twinny - AI Code Completion and Chat

Provides real-time code completion suggestions as developers type by sending the current file context (prefix and suffix) to a locally-hosted or remote AI model via OpenAI-compatible API endpoints. The extension integrates with VS Code's IntelliSense system to display multi-line and single-line completions inline, supporting both localhost Ollama instances and cloud providers (OpenAI, Anthropic, Groq, etc.). Completion triggers automatically during typing without explicit user invocation, with suggestions appearing as ghost text or in the autocomplete menu.

Provides a dedicated sidebar chat interface and full-screen chat mode where developers can ask questions about code, request explanations, or discuss implementation approaches. The chat system maintains conversation history across sessions and can access the current file context to provide code-aware responses. Requests are routed to the configured AI provider (local Ollama or cloud API) using the same OpenAI-compatible endpoint abstraction as code completion, allowing context-aware responses based on the developer's current work.

Allows developers to customize the system prompts and prompt templates used for code completion and chat requests through VS Code settings. This enables fine-tuning of AI behavior to match project-specific requirements, coding standards, or domain-specific patterns. Developers can define custom prompt variables and templates, allowing the extension to inject context (file type, project name, etc.) into prompts before sending to the AI model. This customization approach enables advanced users to optimize AI behavior without forking the extension.

Supports fully offline operation by routing all requests through locally-hosted inference servers (Ollama, vLLM, etc.) without requiring cloud API connectivity. The extension can operate entirely within a local network or on a single machine, enabling code completion and chat without internet access. This offline capability is critical for organizations with strict data privacy requirements, air-gapped networks, or unreliable internet connectivity. The extension automatically falls back to local inference if cloud providers are unavailable or misconfigured.

Optionally integrates with Symmetry Network, a decentralized peer-to-peer inference network, to distribute inference workloads across a network of nodes. This feature allows developers to leverage distributed computing resources for faster inference or to contribute their own hardware to the network. The integration is opt-in and transparent — developers can enable it through settings to participate in the P2P network while maintaining the same completion and chat interface.

Automatically indexes the developer's workspace by generating vector embeddings of code files, enabling the AI model to retrieve contextually relevant code snippets when generating completions or chat responses. The embeddings system scans the workspace on extension activation and maintains an index that can be queried to surface similar code patterns, function definitions, or architectural patterns relevant to the current task. This retrieval-augmented approach improves suggestion relevance by grounding AI responses in the project's actual codebase rather than relying solely on the model's training data.

Abstracts AI provider differences behind a unified OpenAI-compatible API interface, allowing developers to configure and switch between 9+ providers (localhost Ollama, OpenAI, Anthropic, Groq, Deepseek, Cohere, Mistral, Perplexity, OpenRouter) without changing extension code or prompts. The extension manages provider-specific authentication (API keys), endpoint configuration, and model selection through VS Code settings, enabling rapid experimentation with different models and providers. This abstraction layer allows the same completion and chat logic to work across all providers, reducing code duplication and enabling provider-agnostic feature development.

Analyzes staged or modified code changes in the current Git repository and generates descriptive commit messages using the configured AI provider. The feature integrates with VS Code's Git context to identify changed files and diffs, then sends this information to the AI model to produce commit messages following conventional commit formats or project-specific conventions. This automation reduces the cognitive load of writing commit messages while maintaining code quality and repository history clarity.

Analyzes selected code and generates refactoring suggestions through the chat interface, presenting proposed changes in a side-by-side diff view for easy comparison. The developer can select code in the editor, request refactoring assistance (e.g., 'extract this function', 'simplify this logic'), and the AI generates improved code. The diff view allows developers to review changes before applying them, reducing the risk of unintended modifications. This feature supports various refactoring patterns (function extraction, variable renaming, logic simplification, etc.) without requiring manual code manipulation.

Generates unit tests or integration tests from selected code or entire files by analyzing the code structure and logic, then producing test cases that cover common scenarios and edge cases. The generated tests are presented as copyable code blocks in the chat interface, allowing developers to paste them into test files. The feature supports multiple testing frameworks (Jest, Pytest, etc.) depending on the project's language and configuration, with the AI model inferring the appropriate framework from file context or explicit specification.

Analyzes selected code or entire files and generates natural language explanations of what the code does, how it works, and why certain patterns were chosen. The feature can produce documentation in multiple formats (docstrings, comments, markdown) and supports various documentation styles (JSDoc, Sphinx, etc.). Developers can request explanations at different levels of detail (high-level overview, line-by-line breakdown, architectural context) through the chat interface, with responses appearing as formatted text or code comments.

Allows developers to request code generation for new files or functions through the chat interface, with generated code presented as copyable blocks that can be inserted into new or existing documents. The feature supports creating boilerplate code, scaffolding for new projects, or implementation of specific algorithms or patterns. Developers can iteratively refine generated code through chat conversation, requesting modifications or improvements before copying the final result into their project.

Allows developers to configure custom API endpoints and ports for localhost inference servers (Ollama, vLLM, etc.) through VS Code settings, enabling flexible deployment of local models. The extension supports endpoint customization for both completion and chat requests, with automatic fallback or error handling for unreachable endpoints. This configuration approach enables developers to run inference servers on different machines, ports, or with custom configurations without modifying extension code.