Inspiration
Production outages happen at 3 AM when engineers are fast asleep. Standard Slack or email alerts are easily missed, leading to prolonged downtime. We created AutoPatch AI — Voice On-Call SRE Agent with CALL-E to turn passive notifications into real-time emergency voice calls that can wake up an engineer and execute immediate fixes.
What it does
AutoPatch AI monitors server health and infrastructure:
- Detects Anomalies: Captures 5xx errors, memory spikes, or database drops.
- Generates Patch: Uses Gemini AI to analyze root causes and generate fix recommendations.
- Emergency Voice Call: Uses CALL-E API to physically call the on-call engineer's mobile phone.
- Voice Authorization: Reads the emergency report out loud and requests voice/DTMF confirmation to apply the patch autonomously.
How we built it
- CALL-E REST API & SDK: Outbound voice dispatch and real-time speech interaction.
- Python / FastAPI: Event handling and incident automation workflow.
- Gemini API: Diagnostic analysis and patch generation.
Challenges we faced
Structuring prompt handling to ensure clear voice synthesis during emergency alerts and validating safe user authorization flows.
What's next for AutoPatch AI
- Multi-channel voice escalation for global DevOps teams.
- Integration with PagerDuty, Datadog, and OpsGenie workflows.
Built With
- ai-agent
- call-e
- fastapi
- gemini-api
- python
- sre
Log in or sign up for Devpost to join the conversation.