How it works
Two paths to a diagnosis — describe a problem when something's wrong, or let Watchdog detect and respond on its own
SigSentry turns a short description and a time window into a structured diagnosis. There are two ways to get one:
- Reactive — you describe a symptom, you get a diagnosis back. Use this when you already know something is wrong.
- Proactive — Watchdog watches your logs around the clock and runs the diagnosis automatically when a rule fires. Use this so the team hears about issues before customers do.
Most teams start with the reactive path and graduate to Watchdog rules once they've seen enough analyses to know which patterns matter.
Reactive — when something's wrong now
You give SigSentry two inputs and pick where to ask:
| Input | Notes |
|---|---|
| Description | A short statement of what's wrong. "Checkout errors" or "database timeouts in the last 30m" is enough |
| Time window | When to look — a few minutes, hours, or a specific window |
You can trigger a reactive analysis from the dashboard, the API, a chat command in Slack / Teams / Discord / Google Chat, or directly from an incoming support ticket.
A few seconds later you get back a structured diagnosis your team can scan at a glance:
- A one-paragraph summary of what happened
- Severity with a confidence score
- Root cause — service, error type, category
- Affected services and their roles
- A reconstructed timeline of relevant events
- Prioritized suggested actions
- Code correlation when a repo is connected — pointing at the likely offending file and the pull request that introduced the issue
You can generate a Markdown postmortem from any result, ask follow-up questions in a thread, and rate accuracy to feed your team's quality metrics over time.
Proactive — Watchdog monitors 24/7
If you'd rather know about issues before customers report them, set up Watchdog. It watches your log sources for the patterns you care about and, when a rule fires, runs a full analysis automatically and posts the diagnosis to your chosen channels — without anyone opening the dashboard.
- Configure — define rules targeting your log sources: error-rate thresholds, specific pattern matches, or anomaly detection. No code required. See Watchdog for the rule types and scheduling options.
- Detect — Watchdog runs around the clock and fires when a rule's conditions are met.
- Respond — a full analysis runs automatically, the diagnosis lands in Slack, Teams, Discord, or a webhook, and a support ticket can be opened — all while your team is still waking up.
What sharpens the diagnosis
Three configuration choices materially improve quality regardless of which path you use:
Connect a code repo
Adds code correlation — the diagnosis identifies the offending file and pull request.
Add project context
A short paragraph about your stack and known quirks tells the AI what matters in your logs.
Use precise time windows
Tighter windows give cleaner results. 'Last 30 minutes' beats 'last 24 hours' when you know roughly when an issue started.
Privacy
Your raw logs stay in your log aggregator. SigSentry reads them only for the duration of an analysis and doesn't store them. See our Privacy Policy for full data-handling details.
The Quick start walks through your first real diagnosis in five minutes.