critiq 1.2.0

AI-powered local code reviewer — catch issues before you push

critiq

AI-powered local code reviewer — catch issues before you push.

critiq reads your git diff and runs an AI review before you push. It flags security vulnerabilities, bugs, and performance issues with severity ratings so you can fix what matters most.

$ critiq
Reviewing staged changes · +28/-6 lines · src/auth.py, src/db.py

┌─────────────────── Summary ────────────────────────────────┐
│ 🚨 Needs work                                              │
│                                                            │
│ The change introduces direct SQL string interpolation,     │
│ a critical security vulnerability.                         │
└────────────────────────────────────────────────────────────┘

┌── 🚨 [CRITICAL] SQL Injection vulnerability ──────────────┐
│ src/auth.py  L42  security                                 │
│                                                            │
│ **Issue:** User input interpolated into SQL string         │
│ **Fix:** Use parameterized queries:                        │
│   db.execute("SELECT * FROM users WHERE name=?", (user,)) │
└───────────────────────────────────────────────────────────┘

┌── ⚠️ [WARNING] Missing input validation ──────────────────┐
│ src/db.py  L15  correctness                                │
│                                                            │
│ **Issue:** `user_id` can be None; no null check before use │
│ **Fix:** Add `if user_id is None: raise ValueError(...)` │
└───────────────────────────────────────────────────────────┘

Install

pip install critiq

Set your API key (or use Ollama for zero-cost local review):

export ANTHROPIC_API_KEY=your-key   # Claude (default)
export OPENAI_API_KEY=your-key      # or OpenAI
# or use --provider ollama           # local, no API key needed

Usage

# Review staged changes (most common — run before git push)
critiq

# Review and interactively fix issues
critiq --fix

# Review and automatically apply all fixes (no prompts)
critiq --fix-all

# Watch mode: auto-review when staged files change
critiq --watch

# Review all changes vs main branch
critiq --diff main

# Review vs main, then fix what's found
critiq --diff main --fix

# Review a specific file
critiq --file src/auth.py

# Focus on a specific concern
critiq --focus security
critiq --focus performance
critiq --focus readability
critiq --focus correctness

# Only show critical issues
critiq --severity critical

# Compact output (good for scripts/CI)
critiq --compact

# Add context for the AI reviewer
critiq --context "This module handles payments — be strict about error handling"

# Use local Ollama (no API key)
critiq --provider ollama --model llama3.2

# Use OpenAI
critiq --provider openai --model gpt-4o

Auto-Fix (v1.0)

critiq --fix closes the review loop: find issues and fix them in one command.

$ critiq --fix

🚨 [CRITICAL] SQL Injection in login()      src/auth.py  line 6
🚨 [CRITICAL] Plaintext Password Storage    src/auth.py  line 22
⚠️  [WARNING]  Weak Token Generation         src/auth.py  line 10

Fix 3 issue(s) in src/auth.py? (a=fix all / s=select / n=skip)  > a

Generating fix... ✓

╭─── Changes to src/auth.py ─────────────────────────────────────────╮
│ - query = f"SELECT * FROM users WHERE name='{username}'"           │
│ + query = "SELECT * FROM users WHERE name=? AND password=?"        │
│ + db.execute(query, (username, password))                           │
│                                                                     │
│ - return {"token": hashlib.md5(username.encode()).hexdigest()}      │
│ + return {"token": secrets.token_urlsafe(32)}                      │
╰─────────────────────────────────────────────────────────────────────╯

Apply this fix? [Y/n]  y
✅ Applied  (backup: src/auth.py.critiq.bak)

How it works:

1. critiq reviews your diff and finds issues
2. For each file with CRITICAL/WARNING issues, it asks: fix all / select / skip
3. The AI reads the full file + all issues and generates a fixed version
4. You see a colorized diff before applying
5. Original files are backed up as .critiq.bak
6. Run git diff to review all changes before committing

Flags:

• --fix — interactive mode (prompts for each file)
• --fix-all — auto-apply all fixes without prompting
• --fix-severity warning — also fix WARNING issues (default: CRITICAL + WARNING)

Watch Mode (v1.1)

critiq --watch monitors your working directory and automatically re-runs a review every time your staged files change.

# Start watch mode: reviews run automatically when you git add
critiq --watch

# Watch with a specific focus
critiq --watch --focus security

# Adjust re-run delay (default: 2s after last change)
critiq --watch --debounce 5

# For faster file detection (optional):
pip install 'critiq[watch]'   # adds watchfiles for inotify/FSEvents support

This is useful when you're iterating on a feature: stage your changes and immediately get feedback, without leaving the terminal.

Project Preferences (critiq-learn)

Teach critiq your project's conventions so it stops flagging things you don't care about — and always checks the things you do.

# Don't flag these (project uses JS, type hints not required)
critiq-learn ignore "Missing type annotations"
critiq-learn ignore "No docstrings on private methods"

# Always check these extra rules
critiq-learn rule "Never use raw SQL strings — always use parameterized queries"
critiq-learn rule "All API endpoints must have rate limiting"

# Set project defaults
critiq-learn set focus security         # always focus on security
critiq-learn set provider ollama        # use local LLM by default

# Show current config
critiq-learn show

# Remove an ignore rule
critiq-learn unignore "Missing type annotations"

# Reset everything
critiq-learn reset

Preferences are saved to .critiq.yaml in your project root. critiq automatically picks it up on every run:

# .critiq.yaml (example)
ignore_patterns:
  - Missing type annotations
  - No docstrings on private methods
custom_rules:
  - Never use raw SQL strings — always use parameterized queries
default_focus: security

Add .critiq.yaml to git to share project preferences with your team.

Focus Areas

Flag	Reviews
--focus all	Everything (default)
--focus security	SQL injection, auth, XSS, SSRF, secrets exposure
--focus performance	N+1 queries, blocking I/O, inefficient algorithms
--focus correctness	Logic bugs, null handling, edge cases, race conditions
--focus readability	Naming, complexity, dead code, missing docs
--focus style	Formatting, conventions, unused imports

Severity Levels

Level	Meaning
🚨 CRITICAL	Must fix before merging (critiq exits with code 1)
⚠️ WARNING	Should fix
ℹ️ INFO	Consider fixing
💡 SUGGESTION	Nice to have

critiq exits with code 1 if any CRITICAL issues are found, making it easy to use in pre-push hooks or CI.

Pre-push Hook

Add to .git/hooks/pre-push to automatically review before every push:

#!/bin/sh
critiq --diff origin/main --severity critical --compact

chmod +x .git/hooks/pre-push

Now every git push automatically runs a security review. The push is blocked only if CRITICAL issues are found.

GitHub Actions (CI)

Add AI code review to every pull request with critiq-action:

# .github/workflows/critiq.yml
name: critiq Code Review
on:
  pull_request:
    branches: [main, master]
permissions:
  pull-requests: write
jobs:
  review:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0
      - uses: faw21/critiq-action@v1
        with:
          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}

See critiq-action for full configuration options.

Providers

Provider	Command	Notes
Claude (default)	--provider claude	Best results; requires ANTHROPIC_API_KEY
OpenAI	--provider openai	Requires OPENAI_API_KEY
Ollama	--provider ollama	Free, runs locally; no API key needed

# Default model per provider
critiq --provider claude    # claude-opus-4-6
critiq --provider openai    # gpt-4o
critiq --provider ollama    # llama3.2

# Custom model
critiq --provider claude --model claude-haiku-4-5-20251001  # faster + cheaper
critiq --provider ollama --model codellama

Developer Workflow Integration

critiq fits into the AI-powered git workflow:

# 1. Morning: generate standup from yesterday's commits
standup-ai ~/projects/myapp

# 2. Write code, then review before committing
critiq                          # AI review of staged changes
git add -p                      # stage what looks good

# 3. Generate conventional commit message
gpr --commit-run

# 4. Pack codebase context for LLM-assisted PR review
gitbrief . --budget 8000 --clipboard

# 5. Generate PR description
gpr

# 6. Review a teammate's PR
prcat 42                        # AI review of their changes

# 7. At release: generate CHANGELOG
changelog-ai --from v0.1.0 --prepend CHANGELOG.md

Related Tools

• critiq-action — GitHub Action: run critiq in CI on every PR
• gitbrief — git-history-aware context packer for LLMs
• gpr — AI commit messages + PR descriptions
• standup-ai — daily standup from git commits
• changelog-ai — AI-generated CHANGELOG
• prcat — AI reviewer for teammates' pull requests
• git-chronicle — AI git history narrator (understand WHY code changed)

License

MIT