Failure to parse regex ^(?i:(cs|cz)) (unknown extension) #18

Workflow file for this run

.github/workflows/diplomatize-comments.yml at 265aa65

	name: Diplomatize Comments

	on:
	issue_comment:
	types: [created]
	pull_request_review_comment:
	types: [created]
	# Manual trigger to process existing comments
	workflow_dispatch:
	inputs:
	issue_number:
	description: 'Issue/PR number to scan for hostile comments'
	required: true
	type: number
	dry_run:
	description: 'Dry run - analyze but do not hide/rewrite'
	required: false
	type: boolean
	default: true

	jobs:
	# Job for automatic comment moderation (on new comments)
	diplomatize:
	if: github.event_name != 'workflow_dispatch'
	runs-on: ubuntu-latest

	permissions:
	issues: write
	pull-requests: write
	models: read

	steps:
	- name: Moderate and diplomatize hostile comments
	uses: actions/github-script@v7
	env:
	OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
	with:
	github-token: ${{ secrets.GITHUB_TOKEN }}
	script: \|
	const originalComment = context.payload.comment.body;
	const commentAuthor = context.payload.comment.user.login;
	const commentNodeId = context.payload.comment.node_id;

	// Don't process bot comments
	if (context.payload.comment.user.type === 'Bot') {
	console.log('Skipping bot comment');
	return;
	}

	// Don't process maintainer comments
	const maintainers = ['dvershinin'];
	if (maintainers.includes(commentAuthor)) {
	console.log(`Skipping maintainer comment from ${commentAuthor}`);
	return;
	}

	console.log(`Checking comment from ${commentAuthor}`);

	// Step 0: Check for acronyms/patterns OpenAI doesn't recognize
	const knownBadPatterns = [
	{ pattern: /\bgfy\b/i, reason: 'profanity acronym' },
	{ pattern: /\bstfu\b/i, reason: 'profanity acronym' },
	{ pattern: /\bgtfo\b/i, reason: 'profanity acronym' },
	{ pattern: /\bai slop\b/i, reason: 'dismissive language' },
	];

	let patternMatch = null;
	for (const { pattern, reason } of knownBadPatterns) {
	if (pattern.test(originalComment)) {
	patternMatch = reason;
	break;
	}
	}

	// Step 1: Use OpenAI's FREE moderation endpoint to check for violations
	const moderationResponse = await fetch('https://api.openai.com/v1/moderations', {
	method: 'POST',
	headers: {
	'Authorization': `Bearer ${process.env.OPENAI_API_KEY}`,
	'Content-Type': 'application/json'
	},
	body: JSON.stringify({
	input: originalComment
	})
	});

	if (!moderationResponse.ok) {
	console.log(`Moderation API error: ${moderationResponse.status}`);
	return;
	}

	const moderationData = await moderationResponse.json();
	const result = moderationData.results[0];

	// Check if any category is flagged
	const flaggedCategories = Object.entries(result.categories)
	.filter(([_, flagged]) => flagged)
	.map(([category, _]) => category);

	// Also check for high scores in harassment even if not flagged
	const harassmentScore = result.category_scores.harassment \|\| 0;
	const hateScore = result.category_scores.hate \|\| 0;

	const isFlagged = result.flagged \|\| harassmentScore > 0.3 \|\| hateScore > 0.3 \|\| patternMatch;

	if (!isFlagged) {
	console.log('Comment passed moderation - no action needed');
	console.log(`Scores: harassment=${harassmentScore.toFixed(3)}, hate=${hateScore.toFixed(3)}`);
	return;
	}

	let reason;
	if (patternMatch) {
	reason = patternMatch;
	} else if (flaggedCategories.length > 0) {
	reason = flaggedCategories.join(', ');
	} else {
	reason = `high toxicity score (harassment: ${harassmentScore.toFixed(2)}, hate: ${hateScore.toFixed(2)})`;
	}

	console.log(`Comment flagged: ${reason}`);

	// Step 2: Use GitHub Models to rewrite the comment professionally
	const rewriteResponse = await fetch('https://models.github.ai/inference/chat/completions', {
	method: 'POST',
	headers: {
	'Authorization': `Bearer ${{ secrets.GITHUB_TOKEN }}`,
	'Content-Type': 'application/json',
	'Accept': 'application/json'
	},
	body: JSON.stringify({
	model: 'openai/gpt-4o-mini',
	messages: [
	{
	role: 'system',
	content: `Rewrite this comment to be professional and constructive.

	Rules:
	- PRESERVE all technical content, bug reports, criticisms, and valid points
	- REMOVE profanity, insults, personal attacks, and hostile language
	- Keep the same meaning and structure
	- Do NOT add emojis or excessive pleasantries
	- Output ONLY the rewritten comment, nothing else`
	},
	{
	role: 'user',
	content: originalComment
	}
	],
	temperature: 0.2,
	max_tokens: 2000
	})
	});

	if (!rewriteResponse.ok) {
	const errorText = await rewriteResponse.text();
	console.log(`GitHub Models API error: ${rewriteResponse.status} - ${errorText}`);
	return;
	}

	const rewriteData = await rewriteResponse.json();
	const rewritten = rewriteData.choices[0].message.content.trim();

	// Step 3: Hide the original comment using GraphQL
	console.log('Hiding original comment...');
	await github.graphql(`
	mutation($id: ID!, $classifier: ReportedContentClassifiers!) {
	minimizeComment(input: {subjectId: $id, classifier: $classifier}) {
	minimizedComment {
	isMinimized
	}
	}
	}
	`, {
	id: commentNodeId,
	classifier: 'ABUSE'
	});

	// Step 4: Post the diplomatic version
	const issueNumber = context.payload.issue?.number \|\| context.payload.pull_request?.number;

	await github.rest.issues.createComment({
	owner: context.repo.owner,
	repo: context.repo.repo,
	issue_number: issueNumber,
	body: `Comment from @${commentAuthor} (reformatted for constructive discussion):

	---

	${rewritten}

	---

	<sub>🤖 Original hidden due to: ${reason}. Technical content preserved.
	[Community guidelines](https://github.com/${context.repo.owner}/${context.repo.repo}/blob/master/CODE_OF_CONDUCT.md)</sub>`
	});

	console.log('Done - comment hidden and diplomatic version posted');

	# Job for manual scanning of existing comments
	scan-existing:
	if: github.event_name == 'workflow_dispatch'
	runs-on: ubuntu-latest

	permissions:
	issues: write
	pull-requests: write
	models: read

	steps:
	- name: Scan and moderate existing comments
	uses: actions/github-script@v7
	env:
	OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
	DRY_RUN: ${{ inputs.dry_run }}
	with:
	github-token: ${{ secrets.GITHUB_TOKEN }}
	script: \|
	const issueNumber = ${{ inputs.issue_number }};
	const dryRun = process.env.DRY_RUN === 'true';

	console.log(`\n${'='.repeat(60)}`);
	console.log(`Scanning issue/PR #${issueNumber} for hostile comments`);
	console.log(`Mode: ${dryRun ? '🔍 DRY RUN (analyze only)' : '⚡ LIVE (will hide and rewrite)'}`);
	console.log(`${'='.repeat(60)}\n`);

	// Maintainers to skip
	const maintainers = ['dvershinin'];

	// Bad patterns OpenAI misses
	const knownBadPatterns = [
	{ pattern: /\bgfy\b/i, reason: 'profanity acronym' },
	{ pattern: /\bstfu\b/i, reason: 'profanity acronym' },
	{ pattern: /\bgtfo\b/i, reason: 'profanity acronym' },
	{ pattern: /\bai slop\b/i, reason: 'dismissive language' },
	];

	// Get all comments on the issue
	const comments = await github.rest.issues.listComments({
	owner: context.repo.owner,
	repo: context.repo.repo,
	issue_number: issueNumber,
	per_page: 100
	});

	console.log(`Found ${comments.data.length} comments to analyze\n`);

	let flaggedCount = 0;
	let processedCount = 0;

	for (const comment of comments.data) {
	const author = comment.user.login;
	const nodeId = comment.node_id;
	const body = comment.body;
	const commentId = comment.id;

	// Skip bots and maintainers
	if (comment.user.type === 'Bot') continue;
	if (maintainers.includes(author)) continue;

	processedCount++;

	// Check patterns
	let patternMatch = null;
	for (const { pattern, reason } of knownBadPatterns) {
	if (pattern.test(body)) {
	patternMatch = reason;
	break;
	}
	}

	// Call OpenAI moderation
	const moderationResponse = await fetch('https://api.openai.com/v1/moderations', {
	method: 'POST',
	headers: {
	'Authorization': `Bearer ${process.env.OPENAI_API_KEY}`,
	'Content-Type': 'application/json'
	},
	body: JSON.stringify({ input: body })
	});

	if (!moderationResponse.ok) {
	console.log(`[${commentId}] Moderation API error: ${moderationResponse.status}`);
	continue;
	}

	const moderationData = await moderationResponse.json();
	const result = moderationData.results[0];
	const harassmentScore = result.category_scores.harassment \|\| 0;
	const hateScore = result.category_scores.hate \|\| 0;

	const isFlagged = result.flagged \|\| harassmentScore > 0.3 \|\| hateScore > 0.3 \|\| patternMatch;

	const preview = body.substring(0, 60).replace(/\n/g, ' ') + (body.length > 60 ? '...' : '');

	if (!isFlagged) {
	console.log(`✅ [${author}] PASS (h=${harassmentScore.toFixed(2)})`);
	console.log(` "${preview}"\n`);
	continue;
	}

	flaggedCount++;

	let reason;
	if (patternMatch) {
	reason = patternMatch;
	} else if (result.flagged) {
	reason = Object.entries(result.categories).filter(([_, v]) => v).map(([k]) => k).join(', ');
	} else {
	reason = `harassment: ${harassmentScore.toFixed(2)}, hate: ${hateScore.toFixed(2)}`;
	}

	console.log(`🚨 [${author}] FLAGGED - ${reason}`);
	console.log(` "${preview}"`);
	console.log(` URL: ${comment.html_url}`);

	if (dryRun) {
	console.log(` [DRY RUN] Would hide and rewrite this comment\n`);
	continue;
	}

	// LIVE MODE: Rewrite and hide
	console.log(` Rewriting...`);

	const rewriteResponse = await fetch('https://models.github.ai/inference/chat/completions', {
	method: 'POST',
	headers: {
	'Authorization': `Bearer ${{ secrets.GITHUB_TOKEN }}`,
	'Content-Type': 'application/json',
	'Accept': 'application/json'
	},
	body: JSON.stringify({
	model: 'openai/gpt-4o-mini',
	messages: [
	{
	role: 'system',
	content: `Rewrite this comment to be professional and constructive.
	Rules:
	- PRESERVE all technical content, bug reports, criticisms, and valid points
	- REMOVE profanity, insults, personal attacks, and hostile language
	- Keep the same meaning and structure
	- Do NOT add emojis or excessive pleasantries
	- Output ONLY the rewritten comment, nothing else`
	},
	{ role: 'user', content: body }
	],
	temperature: 0.2,
	max_tokens: 2000
	})
	});

	if (!rewriteResponse.ok) {
	console.log(` GitHub Models error: ${rewriteResponse.status}\n`);
	continue;
	}

	const rewriteData = await rewriteResponse.json();
	const rewritten = rewriteData.choices[0].message.content.trim();

	// Hide the comment
	console.log(` Hiding original...`);
	await github.graphql(`
	mutation($id: ID!, $classifier: ReportedContentClassifiers!) {
	minimizeComment(input: {subjectId: $id, classifier: $classifier}) {
	minimizedComment { isMinimized }
	}
	}
	`, { id: nodeId, classifier: 'ABUSE' });

	// Post diplomatic version
	console.log(` Posting diplomatic version...`);
	await github.rest.issues.createComment({
	owner: context.repo.owner,
	repo: context.repo.repo,
	issue_number: issueNumber,
	body: `Comment from @${author} (reformatted for constructive discussion):

	---

	${rewritten}

	---

	<sub>🤖 Original hidden due to: ${reason}. Technical content preserved.
	[Community guidelines](https://github.com/${context.repo.owner}/${context.repo.repo}/blob/master/CODE_OF_CONDUCT.md)</sub>`
	});

	console.log(` ✅ Done!\n`);
	}

	console.log(`\n${'='.repeat(60)}`);
	console.log(`SUMMARY`);
	console.log(`${'='.repeat(60)}`);
	console.log(`Comments analyzed: ${processedCount}`);
	console.log(`Comments flagged: ${flaggedCount}`);
	if (dryRun && flaggedCount > 0) {
	console.log(`\n⚠️ Re-run with dry_run=false to actually hide and rewrite these comments`);
	}

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Failure to parse regex ^(?i:(cs|cz)) (unknown extension) #18

Workflow file

Failure to parse regex ^(?i:(cs|cz)) (unknown extension) #18

Uh oh!

Workflow file for this run