ChatGPT - Genie AI

Maintains persistent, multi-turn conversations within a VS Code sidebar panel that streams responses token-by-token from OpenAI or Azure OpenAI APIs. The extension preserves conversation history to disk in a local state store, enabling users to resume previous discussions across editor sessions. Streaming is implemented with cancellation support to allow users to stop token generation mid-response, reducing API costs for long-running queries.

Generates new code files directly into the VS Code workspace by sending the current editor context and user prompt to the selected LLM model, then automatically creates the file with the generated content. The extension integrates with VS Code's file creation APIs to place generated files in the workspace root or a user-specified directory, bypassing manual file creation steps.

Supports code analysis and generation for 40+ programming languages (JavaScript, Python, Java, C++, Go, Rust, etc.) by leveraging the underlying LLM's multilingual code understanding. The extension does not perform language-specific parsing or validation — instead, it sends raw code to the LLM and relies on the model's training data to understand syntax and semantics. Language detection is implicit based on file extension or user specification.

Stores all conversations to the local file system in an unencrypted format, allowing users to resume conversations across editor sessions without relying on cloud storage or external services. Conversation data is serialized to disk automatically after each message, and users can browse saved conversations in the sidebar. The storage location is managed by VS Code's extension storage API, typically in the user's home directory under `.vscode/extensions/genieai.chatgpt-vscode-*/`.

Generates unit tests, integration tests, or test cases based on existing code by sending the code and a test generation prompt to the LLM. The extension can analyze code for potential bugs, edge cases, or quality issues and suggest test cases to cover them. Generated tests are returned as code snippets that users can apply to their test files using the diff-and-apply mechanism.

Analyzes code for potential bugs, security vulnerabilities, performance issues, or code smell by sending code snippets to the LLM. The extension can review code in the editor, analyze error messages, or examine diffs to identify issues and suggest fixes. Code review is conversational — users can ask follow-up questions about detected issues and request explanations or alternative solutions.

Generates code modifications and displays them in VS Code's built-in diff viewer, showing original code on the left and AI-suggested changes on the right. Users can review the diff and apply changes with a single click, which updates the editor buffer. The extension uses VS Code's `TextEditor.edit()` API to apply changes atomically, ensuring undo/redo compatibility.

Integrates with VS Code's Problems window to detect compile-time errors and warnings, then sends the error message, file context, and code snippet to the LLM to generate explanations and suggested fixes. The extension registers Quick Fix actions in the Problems panel, allowing users to apply AI-suggested fixes directly from the error diagnostic. Fixes are applied using the same diff-and-apply mechanism as code modification.

Analyzes staged files in the git index by reading the git diff output, then sends the diff to the LLM to generate a semantically meaningful commit message. The extension integrates with VS Code's Source Control panel and can be invoked via the command palette (`Genie: Generate a commit message`) or a custom keybinding. Generated messages follow conventional commit format (e.g., 'feat: add user authentication') and can be customized via the `genieai.promptPrefix.commit-message` setting.

Allows users to select from a curated list of OpenAI models (GPT-4o, GPT-4 Turbo, GPT-3.5-turbo, o1-preview, o1-mini) or deploy custom models via Azure OpenAI Service. The extension stores the selected model in VS Code settings and routes all API calls through the chosen model's endpoint. Model selection is configurable via the extension settings UI, with fallback to a default model if none is explicitly selected.

Allows users to define a custom system message via the `genieai.systemMessage` setting, which is prepended to every conversation and code generation request. Additionally, commit message generation can be customized via `genieai.promptPrefix.commit-message`, enabling teams to enforce specific conventions or add context (e.g., 'Always include Jira ticket ID'). These settings are stored in VS Code's user or workspace settings, allowing per-project customization.

Exports all saved conversations to markdown format, allowing users to archive, share, or version-control their chat history. The export includes the full conversation thread with timestamps, model information, and code snippets. Conversations are stored locally on disk in an unencrypted format, and the extension provides a bulk export feature to save all conversations to a single markdown file or individual files per conversation.

Allows users to include the current editor file content, selected code snippets, or specific line ranges in their prompts to the AI. The extension automatically includes editor context when generating code or analyzing errors, and users can explicitly select code to include additional context. Context is sent as part of the API request to the LLM, enabling the AI to provide file-aware suggestions.

Implements token-level cancellation for streaming responses, allowing users to stop the AI mid-generation by clicking a cancel button or pressing Escape. When cancelled, the extension stops consuming tokens from the OpenAI API, reducing costs for long-running or verbose responses. Cancellation is implemented via the OpenAI streaming API's abort mechanism, which closes the HTTP connection and stops token generation.