specific-ai 0.1.2

specific.ai python sdk

Specific.ai SDK

Installation

pip install specific-ai

Quick Start

from specific_ai import OpenAI

# Initialize the client
client = OpenAI(
    specific_ai_url="<your-specific-ai-service-URL>",
    api_key="<your-openai-api-key>",
    use_specific_ai_inference=False,
)

# Use like regular OpenAI client with the SpecificAI additional field
response = client.chat.completions.create(
    model="gpt-3.5-turbo",
    messages=[
        {"role": "user", "content": "Hello!"}
    ],
    specific_ai={
        "task_name": "greeting",
    }
)

Features

• 🎯 Custom-tailored models optimized for your specific use cases
• 🚀 Drop-in replacement for OpenAI's official Python client
• 📊 Automatic request and response logging
• 🎯 Response optimization capabilities
• 📈 Advanced analytics integration

Core Features Explained

Model Inference Options

The SDK offers two approaches to model inference, controlled by use_specific_ai_inference:

client = OpenAI(
    api_key="your-key",
    use_specific_ai_inference=True  # Enable SpecificAI's optimized models
)

• Standard Inference (use_specific_ai_inference=False):

  • Uses OpenAI's models directly
  • Standard pricing and performance
  • No optimization layer

• SpecificAI Inference (use_specific_ai_inference=True):

  • Uses task-specific models optimized for your use cases
  • Potential improvements in performance and cost
  • Automatic fallback to standard OpenAI models if needed
  • Requires prior data collection for model training

Best Practices

1. Start with Data Collection:

   • Keep use_specific_ai_inference=False initially
   • This builds up training data for your use cases

2. Transition to Optimized Models:

   • Once sufficient data is collected
   • Enable use_specific_ai_inference=True
   • Monitor performance improvements

3. Testing New Features:

   • Use A/B testing between standard and optimized models
   • Compare performance metrics
   • Gradually roll out optimization to production

Logging

This SDK uses Python's standard logging module. To capture logs from the SDK, configure logging in your application:

import logging

# Basic console logging
logging.basicConfig(level=logging.INFO)

# Or for more detailed logs
logging.basicConfig(
    level=logging.INFO,
    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
)

Support

Need help? Contact our support team at [e-mail]