raven-sdk 0.1.3

Async Kafka and HTTP producer SDK for Raven AI logs

Raven SDK

Async Kafka and HTTP producer SDK for AI logs with support for both streaming and batch processing.

Features

• Kafka Producer: Async Kafka producer for streaming AI logs
• HTTP Producer: Simple HTTP producer for batch AI log submissions
• Pydantic Models: Type-safe AI log models with automatic JSON serialization
• UTC Timestamps: Automatic UTC timestamp formatting with ISO-8601 standard

Installation

pip install raven-sdk

Quick Start

Kafka Producer

import asyncio
from datetime import datetime
from raven import AiLog, AiLogKafkaProducer

async def main():
    # Create producer
    producer = AiLogKafkaProducer("localhost:9092", topic="ai-logs")
    
    # Start producer
    await producer.start()
    
    try:
        # Create AI log
        log = AiLog(
            modelId="gpt-4",
            confidenceScore=0.95,
            responseTimeMs=150,
            timestamp=datetime.utcnow(),
            numericFeatures={"temperature": 0.7},
            categoricalFeatures={"type": "completion"}
        )
        
        # Send logs
        await producer.produce_ai_logs([log])
        
    finally:
        await producer.stop()

asyncio.run(main())

HTTP Producer

import asyncio
from datetime import datetime
from raven import AiLog, AiLogHttpProducer

async def main():
    # Create producer
    producer = AiLogHttpProducer("https://api.example.com/logs")
    
    # Create AI log
    log = AiLog(
        modelId="gpt-4",
        confidenceScore=0.95,
        responseTimeMs=150,
        timestamp=datetime.utcnow(),
        numericFeatures={"temperature": 0.7},
        categoricalFeatures={"type": "completion"}
    )
    
    # Send logs (no start/stop needed)
    await producer.produce_ai_logs([log])

asyncio.run(main())

API Reference

AiLog

Pydantic model representing an AI log entry.

Fields:

• modelId: str - Identifier for the AI model
• confidenceScore: float - Confidence score of the prediction
• responseTimeMs: int - Response time in milliseconds
• timestamp: datetime - UTC timestamp of the log
• numericFeatures: Dict[str, float] - Numeric features
• categoricalFeatures: Dict[str, str] - Categorical features

AiLogKafkaProducer

Async Kafka producer for AI logs.

Methods:

• __init__(kafka_bootstrap_servers: str, topic: str = "ai-logs")
• async start() - Start the producer
• async stop() - Stop the producer
• async produce_ai_logs(logs: List[AiLog]) - Send batch of logs
• async produce_stream_logs(logs: AsyncIterable[AiLog]) - Send streaming logs

AiLogHttpProducer

Simple HTTP producer for AI logs.

Methods:

• __init__(endpoint_url: str, timeout: int = 30)
• async produce_ai_logs(logs: List[AiLog]) - Send batch of logs via POST

Requirements

• Python 3.8+
• aiokafka>=0.8
• pydantic>=2.0
• aiohttp>=3.8

License

MIT License