Roo Code Chinese（原Roo Cline）

Generates code completions and implementations by analyzing the current file and project context, then routing requests to configured LLM endpoints (DeepSeek, Claude, or custom APIs) with system prompts translated and optimized for Chinese language models. The extension maintains conversation history within the VS Code editor to enable multi-turn code generation workflows without losing context between requests.

Provides an integrated chat panel in the VS Code sidebar that maintains multi-turn conversation history with the configured LLM. Messages are sent to the LLM endpoint with current file context automatically injected, and responses are rendered in the chat UI with syntax highlighting for code blocks. The conversation state persists within the current VS Code session.

Maintains synchronization with the upstream Roo Code project by merging updates and bug fixes from the original repository. The extension is a localized fork that inherits core functionality from Roo Code while adding Chinese language support and optimizations. Maintenance is performed by individual developer (Leo) with explicit disclaimers about update frequency and project continuity.

Abstracts LLM provider differences behind a unified API interface, allowing support for multiple providers (SiliconFlow, OpenRouter, OpenAI-compatible APIs) without duplicating code. The extension implements a provider adapter pattern that translates between the unified internal API and provider-specific request/response formats, enabling easy addition of new providers.

Allows users to configure custom LLM API endpoints and select between multiple providers (SiliconFlow, OpenRouter, OpenAI-compatible APIs, or local endpoints). The extension routes all inference requests to the configured endpoint using the selected model, with API key management handled through VS Code settings. Supports both cloud-hosted and self-hosted LLM services via standard API protocols.

Automatically captures and injects the current file's content, file path, and language information into LLM requests without requiring manual context specification. The extension detects the active editor tab and includes this context in the system prompt or request payload, enabling the LLM to generate code that aligns with the current file's syntax, style, and imports.

Implements prompt engineering and system message optimization specifically for lightweight Chinese LLMs (7B-14B parameters), particularly DeepSeek-R1-Distill series. The extension translates system prompts to Chinese and adjusts instruction formatting to match the training patterns of Chinese-optimized models, enabling better code generation quality from smaller models compared to using English prompts.

Exposes AI capabilities through VS Code command palette, allowing users to trigger code generation, refactoring, and chat actions via keyboard shortcuts or command search. Commands are registered in the extension's activation context and can be invoked without using the sidebar chat interface, enabling power users to work entirely through keyboard-driven workflows.

Maintains conversation history and context across multiple LLM interactions within a single VS Code session, allowing users to ask follow-up questions and reference previous responses without losing context. The extension manages message history, token counting (if applicable), and context window management to ensure coherent multi-turn conversations.

Renders code blocks in the chat interface with syntax highlighting based on detected language, making code suggestions readable and visually distinct from natural language responses. The extension parses markdown code blocks from LLM responses and applies VS Code's built-in syntax highlighting engine to provide language-specific formatting.

Implements a privacy-first architecture that does not collect, store, or transmit user code or conversation data to the extension developer's servers. All LLM requests are routed directly to the configured endpoint (SiliconFlow, OpenRouter, etc.), and the extension itself does not log or persist user interactions. Privacy policy explicitly disclaims data collection and advises users to protect sensitive information.

Provides complete Chinese-language user interface for the extension, including sidebar labels, chat messages, command names, and settings. System prompts sent to LLMs are translated from English to Chinese, optimized for Chinese language models' training patterns. This enables native Chinese-language interaction without requiring English proficiency from users.