Mirascope

Mirascope is a powerful, flexible, and user-friendly Python library that simplifies working with Large Language Models (LLMs) through a unified interface. It provides a comprehensive toolkit for building AI applications, from simple text generation to complex autonomous agent systems. The library is designed around the principle of "abstractions that aren't obstructions," meaning it provides helpful tools without getting in the way of your development process.

Mirascope works with a wide range of LLM providers, including OpenAI, Anthropic, Mistral, Google (Gemini/Vertex AI), Groq, Cohere, LiteLLM, Azure AI, and Amazon Bedrock. This provider-agnostic approach allows you to seamlessly switch between different models and providers while maintaining consistent workflow and code structure.

Mirascope offers several powerful features for LLM development including Prompt Templates for efficient prompt management, Streaming for real-time LLM responses, Response Models for structured output validation, JSON Mode for structured JSON data, Output Parsers for custom output transformation, Tools to extend capabilities by retrieving information, performing calculations, interacting with APIs, and executing actions, Agents for building autonomous AI agents, Asynchronous Processing for efficiency, Retries for failed API calls, and Evaluation Tools for LLM application strategies. Additionally, Mirascope is designed to be Pythonic by default, providing rich autocomplete, inline documentation, and type hints to catch errors before runtime, and offers both provider-agnostic and provider-specific engineering capabilities.

To get started with Mirascope, first install it using pip with your chosen provider. Then, set your provider's API key as an environment variable or directly within your Python code.

Mirascope supports a variety of use cases, including Text Generation for natural language content, Structured Information Extraction for managing structured data, Complex AI-Driven Agent Systems for autonomous agents, RAG Applications, Semantic Data Processing to remove duplicates and process similar content, Database Reporting from structured data, and Knowledge Management Systems to organize information.

Tools in Mirascope extend LLM capabilities by allowing them to perform specific tasks such as retrieving information from external sources, performing calculations or data processing, interacting with APIs or databases, and executing specific actions based on the LLM's decisions.

The Mirascope community offers several resources including extensive Documentation for features and usage, a Community for asking questions and chatting with other developers, Issues and Discussions to search for similar problems, and the GitHub Repository for source code and reporting issues. When seeking help, it is recommended to be as specific as possible, provide a minimal reproducible example, and list what you have already tried.

Yes, Mirascope allows you to use the same prompt logic across multiple providers. This enables you to easily try different models and optimize for performance or cost while maintaining a consistent workflow.