api-development-api-impl

Skill: api_impl

Purpose: Implement inference endpoints consistently

Use when:

• Adding new API endpoints
• Need consistent patterns (models, errors, logging, config)
• Docker/runtime requirements apply

---

Process

When implementing an inference endpoint:

1. Ask for required information:

   • Route path
   • Request schema
   • Response schema
   • Latency budget
   • Model/service location

2. Ensure implementation includes:

   • Request/response Pydantic models
   • Input validation + clear error messages
   • Structured logs + trace IDs
   • Timeouts, retries if calling downstream services
   • Health check consideration

3. Provide deliverables:

   • Handler code
   • Unit tests
   • Update to OpenAPI/docs
   • Docker/runtime notes if needed

4. End with "how to test locally" commands

---

Output Format

## API Implementation: [Route]

### Specification
- **Route**: [METHOD /path]
- **Request**: [schema summary]
- **Response**: [schema summary]
- **Latency budget**: [target]

### Models

```python
# Request model
class [Name]Request(BaseModel):
    ...

# Response model
class [Name]Response(BaseModel):
    ...

Handler

@router.[method]("[path]")
async def [handler_name](...):
    ...

Tests

def test_[name]_success():
    ...

def test_[name]_validation_error():
    ...

OpenAPI Update

[Any documentation changes needed]

Runtime Notes

• [Docker considerations]
• [Environment variables]
• [Dependencies]

Test Locally

# Start server
[command]

# Test endpoint
curl -X [METHOD] http://localhost:8000/[path] ...

---

## Example Usage

/api_development_api_impl Route: /predict/sentiment Request: {"text": str} Response: {"sentiment": str, "confidence": float} Latency: <200ms p95

Or:

/api_development_api_impl Route: POST /api/v1/embeddings Request: {"texts": list[str], "model": str} Response: {"embeddings": list[list[float]]} Latency: <500ms for batch of 10