mrhavens
16cf8cd5ef
Created code_review_crew.py - 5 expert agents: - SecurityExpert - PerformanceExpert - ArchitectureExpert - TestingExpert - ErrorHandlingExpert Each audits specific domain and reports issues. Also creates initial code review findings as issues.
297 lines
10 KiB
Python
297 lines
10 KiB
Python
"""Code Review CrewAI Agents.
|
|
|
|
Five hyperfocused expert agents that rigorously review the Opus codebase.
|
|
Each agent specializes in a specific domain and finds bugs/issues.
|
|
"""
|
|
|
|
from crewai import Agent, Task, Crew, Process
|
|
from dotenv import load_dotenv
|
|
import os
|
|
|
|
load_dotenv()
|
|
|
|
OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY")
|
|
|
|
|
|
# =============================================================================
|
|
# AGENT 1: SECURITY EXPERT
|
|
# =============================================================================
|
|
|
|
security_expert = Agent(
|
|
name="SecurityExpert",
|
|
role="Security & Vulnerability Researcher",
|
|
goal="Find security vulnerabilities, auth issues, data exposure, and unsafe practices",
|
|
backstory="""You are a security researcher with 15 years of experience in application security.
|
|
You specialize in OWASP Top 10, API security, authentication, authorization, and data protection.
|
|
You find vulnerabilities that others miss and provide actionable remediation.
|
|
Your reviews are thorough and your recommendations are precise."""
|
|
)
|
|
|
|
security_task = Task(
|
|
description="""Conduct a comprehensive SECURITY audit of the Opus Orchestrator codebase.
|
|
|
|
Focus on:
|
|
1. Authentication/Authorization - Any security gaps?
|
|
2. Data Exposure - Sensitive data in logs, errors, responses?
|
|
3. Injection Risks - SQL, command, code injection?
|
|
4. API Security - Rate limiting, input validation?
|
|
5. Dependency Vulnerabilities - Known CVEs in requirements?
|
|
6. Secrets Management - API keys, tokens properly handled?
|
|
7. File Access - Path traversal, file inclusion risks?
|
|
8. Input Validation - All user inputs sanitized?
|
|
|
|
Files to review:
|
|
- opus_orchestrator/utils/llm.py
|
|
- opus_orchestrator/server.py
|
|
- opus_orchestrator/cli.py
|
|
- opus_orchestrator/utils/*.py
|
|
- Any authentication/authorization code
|
|
|
|
For each issue found, provide:
|
|
- File and line number
|
|
- Severity (Critical/High/Medium/Low)
|
|
- Description
|
|
- Impact
|
|
- Remediation""",
|
|
agent=security_expert,
|
|
expected_output="Security audit report with findings, severity, and remediation"
|
|
)
|
|
|
|
|
|
# =============================================================================
|
|
# AGENT 2: PERFORMANCE EXPERT
|
|
# =============================================================================
|
|
|
|
performance_expert = Agent(
|
|
name="PerformanceExpert",
|
|
role="Performance & Scalability Architect",
|
|
goal="Find performance bottlenecks, memory issues, inefficient algorithms, and scaling problems",
|
|
backstory="""You are a performance architect with experience optimizing large-scale systems.
|
|
You specialize in profiling, caching strategies, database queries, async patterns, and scalability.
|
|
You find the bottlenecks that cause systems to slow down under load.
|
|
Your analysis includes both immediate issues and architectural concerns."""
|
|
)
|
|
|
|
performance_task = Task(
|
|
description="""Conduct a comprehensive PERFORMANCE audit of the Opus Orchestrator codebase.
|
|
|
|
Focus on:
|
|
1. Async/Await Issues - Blocking calls, missing awaits, thread pool exhaustion?
|
|
2. Memory Leaks - Unclosed resources, growing collections?
|
|
3. Database/API Calls - N+1 queries, redundant calls, missing batching?
|
|
4. Caching Opportunities - Repeated computations that could be cached?
|
|
5. Large Data Handling - Streaming vs loading into memory?
|
|
6. Concurrency Issues - Race conditions, deadlocks?
|
|
7. Algorithmic Complexity - O(n²) where O(n) possible?
|
|
8. Resource Cleanup - Connections, files, threads properly closed?
|
|
|
|
Files to review:
|
|
- opus_orchestrator/langgraph_workflow.py
|
|
- opus_orchestrator/orchestrator.py
|
|
- opus_orchestrator/agents/*.py
|
|
- opus_orchestrator/utils/*.py
|
|
|
|
For each issue:
|
|
- File and line number
|
|
- Severity (Critical/High/Medium/Low)
|
|
- Description
|
|
- Performance Impact
|
|
- Optimization Suggestion""",
|
|
agent=performance_expert,
|
|
expected_output="Performance audit report with findings and optimizations"
|
|
)
|
|
|
|
|
|
# =============================================================================
|
|
# AGENT 3: ARCHITECTURE EXPERT
|
|
# =============================================================================
|
|
|
|
architecture_expert = Agent(
|
|
name="ArchitectureExpert",
|
|
role="Software Architect & Design Patterns Specialist",
|
|
goal="Find architectural weaknesses, design pattern violations, and structural issues",
|
|
backstory="""You are a software architect with expertise in clean code, SOLID principles, and design patterns.
|
|
You specialize in identifying tight coupling, god objects, missing abstractions, and architectural smells.
|
|
Your reviews improve code maintainability and long-term viability."""
|
|
)
|
|
|
|
architecture_task = Task(
|
|
description="""Conduct a comprehensive ARCHITECTURE audit of the Opus Orchestrator codebase.
|
|
|
|
Focus on:
|
|
1. SOLID Violations - Single Responsibility, Open/Closed, Liskov Substitution, Interface Segregation, Dependency Inversion?
|
|
2. Design Patterns - Missing patterns, anti-patterns, over-engineering?
|
|
3. Coupling - Tight coupling, hidden dependencies, circular imports?
|
|
4. Abstraction - Missing abstractions, leaky abstractions?
|
|
5. God Objects - Classes doing too much?
|
|
6. Feature Envy - Classes more interested in other classes' data?
|
|
7. Shotgun Surgery - Changes require many small changes?
|
|
8. Parallel Inheritance - Two class hierarchies that mirror each other?
|
|
9. Lazy Classes - Classes doing almost nothing?
|
|
10. Speculative Generality - Code for "future" features that don't exist?
|
|
|
|
Files to review:
|
|
- opus_orchestrator/orchestrator.py
|
|
- opus_orchestrator/langgraph_workflow.py
|
|
- opus_orchestrator/cli.py
|
|
- opus_orchestrator/agents/base.py
|
|
- opus_orchestrator/nonfiction/*.py
|
|
|
|
For each issue:
|
|
- File and location
|
|
- Principle/violation
|
|
- Description
|
|
- Refactoring Suggestion""",
|
|
agent=architecture_expert,
|
|
expected_output="Architecture audit with violations and refactoring suggestions"
|
|
)
|
|
|
|
|
|
# =============================================================================
|
|
# AGENT 4: TESTING EXPERT
|
|
# =============================================================================
|
|
|
|
testing_expert = Agent(
|
|
name="TestingExpert",
|
|
role="Test Automation & QA Specialist",
|
|
goal="Find missing tests, coverage gaps, and quality issues in test suite",
|
|
backstory="""You are a QA specialist with expertise in test strategy, coverage analysis, and test automation.
|
|
You specialize in identifying what isn't tested, what should be tested, and test quality issues.
|
|
Your recommendations improve confidence in code correctness."""
|
|
)
|
|
|
|
testing_task = Task(
|
|
description="""Conduct a comprehensive TESTING audit of the Opus Orchestrator codebase.
|
|
|
|
Focus on:
|
|
1. Test Coverage - What's NOT tested? Coverage gaps?
|
|
2. Edge Cases - What boundary conditions are untested?
|
|
3. Error Paths - Are exceptions properly tested?
|
|
4. Integration Tests - Do components work together?
|
|
5. Mock Usage - Are mocks overused (hiding bugs)?
|
|
6. Test Quality - Flaky tests, assertions, setup/teardown?
|
|
7. Test Data - Realistic vs mocked data?
|
|
8. Happy Path Bias - Only success cases tested?
|
|
9. Regression Coverage - Can we detect breaking changes?
|
|
10. Performance Tests - Any load/stress testing?
|
|
|
|
Files to review:
|
|
- tests/*.py
|
|
- Any test-related files
|
|
|
|
Also review:
|
|
- Are critical paths in orchestrator tested?
|
|
- Are the agents tested?
|
|
- Is the CLI tested?
|
|
- Are the frameworks tested?
|
|
|
|
For each gap:
|
|
- What should be tested
|
|
- Why it's important
|
|
- Suggested test approach""",
|
|
agent=testing_expert,
|
|
expected_output="Testing audit with coverage gaps and test recommendations"
|
|
)
|
|
|
|
|
|
# =============================================================================
|
|
# AGENT 5: ERROR HANDLING EXPERT
|
|
# =============================================================================
|
|
|
|
error_expert = Agent(
|
|
name="ErrorHandlingExpert",
|
|
role="Exception Handling & Reliability Specialist",
|
|
goal="Find error handling anti-patterns, uncaught exceptions, and reliability issues",
|
|
backstory="""You are a reliability specialist with expertise in exception handling and fault tolerance.
|
|
You specialize in finding swallowed exceptions, improper error messages, and reliability gaps.
|
|
Your reviews make systems more robust and debuggable."""
|
|
)
|
|
|
|
error_task = Task(
|
|
description="""Conduct a comprehensive ERROR HANDLING audit of the Opus Orchestrator codebase.
|
|
|
|
Focus on:
|
|
1. Swallowed Exceptions - try/except with pass or empty except?
|
|
2. Bare Except - except: catching everything?
|
|
3. Error Messages - Generic vs specific, exposing internals?
|
|
4. Logging Issues - Sensitive data in logs? Missing context?
|
|
5. Retry Logic - Failed operations retried properly?
|
|
6. Circuit Breakers - External API failures handled?
|
|
7. Timeout Handling - Long-running operations have timeouts?
|
|
8. Graceful Degradation - What happens when components fail?
|
|
9. Error Recovery - Can the system recover from errors?
|
|
10. Debug Info - Enough info to diagnose issues?
|
|
|
|
Files to review:
|
|
- opus_orchestrator/orchestrator.py
|
|
- opus_orchestrator/langgraph_workflow.py
|
|
- opus_orchestrator/agents/*.py
|
|
- opus_orchestrator/utils/*.py
|
|
- opus_orchestrator/server.py
|
|
|
|
For each issue:
|
|
- File and location
|
|
- Issue type
|
|
- Description
|
|
- Reliability Impact
|
|
- Better Approach""",
|
|
agent=error_expert,
|
|
expected_output="Error handling audit with reliability issues and fixes"
|
|
)
|
|
|
|
|
|
# =============================================================================
|
|
# RUN ALL AGENTS
|
|
# =============================================================================
|
|
|
|
def run_code_review() -> dict:
|
|
"""Run all 5 code review agents."""
|
|
|
|
crews = [
|
|
("Security", security_expert, security_task),
|
|
("Performance", performance_expert, performance_task),
|
|
("Architecture", architecture_expert, architecture_task),
|
|
("Testing", testing_expert, testing_task),
|
|
("Error Handling", error_expert, error_task),
|
|
]
|
|
|
|
results = {}
|
|
|
|
for name, agent, task in crews:
|
|
print(f"\n{'='*60}")
|
|
print(f"Running {name} Expert Review...")
|
|
print('='*60)
|
|
|
|
crew = Crew(
|
|
agents=[agent],
|
|
tasks=[task],
|
|
verbose=True,
|
|
)
|
|
|
|
try:
|
|
result = crew.kickoff()
|
|
results[name] = {
|
|
"status": "success",
|
|
"findings": result,
|
|
}
|
|
except Exception as e:
|
|
results[name] = {
|
|
"status": "error",
|
|
"error": str(e),
|
|
}
|
|
|
|
return results
|
|
|
|
|
|
if __name__ == "__main__":
|
|
results = run_code_review()
|
|
|
|
print("\n" + "="*60)
|
|
print("CODE REVIEW SUMMARY")
|
|
print("="*60)
|
|
|
|
for name, result in results.items():
|
|
status = result["status"]
|
|
emoji = "✅" if status == "success" else "❌"
|
|
print(f"{emoji} {name}: {status}")
|