AlongWY/Qwen2.5-Coder — reverse-engineered prompt

Reverse engineered prompt

Build me a Python project around the Qwen coder model so I can actually try it end to end without a lot of setup. I want a simple demo I can run locally, plus a basic command line flow, where I can chat with the model about code, ask it to generate or edit code, and point it at a larger codebase for long context understanding. If the model supports tool calling, make that work too so it can handle more agent style coding tasks.

Please include a few clear example scripts for common use cases, and keep the project organized enough that I can also fine tune it and run some basic evaluation later since this repo already seems to head in that direction. Make the setup straightforward, document how to download or connect to the right model checkpoints, and make sure the tokenizer and tool parsing are set up correctly for the current release. If anything is unclear, look up the latest docs online and wire it up in the simplest practical way.

Want more depth? Deep Reverse