ResolveAI — Autonomous Incident Commander for Slack

An autonomous AI incident commander that classifies outages, retrieves historical context, coordinates diagnostics, proposes approved mitigations, and generates postmortems directly in Slack.

Comment 1

Inspiration

Operational incidents are stressful. When systems go down, SREs and developers waste critical minutes context-switching between monitoring tools, wiki runbooks, database consoles, and chat threads. We wanted to build a native Slack incident commander that manages the entire lifecycle safely, keeping humans in the loop for production changes, while leveraging autonomous AI diagnostics.

What it does

ResolveAI sits directly inside your Slack workspace as an interactive incident commander:

  1. Trigger Outages: Launch an incident using /incident <description>.
  2. Context-Aware Retrieval: Instantly matches the current outage characteristics against historical databases (e.g., recognizing credential rotation errors matching past incidents like INC-142).
  3. Hypothesize: Establishes possible root causes with dynamic confidence scoring.
  4. Safety Control: Requires explicit button approvals (Approve Plan / Execute Mitigation) before running production-altering workflows.
  5. Thread Diagnostics: Invites engineers to drop logs, metrics, or clues directly in the Slack thread. ResolveAI analyzes them in real-time to adjust its diagnostic path.
  6. Continuous Monitoring: Checks system health metrics before resolving.
  7. Postmortem Log: Automatically publishes a formatted incident review card in Slack showing timeline, root cause analysis, preventive actions, and reusable runbooks.

How we built it

  • Backend Engine: FastAPI built with Python 3, using SQLite and SQLAlchemy for state management.
  • Slack Integration: Slack Bolt framework running via Socket Mode for secure, firewall-friendly connection.
  • UI/UX: Custom Slack Block Kit layouts, with status updates, action buttons, and beautiful visual indicators.
  • Retrieval System: Dynamic similarity matching comparing new reports against historical runbooks.
  • Visuals: A high-fidelity real-time incident monitor web dashboard ready to deploy to Netlify.

Challenges we faced

Managing bidirectional communication in Slack threads while preserving a strict state machine transition (e.g. going from Triaging -> Investigating -> Resolving -> Closed) was a complex task. We designed a state machine validator to reject illegal operations and prevent message loops.

What we learned

We learned how powerful Slack's Block Kit is for creating structured workflow interfaces. When AI models output JSON payloads, rendering them into custom layouts with interactive controls elevates the user experience from a basic chatbot to a production-grade utility.

What's next for ResolveAI

We plan to hook up ResolveAI to live monitoring tools (Datadog, Prometheus) and cloud providers (AWS, GCP) to dynamically execute infrastructure rollback commands.

Built With

Share this project:

Updates