Player not loading? Watch on YouTube
This sponsored tutorial connects Zed to LM Studio for a local AI coding workflow. The speaker says the setup can work offline and keep code on the computer. Early examples generate a Python greeting function, add documentation, and explain main.py through Zed's sidebar. The speaker cautions that generated explanations need checking.
The setup starts with downloading LM Studio and Qwen 3.5 9B. After enabling developer mode and the local LLM service, the speaker starts the server and loads the model. The model panel supplies its API identifier and address. For this demonstration, the context length is 12,000. A Python system prompt requests code without backticks, explanatory comments, and type annotations; the speaker also disables thinking.
In Zed, the tutorial enables AI features, selects the LM Studio provider, checks the API URL, and chooses the loaded Qwen model in a new thread. A greeting tests the connection before a prompt generates a simple chatbot. The brief test produces a response, but the generated program has no loop. The speaker recommends thorough testing before production use.
The final example lowers the context length to 600 and reloads the model. Even a greeting then triggers an error because the initial prompt exceeds the available context. The suggested remedies are a larger context or shorter input. The speaker returns to 12,000 or more and advises experimenting with the setting.
