Repository navigation

Exception with Android Studio Rabbit 2026.2 #88

Workflow file for this run

.github/workflows/issue-translator.yml at 3809596

	name: Issue Translator

	# Translates non-English issues into English using GitHub Models (the built-in
	# `actions/ai-inference` action — no API key, only the repo's GITHUB_TOKEN).
	#
	# For a non-English issue it does two things:
	# 1. Rewrites the TITLE in place to "<original> (<English translation>)", so the
	# original text is preserved and the English rendering sits beside it.
	# 2. Posts a NEW comment containing a full English translation of the BODY, with
	# every image, attachment link, code block and template header kept verbatim.
	#
	# Triggers:
	# - issues.opened -> automatic on every new issue
	# - workflow_dispatch -> manual, pass an issue number to (re)translate an
	# existing issue (used for back-filling old issues)
	#
	# The step is idempotent: an already-appended title is not touched again, and a
	# second run will not post a duplicate translation comment (detected via a hidden
	# marker in the comment body).

	on:
	issues:
	types: [opened]
	workflow_dispatch:
	inputs:
	issue_number:
	description: "Issue number to (re)translate"
	required: true
	type: number

	permissions:
	# Update the issue title and post the translation comment.
	issues: write
	# Call GitHub Models for inference.
	models: read

	jobs:
	translate:
	runs-on: ubuntu-latest
	steps:
	- name: Resolve issue content
	id: issue
	uses: actions/github-script@v7
	with:
	script: \|
	let number, title, body;
	if (context.eventName === 'workflow_dispatch') {
	number = Number(context.payload.inputs.issue_number);
	const { data } = await github.rest.issues.get({
	...context.repo,
	issue_number: number,
	});
	title = data.title;
	body = data.body \|\| '';
	} else {
	number = context.payload.issue.number;
	title = context.payload.issue.title;
	body = context.payload.issue.body \|\| '';
	}
	core.setOutput('number', String(number));
	core.setOutput('title', title);
	core.setOutput('body', body);

	- name: Translate via GitHub Models
	id: ai
	uses: actions/ai-inference@v1
	with:
	model: openai/gpt-4o
	# The action's `max-tokens` defaults to 200, which truncates the
	# response. Set both the deprecated and current knobs generously.
	max-tokens: 16000
	max-completion-tokens: 16000
	system-prompt: \|
	You are a translation assistant for GitHub issues. You receive an
	issue's TITLE and BODY. Detect the primary natural language.

	Respond with ONLY a single valid JSON object (no markdown code
	fences, no prose before or after) with exactly these keys:
	"is_english": boolean -- true if the content is already primarily English
	"title_en": string -- English translation of the title ("" if is_english)
	"body_en": string -- English translation of the body ("" if is_english)

	Translation rules for body_en:
	- Translate ONLY natural-language prose.
	- Keep the following EXACTLY as in the original, byte for byte,
	translating nothing inside them:
	* <img ...> and any other HTML tags
	* markdown image syntax ![alt](url) and links [text](url)
	* URLs and attachment links (e.g. github.com/user-attachments/...)
	* fenced code blocks ``` ... ``` and inline `code`
	* issue-template section headers that start with "###"
	- Do NOT drop, summarize, or reorder ANY part of the body. Every
	line, image, and section present in the original must be present
	in body_en, in the same order.
	- Preserve blank lines and markdown structure.
	prompt: \|
	TITLE:
	${{ steps.issue.outputs.title }}

	BODY:
	${{ steps.issue.outputs.body }}

	- name: Apply translation
	uses: actions/github-script@v7
	env:
	# Read from the response FILE, not the `response` output: a large
	# translation can exceed the step-output size limit and arrive
	# truncated (which breaks JSON parsing).
	AI_RESPONSE_FILE: ${{ steps.ai.outputs.response-file }}
	ORIG_TITLE: ${{ steps.issue.outputs.title }}
	ISSUE_NUMBER: ${{ steps.issue.outputs.number }}
	with:
	script: \|
	const fs = require('fs');
	const raw = fs.readFileSync(process.env.AI_RESPONSE_FILE, 'utf8');
	const origTitle = process.env.ORIG_TITLE \|\| '';
	const number = Number(process.env.ISSUE_NUMBER);

	// The model is asked for raw JSON, but strip accidental ```json fences.
	const jsonText = raw
	.trim()
	.replace(/^```(?:json)?\s*/i, '')
	.replace(/\s*```$/, '')
	.trim();

	let parsed;
	try {
	parsed = JSON.parse(jsonText);
	} catch (e) {
	core.setFailed('Could not parse model response as JSON: ' + e.message + '\n---\n' + raw);
	return;
	}

	if (parsed.is_english) {
	core.info('Issue is already in English; nothing to do.');
	return;
	}

	// 1) Title: append the English translation, keeping the original.
	// Idempotent -- skip if the translation is already present.
	if (parsed.title_en && !origTitle.includes(parsed.title_en)) {
	await github.rest.issues.update({
	...context.repo,
	issue_number: number,
	title: `${origTitle} (${parsed.title_en})`,
	});
	core.info('Title updated.');
	} else {
	core.info('Title already translated; leaving it as-is.');
	}

	// 2) Body: post a translation comment (once).
	const marker = '<!-- ccg-ai-translation -->';
	const comments = await github.paginate(github.rest.issues.listComments, {
	...context.repo,
	issue_number: number,
	per_page: 100,
	});
	if (comments.some((c) => c.body && c.body.includes(marker))) {
	core.info('Translation comment already exists; skipping.');
	return;
	}

	const note =
	'> [!NOTE]\n' +
	'> This issue was written in another language. Below is an automatic ' +
	'English translation (the original text is preserved above).';
	const commentBody = `${marker}\n${note}\n\n---\n\n${parsed.body_en}`;
	await github.rest.issues.createComment({
	...context.repo,
	issue_number: number,
	body: commentBody,
	});
	core.info('Translation comment posted.');

Navigation Menu

Repository navigation

Exception with Android Studio Rabbit 2026.2 #88

Workflow file

Exception with Android Studio Rabbit 2026.2 #88

Uh oh!

Workflow file for this run