watsonx Code Assistant

Generates code suggestions as developers type, leveraging IBM Granite or IBM Cloud watsonx models to predict next tokens based on current file context and optionally referenced workspace symbols (files, classes, methods) via @-syntax. The extension monitors keystroke patterns and triggers completion suggestions without explicit user invocation, integrating directly into VS Code's IntelliSense pipeline.

Accepts free-form natural language prompts in a chat panel within VS Code and generates code snippets, functions, or entire code blocks using IBM Granite or cloud-based watsonx models. The chat interface maintains conversation history within a session, allowing iterative refinement of generated code through follow-up prompts. Generated code can be inserted directly into the editor or copied manually.

Supports local deployment of IBM's Granite model (via watsonx Code Assistant Individual) for offline, on-device code assistance without cloud connectivity or data transmission. The local model runs on the developer's machine, processing code entirely locally with no external API calls. This option trades cloud model performance for privacy and offline capability. Local Granite deployment is configured separately from cloud deployment and requires local hardware resources (RAM, disk space, GPU optional).

Integrates as a native VS Code extension within the extension sandbox, providing workspace-scoped file access and respecting VS Code's security model. The extension can access files within the opened workspace folder(s) for context and code generation but cannot access system files outside the workspace or execute arbitrary system commands. Integration points include the editor context menu, command palette, chat panel, and inline suggestions. The extension does not provide additional security controls beyond VS Code's built-in sandbox.

Offers a freemium pricing structure where the base watsonx Code Assistant extension is free to install and use with local Granite model deployment (watsonx Code Assistant Individual), while cloud-based IBM Cloud watsonx service deployment requires separate provisioning and pricing (unspecified in marketplace listing). This allows free access to core capabilities via local model while offering premium cloud deployment for organizations. Pricing details for cloud service are not documented in the marketplace listing.

Analyzes existing functions, methods, or classes in the current file and generates corresponding unit tests using the model's understanding of code behavior and common testing patterns. The extension identifies test-worthy code units and generates test cases covering typical scenarios, edge cases, and error conditions. Generated tests are formatted for the detected language's testing framework (Jest for JavaScript, pytest for Python, JUnit for Java, etc.).

Analyzes functions, methods, classes, or code blocks and generates descriptive comments, docstrings, and documentation in language-appropriate formats (JSDoc for JavaScript, docstrings for Python, Javadoc for Java, etc.). The generator understands code intent and produces documentation that explains parameters, return types, side effects, and usage examples. Documentation is inserted inline or presented for manual insertion.

Analyzes selected code blocks, functions, or entire files and generates natural language explanations of what the code does, how it works, and what its intent is. The model breaks down complex logic into understandable steps, identifies potential issues, and explains algorithm behavior. Explanations are presented in a chat or side panel and can be iteratively refined through follow-up questions.

Converts code from one programming language to another using a structured syntax: `translate [from <source_language>] to <target_language> <code_reference>`. The model understands language-specific idioms, standard libraries, and syntax conventions to produce idiomatic code in the target language. Translation preserves logic and functionality while adapting to target language best practices. Supports translation between Java, Python, JavaScript, TypeScript, Go, C/C++, and other supported languages.

Available via optional companion extension (IBM.wca-enterprise-java), this capability provides specialized analysis and code generation for Java applications, including detection of outdated patterns, suggestions for modernization (e.g., Java 8+ features, Spring Boot upgrades), and generation of refactored code. Enhanced explanation and testing capabilities are tailored to Java-specific frameworks and patterns (Spring, Jakarta EE, etc.). This extends the base watsonx Code Assistant with Java-specific domain knowledge.

Detects the programming language of the current file (based on file extension and content) and adapts all capabilities (completion, generation, explanation, testing) to language-specific syntax, conventions, and idioms. Supports Java, Python, JavaScript, TypeScript, Go, C, C++, Kotlin, Ruby, PHP, Perl, R, Swift, Bash, and C#. Language detection is automatic and transparent; no manual language selection required. Context includes language-specific standard libraries, frameworks, and testing conventions.

Enables developers to explicitly reference workspace symbols (files, classes, methods, functions) using @-syntax (e.g., @ClassName, @methodName, @filename.py) within chat prompts, code generation requests, or completion context. The extension resolves these references to actual code definitions and includes them in the model's context window, allowing precise anchoring of AI-generated code to specific codebase artifacts. This mechanism bridges the gap between implicit context (current file) and explicit, developer-controlled context.

Integrates with IBM Cloud's watsonx Code Assistant service for cloud-based model inference, allowing organizations to provision a managed service instance and configure the extension to use cloud-based models instead of local deployment. Cloud deployment provides centralized management, organization-level configuration, and access to IBM's latest models. Authentication is handled via IBM Cloud API keys or service credentials. Cloud deployment is recommended for 'best performance and full set of features' according to documentation.