AI-Powered Issue Labeling: 6 Open-Source Tools for Automated GitHub Triage

The githubnext/awesome-continuous-ai repository curates six open-source tools that use LLMs to automatically label GitHub issues by analyzing title and body text through GitHub Actions workflows.

Maintaining accurate labels on GitHub issues is essential for project organization, but manual triage consumes significant maintainer time. The githubnext/awesome-continuous-ai repository provides a curated collection of AI-powered issue labeling tools under its Continuous Triage section, offering ready-to-use GitHub Actions that integrate large language models directly into your workflow.

Available AI-Powered Issue Labeling Tools

Ultralytics Actions

The Ultralytics Actions tool operates as a monolithic GitHub Action that runs a lint-and-format pipeline. Internally, it invokes an LLM via actions/ai-inference to generate label suggestions based on the issue body, then calls the GitHub REST API to apply the labels. You declare it as a workflow step that triggers on issues events.

Automattic Issue Triage

Found in the automattic/jetpack repository, this custom GitHub Action parses the issue payload and feeds the text to either GitHub Models or an external LLM. After receiving a list of labels, it writes them back via the issues scope, typically triggering on issues: opened,reopened,edited.

GenAI Issue Labeller

The GenAI Issue Labeller (pelikhan/action-genai-issue-labeller) provides a minimal wrapper around GenAIScript. The action executes a GenAIScript that sends the issue description to a model, receives a JSON-structured label list, and calls github.issues.addLabels. The reference implementation lives in .github/workflows/genai-issue-labeller.yml within the awesome-continuous-ai repository.

Detect Duplicate Issues

Built as action-genai-issue-dedup, this tool extends basic labeling by adding a deduplication step. After labeling, it queries recent issues, generates embeddings to compute similarity, and flags potential duplicates. This reusable action can be chained into any workflow that already labels issues.

Detect Non-English Issues

This specialized tool uses a language-identification model via GitHub Models to determine if an issue is written in a non-English language. When detected, it automatically adds a non-english label. The standalone workflow file appears in the Home-Assistant core repository and is referenced in the README.

Continuous AI Resolver

A more elaborate bot that not only labels issues but also attempts to auto-resolve stale or already-fixed issues. It searches recent PRs and applies a "resolved" label when it identifies matching fixes, running as a scheduled workflow that combines issue triage with PR search APIs.

How AI Issue Labeling Works

All tools in the awesome-continuous-ai collection follow a standardized five-step pipeline:

1. Trigger: Capture GitHub events (issues: opened|reopened|edited).
2. Payload Extraction: Receive the issue title, body, and metadata.
3. LLM Inference: Process text through actions/ai-inference, GitHub Models, or self-hosted models.
4. Post-Processing: Transform raw model output into canonical GitHub labels using JSON schema validation.
5. API Call: Execute POST /repos/{owner}/{repo}/issues/{issue_number}/labels with appropriate issues: write permissions.

Required Permissions

According to the repository's SECURITY.md, these workflows require specific GitHub token permissions. The genai-issue-labeller.yml workflow demonstrates the minimal permission set:

permissions:
  contents: read
  issues: write        # needed to add labels

  models: read         # needed to call GitHub Models

Implementation Examples

Minimal GenAI Issue Labeller Setup

To implement the GenAI Issue Labeller from the repository's own workflow at .github/workflows/genai-issue-labeller.yml, use this configuration:

name: AI Issue Labeller
on:
  issues:
    types: [opened, reopened, edited]

permissions:
  contents: read
  issues: write        # needed to add labels

  models: read         # needed to call GitHub Models

jobs:
  label:
    runs-on: ubuntu-latest
    steps:
      - uses: pelikhan/action-genai-issue-labeller@v0
        with:
          github_token: ${{ secrets.GITHUB_TOKEN }}

Ultralytics Actions Configuration

For projects using the Ultralytics ecosystem, configure the action as shown in the README at lines 11-12:

steps:
  - name: Run Ultralytics AI labeling
    uses: ultralytics/actions@v1
    with:
      model: github-models  # chooses the built‑in GitHub Models inference service

      prompt: |
        Issue title: ${{ github.event.issue.title }}
        Issue body: ${{ github.event.issue.body }}
        Suggest appropriate GitHub labels, return a JSON array.

Non-English Issue Detection

To automatically flag non-English issues using the Home-Assistant approach:

steps:
  - name: Detect language
    uses: home-assistant/detect-non-english-issues@v1
    env:
      GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

Extending With Duplicate Detection

The repository includes .github/workflows/detect-duplicate-tools.yml demonstrating how to chain the duplicate detection action after initial labeling. This workflow first applies labels, then queries recent issues, embeds them for similarity comparison, and flags duplicates for maintainer review.

Summary

• The githubnext/awesome-continuous-ai repository provides six distinct AI-powered issue labeling tools under its Continuous Triage section.
• Tools range from simple labelers like GenAI Issue Labeller to complex systems like Continuous AI Resolver that auto-close stale issues.
• All implementations require issues: write permission and follow a standard pipeline: trigger → extract → infer → validate → apply.
• Reference implementations exist in .github/workflows/genai-issue-labeller.yml and automattic/jetpack for production-ready examples.
• Security considerations for external LLM invocations are documented in SECURITY.md.

Frequently Asked Questions

What permissions are required for AI-powered issue labeling?

Workflows require issues: write permission to add labels via the GitHub REST API, plus models: read when using GitHub Models. The genai-issue-labeller.yml file in the repository shows the minimal permission set, while SECURITY.md outlines additional considerations for safe LLM invocation.

Can I use self-hosted models instead of GitHub Models?

Yes. While tools like Ultralytics Actions and the Detect Non-English Issues action default to GitHub Models, the Automattic Issue Triage system supports external LLMs. You can configure the model endpoint in your workflow environment variables or action inputs according to the repository's README.md.

How does duplicate issue detection work alongside labeling?

The Detect Duplicate Issues action (action-genai-issue-dedup) runs after initial labeling. It queries recent issues, generates embeddings to compute similarity scores, and flags potential duplicates. This can be chained into existing workflows as shown in .github/workflows/detect-duplicate-tools.yml.

Which tool is best for small open-source projects?

The GenAI Issue Labeller (pelikhan/action-genai-issue-labeller) offers the simplest setup for small projects, requiring only a single workflow file and minimal configuration. The reference implementation at .github/workflows/genai-issue-labeller.yml provides a drop-in solution that works with GitHub's built-in models without requiring external API keys.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}