The GitHub issue #4424 for Zed relates to the lack of a feature for using local large language models (LLMs). In response to this, I proposed a workaround that enables the integration of local LLMs into Zed. This solution addresses the need for a non-proprietary, offline alternative to mainstream models like ChatGPT, potentially increasing privacy and control for users.
To integrate a custom model in Zed, I bypassed the limitation of only using OpenAI models. I did this by running the Mistral model from the Ollama library and cloning it to appear as "gpt-4-1106-preview." The steps included pulling and running the Mistral model, then using Ollama's commands to clone it. I updated Zed's settings to point to the local API URL of the cloned model. Restarting Zed applied these changes, enabling the use of the local LLM within Zed's environment.
To integrate a custom model in Zed, I bypassed the limitation of only using OpenAI models. I did this by running the Mistral model from the Ollama library and cloning it to appear as "gpt-4-1106-preview." The steps included pulling and running the Mistral model, then using Ollama's commands to clone it. I updated Zed's settings to point to the local API URL of the cloned model. Restarting Zed applied these changes, enabling the use of the local LLM within Zed's environment.
For more details, you can refer to the GitHub issue directly: https://github.com/zed-industries/zed/issues/4424#issuecomme...