hwchase17/langchain — reverse-engineered prompt

Build me a Python library for creating AI agents and LLM powered apps. I want developers to be able to install it, pick a chat model, send a simple prompt, then grow that into bigger workflows without rewriting everything.

The library should make it easy to connect models, prompts, tools, retrievers, embeddings, vector stores, and outside services in a consistent way. A user should be able to swap between model providers, add real time data, and build agents that can plan steps, call tools, use files, and hand work to smaller subagents when needed.

Please include a clean quickstart example, sensible defaults for beginners, and room for advanced users to customize workflows. Add basic hooks for tracing, debugging, evaluation, and deployment support, even if some of those connect to external services later.

Keep the code organized as a real open source Python package with docs, tests, and examples. Look up current docs online if you need to.