DSLForge

DSLForge is a VS Code extension for Copilot-assisted DSL authoring across Langium, ANTLR4, and Xtext workspaces.

It helps language engineers generate, explain, and iterate on DSL grammars inside VS Code without pretending to replace the underlying DSL engine. DSLForge detects the active grammar context, keeps AI features focused on explicit authoring tasks, and routes validation through the workspace's real command path so generated ideas can be checked against the actual project workflow.

Current Status

DSLForge is currently a pre-release extension.

• current branch work supports Langium, ANTLR4, and Xtext workflows
• AI commands are the primary authoring-assistance surface for scaffold, explanation, and sample generation
• Validate Current Grammar is intentionally non-AI
• AI commands require GitHub Copilot or another supported VS Code model environment
• if AI access is unavailable, DSLForge stops and shows guidance instead of inventing a fake fallback

Features

• explains current grammars and generates scaffold or sample proposals through supported VS Code model access
• detects the current Langium, ANTLR4, or Xtext workspace and grammar context
• follows import-aware grammar context selection for Langium, reference-aware context selection for ANTLR4, and workflow-aware context selection for Xtext
• resolves validation from the real workspace in this order:
  1. dslforge.validation.command
  2. auto-detected supported package.json script
  3. auto-detected gradlew task
  4. auto-detected mvnw goal
  5. setup guidance
• publishes normalized diagnostics into Problems
• uses AI only for explicit DSL authoring tasks instead of generic coding assistance
• exposes grammar actions from editor and Explorer context menus in addition to the Command Palette
• lets reviewed AI preview documents move toward workspace files through an explicit diff-and-apply flow

What DSLForge Does Not Do

• it does not replace Langium, ANTLR4, or Xtext
• it does not try to be a generic AI coding copilot
• it does not fake a validation engine when the workspace has not defined one
• it does not fake AI output when no supported model environment is available

Install

After the Marketplace listing is live, install DSLForge from VS Code Marketplace.

Until then, you can evaluate it locally:

1. run npm install
2. run npm exec -- vsce package
3. install the generated dslforge-0.3.2.vsix in VS Code

One-Minute Flow

1. Open a Langium, ANTLR4, or Xtext workspace.
2. Open the grammar file you are working on.
3. Run DSLForge: Explain Current Grammar, Create DSL Scaffold, or Generate Sample DSL when GitHub Copilot or another supported model is available.
4. Run DSLForge: Validate Current Grammar to check the grammar through the workspace's real validation command path.
5. If DSLForge cannot resolve validation, set dslforge.validation.command or add a supported package.json script, gradlew task, or mvnw goal.

Scaffold-only bootstrap flow:

1. Open any trusted workspace folder, even if it does not contain a detected DSL yet.
2. Run DSLForge: Create DSL Scaffold.
3. Review the proposal before creating files manually.

Recommended first local check from this repository:

1. open test-fixtures/langium/configured-command
2. open src/language/configured.langium
3. run DSLForge: Validate Current Grammar

Expected result:

• DSLForge selects the configured validation command first
• validation output is streamed to the Output panel
• normalized diagnostics are surfaced in Problems

Commands

Non-AI:

• DSLForge: Validate Current Grammar

AI-backed authoring assistance:

• DSLForge: Explain Current Grammar
• DSLForge: Create DSL Scaffold
• DSLForge: Generate Sample DSL

Create DSL Scaffold can run in two modes:

• detected-workspace mode when DSLForge recognizes a Langium, ANTLR4, or Xtext context
• bootstrap mode when the workspace is open but no supported DSL framework has been detected yet

Validation Behavior

Validate Current Grammar is intentionally non-AI.

Validation priority:

1. dslforge.validation.command
2. supported package.json script auto-detection
3. supported gradlew task auto-detection
4. supported mvnw goal auto-detection
5. setup guidance

This keeps DSLForge aligned with the workspace's real build and CI behavior instead of inventing an internal validation path that does not match the project.

AI Limits

The following commands require GitHub Copilot or another supported VS Code model environment:

• Explain Current Grammar
• Create DSL Scaffold
• Generate Sample DSL

If model access is unavailable, DSLForge:

• stops before making any model request
• shows setup or sign-in guidance
• records the AI gate result in the Output channel
• does not create a fake preview document

Adapter Scope

Current implemented adapters in this branch:

• Langium
• ANTLR4
• Xtext

Current responsibilities:

• project detection
• context selection
• validation orchestration
• diagnostics presentation

Next adapter work after this branch:

• later target: Generic mode

Settings

Current extension settings:

• dslforge.validation.command
• dslforge.validation.maxCapturedOutputCharacters
• dslforge.ai.maxContextFiles
• dslforge.ai.maxCharactersPerFile
• dslforge.ai.maxContextCharacters
• dslforge.telemetry.enabled
• dslforge.telemetry.endpointOverride

Telemetry

DSLForge can send minimal product telemetry through a server-side proxy when telemetry is enabled.

Telemetry behavior:

• DSLForge always respects VS Code telemetry enablement
• dslforge.telemetry.enabled can further disable extension telemetry
• released builds can embed a proxy endpoint through DSLFORGE_TELEMETRY_PROXY_ENDPOINT at package time
• dslforge.telemetry.endpointOverride or runtime DSLFORGE_TELEMETRY_ENDPOINT can override that endpoint for local validation
• the client never stores the PostHog project key directly

Telemetry scope:

• command and workflow events such as extension activation, AI gate results, validation runs, AI document generation, and reviewed preview apply flows
• coarse counts such as context file count, issue count, duration, and apply target count
• no grammar file content, prompts, workspace file paths, or raw validation output are intentionally sent

Proxy deployment:

1. deploy this repository to Vercel
2. set the Vercel environment variable POSTHOG_PROJECT_API_KEY
3. optionally set POSTHOG_HOST, TELEMETRY_RATE_LIMIT_WINDOW_MS, and TELEMETRY_RATE_LIMIT_MAX_EVENTS
4. set the GitHub repository variable DSLFORGE_TELEMETRY_PROXY_ENDPOINT to your production URL such as https://your-project.vercel.app/api/telemetry
5. publish the extension from CI so the packaged VSIX embeds that proxy endpoint

Development

Useful commands:

• npm run build
• npm run typecheck
• npm run test
• npm run test:diagnostics
• npm run test:projects
• npm run test:manifest
• npm exec -- vsce package

Contributions are welcome. Read CONTRIBUTING.md before opening an issue or pull request. Everyone participating in the project must follow the Code of Conduct.

Release notes live in CHANGELOG.md.