Skip to content

README.md

Latest commit

 

History

History
159 lines (115 loc) · 4.54 KB

README.md

File metadata and controls

159 lines (115 loc) · 4.54 KB

AIDLC Code Reviewer

Automated, language-agnostic code quality analysis for code assets generated through the AI Development Lifecycle (AIDLC). Runs static analysis tools, AI-powered critical code analysis, and business logic review — then produces structured HTML and Markdown reports.

Quick Start

cd scripts/aidlc-codereview

# Install with uv (recommended)
uv sync

# Or install with pip
pip install -e .

# Run a review
aidlc-code-reviewer ./path/to/code

Prerequisites

  • Python 3.11+
  • AWS credentials with Amazon Bedrock access (for AI-powered analysis)
  • Java JDK 17+ (optional, for Java tools)

See docs/SETUP.md for full installation and configuration details.

What It Does

The tool runs in two parallel tracks:

Technical Analysis — static tools + AI critique:

  • Security scanning (bandit, semgrep, gitleaks)
  • Linting and type checking (ruff, mypy, checkstyle, eslint)
  • Complexity, duplication, and dead code analysis
  • AI-powered critical code findings (COMPUTATION, CONTROL_FLOW, DATA_TRANSFORM)
  • Code structure critique across 6 dimensions (logging, scalability, efficiency, complexity, measurability, structure)

Business Logic Analysis — AI-driven domain review:

  • Identifies business rules, formulas, pricing logic, state machines, and 10 other domain categories
  • Flags every finding for human review regardless of tool results
  • Cross-checks for self-consistency (constant drift, logic divergence, naming mismatches)

Reports

Three report files are generated:

Report Format Purpose
code_review_summary_*.html HTML Entry point — two cards linking to detailed reports
code_review_technical_*.html/.md HTML + Markdown Static tool findings, critical code sections, structure critique
code_review_business_*.html/.md HTML + Markdown Business logic findings, consistency issues

Open the summary HTML first — it tells you where to start.

Usage

# Default: both reports
aidlc-code-reviewer <target>

# Technical report only
aidlc-code-reviewer <target> --technical-report

# Business logic report only
aidlc-code-reviewer <target> --business-report

# Custom output directory
aidlc-code-reviewer <target> -o ./my-reports

# Skip auto-generation of missing tool wrappers
aidlc-code-reviewer <target> --no-generate

# Pre-flight check (verify AWS credentials and Bedrock access)
aidlc-code-reviewer --preflight

# Verbose output
aidlc-code-reviewer <target> -v

Configuration

config/review-config.yaml

Defines which tools to run. Just list tool names:

tools:
  - bandit
  - ruff
  - mypy
  - semgrep

The agent auto-generates a wrapper for each tool on first run (requires Amazon Bedrock access). Generated wrappers are cached in src/code_reviewer/tools/ for subsequent runs.

config/agent-config.yaml

Controls the Amazon Bedrock model and AWS settings:

agent:
  model_id: "us.anthropic.claude-sonnet-4-6"
  max_tokens: 16384

aws:
  region: "us-east-1"

Environment variables AWS_REGION, AWS_PROFILE, and BEDROCK_MODEL_ID override the YAML values.

Project Structure

scripts/aidlc-codereview/
├── src/
│   └── code_reviewer/
│       ├── __init__.py             # Package init, project root constants
│       ├── runner.py               # CLI orchestration
│       ├── agent/                  # AI agents (critical findings, structure, business logic)
│       ├── common/                 # Shared utilities, models, report generation
│       └── tools/                  # Tool registry and auto-generated wrappers
├── config/
│   ├── agent-config.yaml           # Agent/Bedrock configuration
│   ├── review-config.yaml          # Tool configuration
│   └── prompts/                    # System prompts for AI agents
├── docs/
│   └── SETUP.md                    # Detailed setup guide
├── pyproject.toml
├── README.md
├── CHANGELOG.md
├── LICENSE
└── NOTICE

Documentation

See docs/SETUP.md for:

  • AWS credential configuration (IAM roles, SSO, profiles)
  • Amazon Bedrock model access setup
  • CLI usage and flags
  • How auto-generation works
  • Troubleshooting

License

MIT-0 (MIT No Attribution)