Question 1

How does AI log analysis work?

Accepted Answer

When you ask to view logs or Bro runs a diagnostic recipe, it fetches logs (journalctl, docker logs, or log files), applies filters (time range, severity, keywords from your request), and sends to Claude/GPT for analysis. AI identifies: error patterns, correlation between events, root causes, severity assessment. Example: sees "504 Gateway Timeout" in nginx + "Slow query: 8.3s" in postgres logs 2 seconds earlier → concludes database is the bottleneck, not nginx. Analysis happens in 3-8 seconds typical.

Question 2

What's the "show by default, change by request" principle?

Accepted Answer

Critical safety rule confirmed by real usage data. When you ask read-only operations ("show logs", "check status", "what's using memory?"), Bro analyzes and reports — never makes changes even if it sees problems. If diagnostic detects issue, it recommends fix but waits for explicit confirmation. Example: you ask "show logs" → Bro finds 23 errors → "Found TypeError exceptions. Restart service?" → you confirm → it executes. This prevents AI from making unwanted changes while keeping UX fast.

Question 3

Can I see what Bro will do before it executes?

Accepted Answer

Yes, all change operations show preview + require approval. Examples: "Restart nginx" → "Going to restart nginx. Brief downtime expected. Proceed?" "Clean disk" → "Can remove: tmp files (~200MB), apt cache (~150MB), old logs (~500MB). Total ~850MB. Proceed?" "Deploy" → "Pre-check passed. Will: pull from main, run migrations, build, restart api + nginx. Snapshot created. Proceed?" You always see the plan before execution.

Question 4

How accurate are the diagnostic recipes?

Accepted Answer

Recipes are deterministic workflows, not free-form AI chaining. Accuracy comes from predefined investigation sequences tested on real production scenarios. AI's role is limited to: (1) choosing the right recipe based on your question, (2) extracting parameters (service name, time range, filters), (3) analyzing aggregated results. The operations themselves are standard Linux commands — no AI hallucination risk. Real user feedback: 94% of WebsiteDown diagnostics correctly identified root cause on first run.

Question 5

What happens if deploy fails mid-process?

Accepted Answer

Automatic rollback to snapshot. Deploy pipeline creates snapshot before starting (database state + code version + service configs). Each step is validated: pre-check ensures resources available, post-check verifies services started and app responds. If any step fails (build error, migration failure, service won't start, health check fails), Bro immediately rolls back to snapshot and reports what went wrong. Example: "Build failed: npm ERR! peer dependency conflict. Rolled back to previous version. All services restored."

Question 6

Can I customize diagnostic recipes?

Accepted Answer

v1 ships with 8 core recipes covering 90% of production incidents (WebsiteDown, HighLatency, PostDeployIssue, DiskFull, ServerErrors, PreDeployCheck, ChangeAudit, PostDeployVerification). Custom recipes are planned for v1.x based on user demand. Workaround for v1: you can chain operations manually via natural language ("check nginx status, then show last 100 error logs, then check SSL cert") — Bro executes sequentially and correlates results.

Question 7

Does it support multi-server management?

Accepted Answer

Deployment Bro tier supports up to 5 servers. In Telegram conversation, you specify which server: "check logs on production", "deploy to staging", "restart nginx on backup". Bro maintains context (if you checked logs on prod, next question defaults to prod unless you specify otherwise). Dashboard shows all servers, click to switch context. No need to prefix every command — context-aware routing.

Capability	Watchdog	Deployment Bro	Deployment Crew
How you interact	Buttons & /commands	Plain English	Plain English
Servers	1	3 (+$15/extra)	9 (+$15/extra)
Learning curve	Zero	Zero	Zero
Auto-restart	✓	✓	✓
Alerts	✓	✓	✓
Dashboard	—	✓	✓
Natural language	—	✓	✓
Smart diagnosis	—	✓	✓
Proactive monitoring	—	✓	✓
Auto-fix	Auto-restart	Simple issues	Simple issues
Git deploys	—	✓	✓
Deploy pipeline	—	✓	✓ + webhooks
Multi-messenger	—	—	✓
Team access	—	—	✓
Provisioning	—	—	✓
Audit retention	7 days	30 days	90 days
Support	Community	Email	Priority (24h)

From monitoring to autonomous DevOps.

Watchdog Mode

Deployment Bro

Quick comparison

Feature Deep Dive

Further Reading