Translate a file-system Markdown .md document into *.zh-CN.md inside VS Code.
What V1 Does
Adds one editor title action for eligible .md files.
Sends the entire source Markdown document in one chat/completions request.
Expects the model to return one outer ``````markdown fenced block.
Reuses an existing translated file when the cache is still valid.
Restores a deleted translated file from the persistent local blob cache when the source and config still match.
Refuses to overwrite a translated file that has unsaved editor changes.
Can optionally delete the workspace .zh-CN.md file when the translated document closes, while keeping the local blob cache.
Configuration
Set these settings in VS Code:
markdownTranslator.model
markdownTranslator.baseUrl (optional; defaults to https://api.openai.com/v1)
markdownTranslator.requestTimeoutMs
markdownTranslator.systemPrompt (optional)
markdownTranslator.localBlobCacheMaxBytes (optional; defaults to 10485760)
markdownTranslator.deleteTranslatedOnClose (optional; defaults to false)
Store the API key with the command:
Markdown Translator: Set API Key
Clear the API key with:
Markdown Translator: Clear API Key
Response Contract
The model must return exactly one outer fenced block using 5 backticks:
`````markdown
<translated markdown>
`````
The extension extracts only the Markdown inside the unique outer fence. JSON, malformed fences, or responses with zero or multiple candidate fences are rejected. If the model adds a little noise before or after the unique fence, that noise is discarded.
Development
npm install
npm run lint
npm run compile
npm test
Run the extension with F5 in VS Code.
Known Limits
Only .md file-system documents are supported.
.mdx, .markdown, untitled documents, and virtual documents are rejected.
V1 relies on prompt rules plus local integrity checks instead of AST rewriting.
