Yesterday, we published our PyPI package named dhenara. This is a lightweight, straightforward framework for integrating multiple AI models into Python applications. It’s similar in spirit to LangChain but with a focus on simplicity and minimal dependencies.

This package is in a decent state, we have already integrated this into our production code for our multi provider chat-bot app (DhenChat) and works smoothly there.

Here are some features:

- Genuinely Open Source: Built from the ground up as a community resource, not an afterthought or internal tool - Unified API: Interact with different AI providers through a consistent interface - Type Safety: Built with Pydantic for robust type checking and validation - Easy Regeneration across Providers: With a unified Pydantic output and built-in prompt formatting, send output from a model to any other model easily - Streaming: First-class support for streaming responses along with accumulated responses similar to non-streaming responses - Async Support: Both synchronous and asynchronous interfaces for maximum flexibility - Resource Management: Automatic handling of connections, retries, and timeouts - Foundation Models: Pre-configured models with sensible defaults - Test Mode: Bring up your app with dummy responses for streaming and non-streaming generation - Cost/Usage Data: Derived cost and usage data along with responses, with optional charge for each model endpoint for commercial deployment

Would appreciate your thought and feedback on this.

Thanks, Ajith