This library provides a Go-idiomatic framework for building generative AI applications and autonomous agents, targeting developers who prefer Go's performance and static typing over Python-based solutions. It aims to simplify the creation of complex AI systems by offering a clean, extensible, and efficient approach.

How It Works

Agency is built around a core concept of composable "operations" that can be chained together to form "processes." This design allows developers to separate business logic from AI provider implementations, promoting clean architecture. Custom operations can be created by implementing a simple interface, and interceptors can be used to observe or modify each step in a process. The library currently offers bindings for OpenAI's API, supporting text-to-text, text-to-image, text-to-speech, and speech-to-text functionalities.

Quick Start & Requirements

• Install package: go get github.com/neurocult/agency
• Requires an OpenAI API key (can be loaded via .env file using github.com/joho/godotenv/autoload).
• Example chat application provided in the README.
• Official documentation and examples are available.

Highlighted Details

• Pure Go implementation for speed and static typing.
• Supports OpenAI API (text-to-text, text-to-image, text-to-speech, speech-to-text).
• Extensible design with custom operation interfaces.
• Composable operations into processes with interceptor support.
• Roadmap includes external function calls, metadata, more providers, image-to-text, and agent API features.

Maintenance & Community

• The project is maintained by neurocult.
• Links to community channels or roadmaps are not explicitly provided in the README.

Licensing & Compatibility

• The README does not explicitly state a license.

Limitations & Caveats

The library is currently focused on OpenAI providers, with plans to expand to others. Some planned features like metadata and advanced agent APIs are still under development according to the roadmap.