ai-cache 0.1.0

Lightweight automatic caching for LLM API responses

ai-cache

Lightweight automatic caching for LLM API responses

Save time, tokens, and API costs by caching LLM responses locally. One line to enable, zero code changes needed.

Features

• ✅ One-line activation - ai_cache.enable()
• ✅ Multi-provider support - OpenAI, Anthropic, Gemini
• ✅ Local SQLite storage - All data stays on your machine
• ✅ Zero dependencies - Only Python standard library
• ✅ Cache expiration - Optional TTL support
• ✅ Cache statistics - Monitor hits and savings

Installation

pip install ai-cache

Quick Start

import ai_cache
ai_cache.enable()

# Use any LLM API as normal - responses are automatically cached
import openai
response = openai.ChatCompletion.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "Hello!"}]
)

# Second identical call returns instantly from cache
response = openai.ChatCompletion.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "Hello!"}]
)

API

# Enable caching
ai_cache.enable()                        # Default: ~/.ai-cache/
ai_cache.enable(cache_dir="./cache")     # Custom directory
ai_cache.enable(ttl=3600)                # With 1-hour expiration

# Manage cache
stats = ai_cache.get_stats()             # Get hit/miss statistics
ai_cache.clear()                         # Clear all cache
ai_cache.invalidate(provider="openai")   # Clear specific provider
ai_cache.invalidate(model="gpt-4")       # Clear specific model
ai_cache.disable()                       # Disable caching

Supported Providers

• OpenAI (ChatGPT, GPT-4, etc.)
• Anthropic (Claude)
• Google Gemini

How It Works

1. Call ai_cache.enable() to activate
2. Library intercepts LLM API calls
3. Requests are fingerprinted (SHA256 of model + prompt + params)
4. Cached responses returned instantly, new requests cached automatically
5. All data stored locally in SQLite

License

MIT