Issue Sentinel

Drop any GitHub issue URL. Get instant classification — type, urgency, sentiment — powered by 35+ regex patterns. Zero API keys. Runs in your browser.

Python 3.10+ ▸ pip install issue-sentinel ⚡ Rule-Based + LLM ⊘ Zero Config 📄 MIT License
0
Regex Patterns
0
Urgency Signals
0
Sentiment Markers
0
Auto-Actions

⚡ Live Demo

// runs locally in your browser

› input.classify()

// or enter manually

› result.toJSON()

🛡️

awaiting input...

paste a URL or click a sample to begin

⚙️ Architecture

// how it works
📥 INPUT GitHub Issue title + body + URL 🔬 PARSER Tokenize & normalize extract signals 🏷️ TYPE bug · feature · question 🎯 URGENCY p0 · p1 · p2 · p3 💬 SENTIMENT frustrated → positive 🧬 MERGE Weighted fusion confidence scoring 📊 OUTPUT labels[] priority routing 🤖 LLM optional · GPT / Claude fallback ─── deterministic flow ┈┈┈ optional LLM path ⬡ parallel analyzers run concurrently

🚀 Quick Start

// 3 ways to run
terminal
# ── Install ────────────────────────────────────────────── $ pip install issue-sentinel # ── CLI: classify a single issue ───────────────────────── $ issue-sentinel classify --title "Login API returns 500" \ --body "Crashes after upgrade to v3, was working in v2.9" type: regression confidence: 0.60 urgency: p1 score: 0.60 sentiment: negative labels: [regression, p1] # ── GitHub Action: auto-triage on every new issue ──────── # .github/workflows/triage.yml uses: kustonaut/issue-sentinel@v1 with: github-token: ${{ secrets.GITHUB_TOKEN }}

🧩 Features

// what you get
🏷️

Type Classification

Detects bug, feature-request, question, regression, and docs using 35+ battle-tested regex patterns.

🎯

Urgency Scoring

4-tier priority system (p0–p3) with weighted signals: security vulns, regressions, production outages, CVE references.

💬

Sentiment Analysis

5-class sentiment detection — frustrated, negative, neutral, constructive, positive — to prioritize unhappy users.

⚡

GitHub Action

Drop kustonaut/issue-sentinel@v1 into any workflow. Auto-labels every new issue. Zero config.

🤖

Optional LLM Mode

Enable --llm for GPT/Claude-powered classification when regex falls short. Graceful fallback to rules.

🧪

Fully Tested

48 unit tests, 100% CI on Python 3.10–3.12, strict mypy, ruff linting. Ship with confidence.