Tencent Cloud CodeBuddy

The Craft Agent capability enables autonomous generation and rewriting of code across multiple files based on natural language instructions. It uses Tencent Hunyuan or configurable third-party models (DeepSeek, GLM) to deeply comprehend instruction semantics and generate executable applications spanning multiple source files. The agent maintains cross-file consistency by understanding project structure context and generates code that is immediately compilable without manual intervention.

Provides real-time code completion suggestions as developers type, leveraging Tencent Hunyuan or configurable models to predict next tokens based on language syntax and project context. The completion engine supports 14+ programming languages (Java, Python, Go, C/C++, JavaScript, TypeScript, HTML, PHP, Ruby, Rust, Swift, Scala, Lua, Dart) with language-specific AST awareness. Suggestions are inserted directly into the editor via one-click acceptance or keyboard shortcuts.

Provides a dedicated sidebar panel within VS Code for accessing CodeBuddy features, maintaining conversation history, and managing code context. The sidebar displays ongoing conversations, allows code selection and insertion from chat, and provides quick access to custom agents and commands. Conversation history is persisted across sessions, enabling users to reference previous interactions. Code context can be selected from the editor and automatically included in conversations for context-aware responses.

Extends CodeBuddy functionality beyond VS Code to JetBrains IDEs (IntelliJ IDEA, Rider, PyCharm, Android Studio), Visual Studio, HarmonyOS DevEco Studio, CloudStudio, and WeChat Mini Program Developer Tools. Each IDE integration is optimized for platform-specific UI patterns, keybindings, and workflows. The extension uses IDE-native APIs for code insertion, diagnostics integration, and sidebar rendering. Platform support is continuously updated, though some IDEs may experience delays due to release schedules.

Analyzes selected code or entire files to identify violations of coding standards, best practices, and normalization rules. The code review engine uses Tencent Hunyuan models to understand code semantics and compare against configurable rule sets. Reviews can be triggered on-demand via command palette or sidebar, with results presented as inline annotations or conversation-style feedback. Custom rules can be managed at the team level for enterprise deployments.

Automatically generates unit tests for selected code or functions using language-specific test frameworks (Jest for JavaScript, pytest for Python, JUnit for Java, etc.). The generation engine analyzes function signatures, logic flow, and edge cases to create comprehensive test cases. Generated tests can be inserted directly into test files or created as new test files within the project structure. Supports both synchronous and asynchronous code patterns.

Detects code errors, compilation failures, and runtime issues, then generates fixes or repair suggestions. The repair engine integrates with VS Code's diagnostic system to identify errors from linters and compilers, then uses Tencent Hunyuan models to understand error context and propose corrections. Repairs can be applied automatically or presented as suggestions for manual review. Supports syntax errors, type mismatches, logic errors, and common anti-patterns.

Provides a chat interface within VS Code for asking technical questions and receiving answers grounded in Tencent Cloud documentation, WeChat development guides, and general programming knowledge. The Q&A engine uses multi-turn conversation to maintain context across questions, allowing follow-up queries and clarifications. Code from the current editor can be selected and inserted into conversations for context-specific advice. Answers can reference Tencent Cloud APIs and services, with links to documentation. Custom team knowledge bases can be integrated for enterprise deployments.

Allows switching between multiple AI models (Tencent Hunyuan, DeepSeek, GLM) for code generation and analysis tasks. The model configuration system enables users to select preferred models at the extension level, with support for third-party model integration via API configuration. Model switching is persistent across VS Code sessions. Enterprise deployments can enforce model policies and restrict available models. The architecture supports flexible model selection without requiring extension restart.

Enables creation of custom agents and commands tailored to team workflows, with centralized management for enterprise deployments. Custom agents can be defined to perform specific coding tasks (e.g., 'generate API endpoint', 'refactor for performance') and invoked via command palette or sidebar. Commands are stored and versioned at the team level, allowing sharing across developers. The agent creation interface supports natural language instruction definition and parameter configuration. Enterprise accounts can enforce agent policies and audit usage.

Integrates with the Model Context Protocol (MCP) open ecosystem, enabling CodeBuddy to work alongside other MCP-compatible tools and services. The MCP integration allows CodeBuddy to leverage external tools, knowledge bases, and services through standardized MCP server interfaces. This enables extensibility without modifying the core extension, supporting custom integrations for specialized domains or internal tools. MCP compatibility is explicitly stated but implementation details are not documented.

Provides centralized rules management system for enterprises to define, enforce, and audit coding standards, security policies, and architectural constraints across teams. Rules can be applied to code review, generation, and completion capabilities. The rules engine supports conditional logic, custom metrics, and integration with CI/CD pipelines. Rules are versioned and can be rolled out gradually across teams. Audit logs track rule violations and enforcement actions.