python-modal

## Quick Reference

| Pattern | Decorator | Use Case |
|---------|-----------|----------|
| Simple function | `@app.function()` | Stateless compute |
| Class with state | `@app.cls()` | ML models, DB connections |
| Web endpoint | `@modal.asgi_app()` | FastAPI/Flask APIs |
| Scheduled | `@app.function(schedule=...)` | Cron jobs |
| Parallel | `.map()`, `.starmap()` | Batch processing |

## When to Use This Skill

Use for **Python on Modal**:
- Type-safe serverless functions with Pydantic
- Async/await patterns for concurrent operations
- GPU workloads (ML inference, training)
- FastAPI web endpoints
- Scheduled tasks and automation
- Testing Modal functions locally

---

# Modern Python on Modal.com (2025)

Complete guide to building type-safe, production-ready Python applications on Modal's serverless platform.

## Type-Safe Modal Functions

### Basic Function with Type Hints

```python
import modal

app = modal.App("typed-example")

image = modal.Image.debian_slim(python_version="3.11").pip_install("pydantic")

@app.function(image=image)
def process_data(
    items: list[str],
    multiplier: int = 1,
) -> dict[str, int]:
    """Process data with full type hints."""
    return {item: len(item) * multiplier for item in items}


@app.local_entrypoint()
def main():
    result: dict[str, int] = process_data.remote(["hello", "world"], multiplier=2)
    print(result)  # {'hello': 10, 'world': 10}
```

### Pydantic Models for Validation

```python
import modal
from pydantic import BaseModel, Field, field_validator
from datetime import datetime
from typing import Literal

app = modal.App("pydantic-example")

image = modal.Image.debian_slim(python_version="3.11").pip_install("pydantic>=2.0")


# === Input/Output Models ===

class ProcessingConfig(BaseModel):
    """Configuration with validation."""
    model_name: str = Field(..., min_length=1, max_length=100)
    batch_size: int = Field(32, ge=1, le=256)
    temperature: float = Field(0.7, ge=0.0, le=2.0)
    mode: Litera