Genie for VS Code: the repository is documentation, not the extension

No source code, by design

The first thing to establish is what this repository is. The README states it directly: this repository is meant for documentation, bug reports and feature requests. The top level listing agrees, holding a README, a CHANGELOG, a LICENSE, images, a CNAME, a _config.yml and an _includes directory, which is a GitHub Pages site rather than an extension project. There is no package.json, no src/ and no build configuration. The extension itself is published as genieai.chatgpt-vscode on the Visual Studio Marketplace, and that listing is where the download count and version badge in the README come from. So if your evaluation process includes reading the code, stop here: you cannot, from this repository.

Bring your own key, or your own Azure deployment

The extension talks to models with credentials you supply. The feature list leads with using GPT-4o and other improved models via your own OpenAI API key, and separately with your own Azure OpenAI Service deployments. Azure support is configured through the genieai.azure.url setting, where you set your full Azure OpenAI deployment URL following the instructions in the setting description, and one instruction is easy to miss: you must also set the extension's model setting to the base model you used for the Azure deployment, because the extension does not infer it.

The editor integration is the product, not the chat

Three features are what make this more than a prompt box in a sidebar. First, a diff between your code and Genie's suggestion, viewable inside the editor with one click, which is the difference between reading a suggestion and evaluating it. Second, use it from the Problems window to explain and suggest a fix for compile-time errors, with the lightbulb quick fix action on underlined error code. Third, generate commit messages from your git changes, with a Genie: Generate a commit message command and shortcut that supports multi-folder workspaces. There is also creating or fixing files with one click or a keyboard shortcut, streaming answers in the editor or the sidebar conversation, the ability to stop a response to save tokens, and the ability to rename and personalise the assistant.

Nearly every behaviour has an opt-out setting

The settings list is unusually complete, which tells you the author expects people to disagree with some of it. genieai.systemMessage customises the system message or context that starts your conversation. genieai.promptPrefix.commit-message holds the prompt used for commit messages, and genieai.quickFix.enable is an explicit opt-out of the Quick Fix actions, while genieai.enableGenerateCommitMessage opts out of the commit message feature entirely. Context menu items were also moved under a Genie submenu, which is a small quality of life change for anyone with a crowded right click menu. Two of these settings are explicit opt-outs, while the other two configure prompts.

Conversations are a file on your disk, and exportable as Markdown

Conversation history is stored on your disk so you can continue at any time, and conversations can be renamed or removed from the sidebar without editing the genie.json file, which was the previous mechanism. The whole history exports at once in Markdown format. Both halves of that matter in different directions. The corresponding risk is the obvious one: your prompts and the code you pasted into them sit in a file on disk, which is a consideration for any machine you do not own outright. The README also describes the history feature as an experiment intended to collect feedback and measure compatibility across machines and operating systems, which is an honest signal about its maturity.

The release history is a model changelog that stopped in September 2024

Reading the What's new section against the tags is instructive. v0.0.9 added GPT-4 and GPT-3.5 Turbo models and deprecated older preview names. v0.0.11 added gpt-4o and the 2024 models with improved maxTokens, fixed Editor View to use your selected model, and fixed the commit message feature that broke when VS Code updated its APIs. v0.0.12 added the custom system message and the o1-mini and o1-preview models, with a note that those models carry usage tier limitations. The last of those was published on 2024-09-13, and the last push to this documentation repository was on 2026-10-08. The listed release history therefore shows no extension release after 2024-09-13.

Against the built-in Copilot chat or your own key in a sidebar

Genie's documented features include editor diffs, Problems window explanations and suggested fixes, commit-message generation, local conversation history and Markdown export.

Editorial conclusion

Genie suits someone who wants a chat panel with editor awareness and refuses to hand a browser tool access to their workspace, because the diff view, the Problems window quick fix and the commit message generator are the features that make it more than a text box with an API key attached. Review the release history before adoption, and if your organisation routes through Azure OpenAI, verify the deployment URL format in genieai.azure.url still matches your tenant. Before installing, read the licence terms, note that the repository is documentation only so you cannot audit the extension source here, and decide whether your prompts and code should go to OpenAI directly at all. The repository is ISC licensed and the last push was on 2026-10-08.

Frequently asked questions

Official sources

1. ai-genie/chatgpt-vscode on GitHub
2. License: ISC
3. Project website
4. README
5. Releases