​
How Maintenance Detection Works

Tembo connects directly to the tools already monitoring your stack — like Sentry, Github, and Postgres logs — and listens for issues as they arise. Instead of pinging engineers with alerts, it triages problems, traces root causes, and writes PRs to fix them.

Tembo’s pipeline looks like this:

  1. Ingest signals
    Pulls real-time errors, performance metrics, and traces from your observability stack.
  2. Analyze context
    Uses your application code, stack traces, and past issue patterns to understand what’s going wrong — and why.
  3. Generate solutions
    Writes potential code-level fixes tailored to your actual tech stack and repo structure.
  4. Submit a pull request
    Pushes clean, review-ready PRs into your existing Git workflows for approval.

​
Types of Maintenance Opportunities

The superhuman developer reacts to production issues wherever they surface.

​
Application Bugs

  • Errors from tools like Sentry (e.g. null references, uncaught exceptions)
  • Stack traces from crashes and exceptions
  • Runtime issues like type mismatches or broken logic

​
Performance Problems

  • Slow database queries from Postgres (via logs or pg_stat_statements)
  • Inefficient endpoints
  • N+1 queries and redundant computation
  • Memory or CPU spikes (via Datadog or OpenTelemetry)

​
Configuration & Best Practices

  • Misused libraries or deprecated patterns
  • Unsafe or unscalable defaults in app or infra code
  • Repetitive patterns Tembo can refactor or improve

​
Pull Request Generation

Once an issue is detected, Tembo:

  • Builds a clear, code-based fix
  • Writes a detailed pull request
  • Routes it through your GitHub workflows
  • Awaits your approval, just like a junior engineer would

​
Customizing Tembo’s Workflow

Tembo learns from your repo, team preferences, and past approvals. Your team can:

  • Set thresholds for which issues Tembo acts on
  • Enable/disable categories of issues (e.g. performance only, not style)
  • Define approval paths and codeowners for sensitive areas
  • Filter by severity or business impact