Skip to content

Deploy to Production

Deploy optimized prompters in Docker, FastAPI, and serverless environments.

FastAPI Service

from fastapi import FastAPI
from pydantic import BaseModel
from dspydantic import Prompter
import dspy

app = FastAPI()

# Load once at startup
dspy.configure(lm=dspy.LM("openai/gpt-4o-mini"))
prompter = Prompter.load(
    "./my_prompter",
    model=MyModel,
    model_id="openai/gpt-4o-mini"
)

class ExtractionRequest(BaseModel):
    text: str

class ExtractionResponse(BaseModel):
    data: dict
    confidence: float

@app.post("/extract")
async def extract(request: ExtractionRequest):
    result = prompter.run(request.text)
    return ExtractionResponse(
        data=result.dict(),
        confidence=1.0
    )

Docker

FROM python:3.10-slim

WORKDIR /app

# Install dependencies
COPY requirements.txt .
RUN pip install -r requirements.txt

# Copy prompter
COPY my_prompter ./my_prompter

# Copy app
COPY app.py .

# Set API key from environment
ENV OPENAI_API_KEY=${OPENAI_API_KEY}

# Run
CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "8000"]

# Build
docker build -t my-prompter-api .

# Run
docker run -e OPENAI_API_KEY=sk-... -p 8000:8000 my-prompter-api

AWS Lambda

import json
import dspy
from dspydantic import Prompter

# Load at cold start (global scope)
dspy.configure(lm=dspy.LM("openai/gpt-4o-mini"))
prompter = Prompter.load(
    "./my_prompter",
    model=MyModel,
    model_id="openai/gpt-4o-mini"
)

def lambda_handler(event, context):
    text = json.loads(event['body'])['text']
    result = prompter.run(text)

    return {
        'statusCode': 200,
        'body': json.dumps(result.dict())
    }

Validation Before Deploy

def validate_prompter():
    """Test prompter before deploying."""
    dspy.configure(lm=dspy.LM("openai/gpt-4o-mini"))
    prompter = Prompter.load("./my_prompter", model=MyModel)

    # Test cases
    test_cases = [
        "sample input 1",
        "sample input 2",
    ]

    for test_input in test_cases:
        try:
            result = prompter.run(test_input)
            assert result is not None
            print(f"✓ {test_input}")
        except Exception as e:
            print(f"✗ {test_input}: {e}")
            return False

    return True

if __name__ == "__main__":
    assert validate_prompter(), "Validation failed"

CI/CD Integration

# .github/workflows/deploy.yml
name: Deploy

on:
  push:
    branches: [main]

jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v3

      - uses: actions/setup-python@v4
        with:
          python-version: "3.10"

      - run: pip install -r requirements.txt

      - name: Validate prompter
        env:
          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
        run: python validate.py

      - name: Run tests
        run: pytest tests/

      - name: Build and push Docker image
        run: docker build -t my-prompter-api . && docker push my-prompter-api

Tips

  • Always validate before deploying
  • Use environment variables for API keys
  • Load prompters once (not per request) for performance
  • Monitor API usage in production
  • Have a rollback plan for new versions

See Also