valuator 1.1.0

A lightweight library to fetch and search AI model pricing data

Valuator

A lightweight Python library to fetch AI model pricing data and perform searches to retrieve input and output costs per token for matching models.

Installation

pip install valuator

Usage

import asyncio
from valuator import Valuator

async def main():
    valuator = Valuator()
    try:
        # Initialize with default force_refresh=True to fetch latest data
        await valuator.initialize()
        print(valuator.get_model_costs("claude.*haiku"))
        print(valuator.get_model_costs("gpt-4"))
        # Use force_refresh=False to prefer cache if unchanged
        await valuator.initialize(force_refresh=False)
        print(valuator.get_model_costs("gpt.*"))
    finally:
        await valuator.close()

asyncio.run(main())

Example

import asyncio
from valuator import Valuator

async def main():
    valuator = Valuator()
    try:
        await valuator.initialize()
        data = valuator.get_model_costs("us.anthropic.claude-3-5-sonnet-20240620-v1:0")
        if data:
            # Extract first matched model and its costs
            model_names = list(data.keys())
            selected_model = model_names[0]
            selected_model_costs = data[selected_model]

            print("Matched Model:")
            print(f"  Name: {selected_model}")
            print(f"  Input cost per token: {selected_model_costs['input_cost_per_token']}")
            print(f"  Output cost per token: {selected_model_costs['output_cost_per_token']}")
        else:
            print("No models found.")
    finally:
        await valuator.close()

asyncio.run(main())

Output:

Matched Model:
  Name: us.anthropic.claude-3-5-sonnet-20240620-v1:0
  Input cost per token: 3e-06
  Output cost per token: 1.5e-05

Features

• Fetches model pricing data.
• Automatically checks if the remote JSON has changed using ETag headers.
• Defaults to fetching the latest data (force_refresh=True) to ensure up-to-date model prices.
• Performs regex-based searches on model names for flexible matching.
• Returns only input_cost_per_token and output_cost_per_token for matched models.
• Optimized for low memory usage with efficient data structures (sets, cached regex).
• Asynchronous HTTP requests for fast data retrieval.

Requirements

• Python 3.8+
• aiohttp>=3.8.0

License

MIT License