dinocheck 0.0.2

Your vibe coding companion - LLM-powered code critic for AI-assisted development

Dinocheck

Your vibe coding companion - LLM-powered code critic

---

Dinocheck is an AI-powered code critic designed to enhance your vibe coding sessions. It's not a traditional linter - those focus on syntax and style. Dinocheck uses GPT, Claude, or local models to understand your code semantically and provide intelligent feedback on the things that matter: logic bugs, security issues, and architectural problems.

$ dino check src/views.py

src/views.py:42 [major] django/n-plus-one
  N+1 query detected: iterating over `Order.objects.filter(user=user)` and
  accessing `order.items.all()` inside the loop causes one query per order.

  → Use `prefetch_related('items')` to fetch all items in a single query.

src/views.py:87 [critical] django/missing-permission-check
  The `delete_account` view modifies user data but has no permission check.
  Any authenticated user could delete any account by guessing the ID.

  → Add ownership validation: `if account.user != request.user: return 403`

✗ 2 issues found (1 critical, 1 major)

Why Dinocheck?

Traditional linters catch syntax errors and style issues. Dinocheck catches logic bugs, security issues, and architectural problems that only an AI can understand:

• Detects N+1 queries that would kill your database
• Spots missing authorization checks before they become CVEs
• Finds race conditions in your async code
• Identifies test mocks that hide real bugs

Philosophy

Dinocheck is a linter, not a fixer. It's designed to be your coding companion:

• Reviews your code with LLM intelligence
• Points out issues and explains why they matter
• Lets you decide how to fix them

This fits the vibe coding workflow: you write code with AI assistance, and Dinocheck provides a second opinion.

Features

Feature	Description
LLM-First Analysis	Uses GPT-4, Claude, or local models for semantic code review
Rule Packs	Python and Django packs with 40+ rules
Smart Caching	SQLite cache avoids re-analyzing unchanged files
Cost Tracking	Monitor LLM usage and costs with dino logs
Multi-Language	Get feedback in English, Spanish, French, etc.
100+ Providers	OpenAI, Anthropic, Ollama, and more via LiteLLM

Quick Start

Installation

pip install dinocheck
# or with uv
uv add dinocheck

Configuration

Initialize your project and configure your LLM provider:

# Create dino.yaml with interactive setup
dino init

# Set your API key
export OPENAI_API_KEY=sk-...

The dino init command creates a dino.yaml file where you can customize your LLM provider and packs:

dinocheck:
  packs:
    - python
    - django

  provider:
    model: gpt-4o-mini          # or claude-3-5-sonnet, ollama/llama3
    api_key_env: OPENAI_API_KEY

  output:
    language: en

Usage

# Analyze current directory
dino check

# Analyze specific files
dino check src/views.py src/models.py

# Only analyze changed files (git diff)
dino check --diff

# Output as JSON
dino check --format json

# View LLM costs
dino logs cost

CLI Reference

Command	Description
dino check [paths]	Analyze code with LLM
dino check --diff	Analyze only changed files
dino packs list	List available packs
dino packs info NAME	Show pack details
dino explain RULE_ID	Explain a rule
dino cache stats	Show cache statistics
dino cache clear	Clear the cache
dino logs list	View LLM call history
dino logs show ID	Show details of a specific LLM call
dino logs cost	View cost summary
dino init	Create dino.yaml
dino version	Show version information

Rule Packs

Python Pack

Category	Rules
Security	SQL injection, insecure deserialization, mass assignment, timing attacks
Logic	Inverted conditions, unreachable code, copy-paste bugs, missing edge cases
Code Quality	Naming mismatches, misleading comments, stale TODOs
Concurrency	Race conditions, deadlocks, resource lifecycle
Testing	Wrong assertions, flaky patterns, missing negative tests, mocks hiding bugs

Django Pack

Category	Rules
ORM	N+1 queries, missing select_related, queryset performance, template queries, signal hidden logic, cache stale data
Transactions	Atomic scope, select_for_update, F-expressions, on_commit side effects
DRF	Permission classes, throttling, async blocking calls, serializer mismatches
Migrations	Data loss, large indexes, NOT NULL two-phase
Tasks	Non-idempotent Celery tasks
Testing	Missing auth tests, business logic coverage

Output Formats

Format	Use Case
text	Colored terminal output (default)
json	Full JSON for tooling integration
jsonl	JSON Lines for streaming

GitHub Actions Integration

name: Dinocheck
on: [pull_request]

jobs:
  review:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: astral-sh/setup-uv@v5

      - run: uv add dinocheck
      - run: uv run dino check --diff
        env:
          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}

Supported LLM Providers

Provider	Model Examples
OpenAI	gpt-4o, gpt-4o-mini, o1-preview
Anthropic	claude-3-5-sonnet, claude-3-opus
Ollama	ollama/llama3, ollama/codellama
Azure	azure/gpt-4o
Google	gemini/gemini-pro

See LiteLLM docs for 100+ supported providers.

Development

# Clone and install
git clone https://github.com/diegogm/dinocheck.git
cd dinocheck
uv sync --dev

# Run tests
fab test

# Run linters
fab lint

# Pre-deployment checks
fab predeploy

License

MIT License - see LICENSE for details.

---

Made with care by the Dinocheck contributors