AdalFlow

AdalFlow is the PyTorch-like, open-source library engineered to build and auto-optimize Large Language Model (LLM) task pipelines, including RAG systems and autonomous agents.

This is the definitive, PyTorch-inspired framework for production-ready LLM applications. AdalFlow provides a lightweight, modular architecture with powerful base classes (Component and DataClass) for maximum control and clarity: no black boxes here. Its core strength is the unified, auto-differentiative optimization engine, leveraging research like LLM-AutoDiff to automatically tune prompts for peak performance in zero-shot or few-shot scenarios. The design is model-agnostic, ensuring you can switch between providers (OpenAI, Groq, Anthropic) via a simple config change, eliminating vendor lock-in and accelerating your AI innovation cycle from research to deployment.

https://adal.ml/