: get_column_stats"
  - "Tool: list_tables"
  # SQL Generation stage - Reviewer
  - "Tool: validate_sql_syntax"
  - "Tool: check_table_exists"
  # Compliance stage (shares get_database_schema / check_table_exists; do not duplicate)
  # Result Interpretation stage
  - "Tool: run_sql_query"
  - "Tool: count_rows"

# Models to evaluate (same as the MCP version; only the path changes to SQLAssistant-A2A)
models:
  - "GPT-5"
  - "GPT-4o-mini"
  - "DeepSeek-V3-1"
  - "DeepSeek-R1"
  - "Gemini-2.5-flash"
  - "Gemini-2.5-flash-nothinking"
  - "Qwen3-235b"
# Usage (brief):
# - Run from: RQ-Failure_Breakdown/SQLAssistant-A2A
#     python3 evaluate_trajectory-Filter_Tools.py --config reference_trajectory.yaml
# - evaluate_trajectory-Filter_Tools.py will automatically read execution_path.md from
#   RESULTS/<model>/SQLAssistant-A2A/test_results based on project_name.