Q: Can Bro fix issues automatically or does it just report?

Bro follows "show by default, change by request" principle. When you ask read-only questions ( "show logs" , "check status" ), it analyzes and reports — never changes anything. If it detects a problem, it recommends action but waits for approval. Example: finds 23 TypeError exceptions in logs → "Node.js crashed 3 minutes ago. Restart the service?" You confirm, it executes with preview. This prevents unwanted changes while maintaining speed.

Q: What operations can Bro execute?

Three layers: Layer 1 (Direct) : healthcheck, disk/memory/port checks, view logs, restart service/container, reboot server, clean disk, reload nginx, renew SSL cert. Layer 2 (Diagnostic recipes) : WebsiteDown, HighLatency, PostDeployIssue, DiskFull, ServerErrors, PreDeployCheck, ChangeAudit, PostDeployVerification. Layer 3 (Deploy) : full managed pipeline with rollback. All operations are validated and change operations require approval.

Question 1

How fast does Deployment Bro diagnose issues?

Accepted Answer

Diagnostic recipes typically complete in 15-45 seconds. Example: "website down" recipe runs 5 operations (check ports 80/443, validate nginx config, check SSL cert, analyze error logs, healthcheck), correlates results, and delivers root cause analysis. Real user data: memory leak detection averaged 15 seconds, post-deploy issue diagnosis 22 seconds, disk space analysis 18 seconds.

Question 2

What are diagnostic recipes exactly?

Accepted Answer

Pre-defined investigation sequences that Bro executes when you describe a problem. Not arbitrary AI chaining — predictable, testable workflows. Examples: WebsiteDown (checks ports → nginx config → SSL cert → error logs), HighLatency (checks load → memory → disk → slow query logs), PostDeployIssue (analyzes recent error logs → memory spikes → service status). AI's role: choose the right recipe, fill parameters from your natural language request, analyze results, explain in human terms.

Question 3

Can Bro fix issues automatically or does it just report?

Accepted Answer

Bro follows "show by default, change by request" principle. When you ask read-only questions ("show logs", "check status"), it analyzes and reports — never changes anything. If it detects a problem, it recommends action but waits for approval. Example: finds 23 TypeError exceptions in logs → "Node.js crashed 3 minutes ago. Restart the service?" You confirm, it executes with preview. This prevents unwanted changes while maintaining speed.

Question 4

How does the managed deploy pipeline work?

Accepted Answer

One-time setup: specify git repo, build command, services to restart, optional migration command. Then deploy with "deploy" in Telegram. Bro runs: (1) Pre-check (disk space, memory, server health — go/no-go decision), (2) Snapshot current state, (3) Git pull + merge, (4) Run migrations if configured, (5) Build (npm/yarn build, composer install, etc.), (6) Restart services, (7) Post-check (services running, app responding). If any step fails → automatic rollback to snapshot. Total time: 45-90 seconds typical.

Question 5

What operations can Bro execute?

Accepted Answer

Three layers: Layer 1 (Direct): healthcheck, disk/memory/port checks, view logs, restart service/container, reboot server, clean disk, reload nginx, renew SSL cert. Layer 2 (Diagnostic recipes): WebsiteDown, HighLatency, PostDeployIssue, DiskFull, ServerErrors, PreDeployCheck, ChangeAudit, PostDeployVerification. Layer 3 (Deploy): full managed pipeline with rollback. All operations are validated and change operations require approval.

Question 6

Can I use my own OpenAI/Anthropic API key?

Accepted Answer

Yes — BYOK (Bring Your Own Key) is supported. Deployment Bro uses Claude 3.5 Sonnet or GPT-4 for analysis (you choose). With BYOK, AI analysis costs ~$0.02-0.15 per diagnostic session depending on log volume. Without BYOK, AI analysis is included in the $99/mo subscription. BYOK gives you full control over AI provider, costs, and audit logs in your own account.

Question 7

Does it work with Docker, PM2, systemd?

Accepted Answer

Yes, all three. Bro detects what you're using and adapts. For Docker: container status, restart container, view container logs, resource usage. For systemd: service status, restart service, journalctl logs. For PM2: process list, restart apps, PM2 logs. You can mix them — e.g., nginx via systemd, app via PM2, database via Docker. Bro understands the full stack and correlates across process managers when diagnosing.

Параметр	Watchdog	SSH Mobile Clients	Grafana / Prometheus
Удобство на мобильном	Высокое (Чат)	Низкое (Консоль)	Среднее (Дашборды)
Самовосстановление	Да (авто-рестарт)	Нет	Нет
Активные действия	Да (кнопками)	Да (набор команд)	Нет
Сложность настройки	2 минуты	Ключи, VPN	Дни/Недели

Твои серверы. Всегда под присмотром.

ERROR: CONNECTION_TIMEOUT

Сбой в дороге

Потеря времени

Тревожность

Watchdog прикроет

Функции Watchdog

> /status

Все серверы — один тап

> /logs

Логи в кармане

> Авто-рестарт

Самовосстановление

> Триггеры

Правила: если X — делай Y

Initialization SETUP_PROTOCOL

STEP 01: Подключение

STEP 02: Агент

STEP 03: Контроль

SECURITY_AUDIT

✓ Требуется подтверждение

✓ Отсутствие входящих портов

✓ Audit logs

Watchdog vs альтернативы

Владей инфраструктурой.

Deployment Bro FAQ

Further Reading

Пусть Watchdog возьмёт рутину на себя.