pyproject.toml

[project]
name = "ai-code-reviewer"
version = "0.1.0"
description = "An AI code review agent that reviews pull requests like a senior engineer."
readme = "README.md"
requires-python = ">=3.11"
dependencies = [
    # OpenAI Python SDK — used as a generic OpenAI-compatible client.
    # Talks to Gemini (free tier, default), OpenAI, Groq, or local Ollama
    # by changing LLM_BASE_URL — no code changes required.
    "openai>=1.50.0",
    # Data validation. Every Finding and Review is a pydantic model so the
    # LLM's output gets type-checked before we trust it.
    "pydantic>=2.8.0",
    # pydantic-settings — loads config from .env cleanly
    "pydantic-settings>=2.5.0",
    # YAML reader for eval cases and configs
    "pyyaml>=6.0",
    # Rich terminal output (pretty JSON, tables). Purely a DX nice-to-have.
    "rich>=13.7.0",
    # --- Week 2 additions ---
    # FastAPI = our webhook server. Async-capable, type-safe, OpenAPI for free.
    "fastapi>=0.115.0",
    # Uvicorn = ASGI server that actually runs FastAPI. [standard] pulls in
    # the optional perf extras (httptools, uvloop on Linux).
    "uvicorn[standard]>=0.30.0",
    # SQLAlchemy 2.0 = ORM + connection pool. We use sync mode for v1 — the
    # LLM call inside the worker is sync, so introducing async DB sessions
    # would mean juggling two concurrency models in one file for no benefit.
    "sqlalchemy>=2.0.30",
    # httpx = HTTP client. Two uses: (a) FastAPI's TestClient depends on it,
    # (b) we'll use it next session to fetch PR diffs from the GitHub API.
    "httpx>=0.27.0",
]

[project.optional-dependencies]
dev = [
    "pytest>=8.0",
    "pytest-asyncio>=0.23",
    "ruff>=0.6",
    "mypy>=1.10",
]

[project.scripts]
# `reviewer --diff ... --repo ...` — week 1 CLI
reviewer = "reviewer.cli:main"
# `reviewer-server` — boots the FastAPI webhook receiver on :8000
reviewer-server = "reviewer.api.app:run"

[build-system]
requires = ["hatchling"]
build-backend = "hatchling.build"

# Tell Hatchling which directory holds the Python package.
# Without this it looks for a folder matching the project name
# (ai_code_reviewer/), but our package folder is named `reviewer/`.
[tool.hatch.build.targets.wheel]
packages = ["reviewer"]

[tool.ruff]
line-length = 100
target-version = "py311"

[tool.pytest.ini_options]
testpaths = ["tests"]