Home

Local Repo Sentinel

A local, evidence-backed repository issue triage system for overnight code health monitoring.

Sentinel runs on your local machine, scans a codebase with deterministic detectors and an optional LLM judgment layer, deduplicates findings across runs, and produces a concise morning report of issues worth reviewing. After explicit approval, selected findings can become GitHub issues.

Key Features

• 18 built-in detectors covering Python, JavaScript/TypeScript, Go, Rust, and polyglot documentation
• Pluggable LLM providers — Ollama (local, default), OpenAI-compatible, Azure AI; optional, degrades gracefully
• Evidence-backed findings — every finding cites concrete code, config, or history evidence
• Fingerprint deduplication across runs via SQLite
• Morning report — scannable markdown grouped by severity
• Human-controlled — no autonomous actions; GitHub issues require explicit approval
• Plugin system — third-party detectors installable via pip install

What it does NOT do

• Implement fixes
• Make architecture plans
• Open pull requests
• Act autonomously on code

Why local?

Client code never leaves your machine. Local execution supports privacy, low marginal cost, offline iteration, and fits naturally into personal and client repositories.

Quick Links

Getting Started	Installation · Quick Start · Configuration
Reference	CLI Reference · Detectors · Model Providers
Advanced	Per-Detector Providers · Writing Custom Detectors · Scheduling
Workflow	Morning Report · Web UI · GitHub Issues