huggingface/smolagents — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me a simple Python library called smolagents for making AI agents that can solve tasks by writing and running small bits of code.

I want someone to be able to install it, create an agent in a few lines, give it tools like web search, connect it to whatever language model they already use, then ask a question and watch the agent work through the answer. It should support code based agents, regular tool calling agents, streaming output, custom tools, and safe execution options like Docker or hosted sandboxes.

Please include a command line experience too, so a user can run an agent from the terminal with a prompt, or open an interactive setup that asks which model and tools to use. Add a web browsing agent command as well.

Make it easy to share or load agents and tools from Hugging Face Hub. Keep the code small, readable, well tested, and documented with quick examples. Look up current docs online if you need to.

Want more depth? Deep Reverse