DevGuard AI

Project : DevGuard AI Problem In modern software engineering, sudden production runtime crashes create major operational bottlenecks. When backend services fail, downtime rapidly increases and impacts business continuity. Traditionally, Site Reliability Engineers (SREs) and developers must manually inspect unstructured terminal logs, identify issues such as syntax errors, variable mismatches, or runtime exceptions, and then write and deploy fixes manually. This reactive debugging workflow is slow, error-prone, and increases operational latency during critical incidents. The Solution DevGuard AI is a 24/7 autonomous, self-healing DevOps SRE assistant that transforms reactive debugging into a proactive AI-driven recovery system. Instead of simply reporting failures, DevGuard AI automatically intercepts crashes, analyzes the root cause, generates fixes, and deploys hot-patches directly to the codebase in real time without requiring manual intervention. The system operates in a closed-loop simulation environment. When a Python backend application crashes due to an unhandled runtime exception or environment issue, DevGuard AI instantly captures the raw terminal crash logs (crash_log.txt). The unstructured logs are then processed using the advanced reasoning capabilities of the gemini-2.5-flash model through the Google GenAI SDK. Using strict Pydantic response schemas, the AI engine generates deterministic structured JSON outputs containing a detailed diagnosis and corrected production-ready code. The autonomous execution layer reads the JSON payload, updates the broken source files, deploys the hot-patch, and re-executes the application automatically. Within seconds, the system restores the server to a stable and fully operational state. Technical Stack • Python 3 for backend orchestration and file-system automation • Google GenAI SDK with gemini-2.5-flash for AI-driven reasoning and code remediation • Pydantic structured schemas for deterministic JSON outputs