AutoPatch AI

• 

  AutoPatch AI intercepting a ZeroDivisionError, generating a code fix via Gemini, and logging to Firestore.

• 

  Core Python code initializing Gemini SDK and Firestore to process error logs and request structured JSON fixes.

• 

  Firestore integration: Parsing Gemini response, saving run payload to 'taskmaster_runs', and printing execution telemetry.

• 

  Cloud Firestore console displaying real-time execution payload and FIXED status in taskmaster_runs collection.

Inspiration

Inspired by the endless, burn-out cycle of SRE teams. A system crash at 3 AM should be patched by an AI, not a human.

What it is

AutoPatch AI is your autonomous, background SRE agent, built for the Taskmaster track. It operates silently, listening for system failures.

How it works

When a crash occurs:

1. It automatically intercepts the error log.
2. It leverages Gemini Pro (Flash) to analyze the root cause and generate a clean, verifiable code patch.
3. It logs the entire execution state, including the fix, directly into Google Cloud Firestore.

Challenges

• Tuning Gemini Pro to consistently output valid code patches.
• Structuring the Firestore write from the LLM output.

This is fully automated, self-healing software.

Built With

• ai
• automation
• devops
• firebase
• firestore
• gemini-pro
• google-cloud
• python
• sre