hh | Devpost

Dashboard

Inspiration

• Thousands of logs & alerts every day make DevOps incident detection overwhelming. • Engineers often solve the same problems repeatedly, wasting valuable time. • Existing monitoring tools detect and notify, but don't remember past incidents or solutions. • We wanted an AI DevOps Assistant with persistent memory that never forgets previous incidents. • It uses past incident knowledge to give context-aware recommendations for new problems. • The system learns from every incident and engineer feedback. • Combines AI + Persistent Memory + Cloud to enable faster incident resolution and reduce downtime.

What it does

• AI analyzes production incidents and generates clear summaries. • Identifies possible root causes and recommends fixes. • Stores incidents, AI responses, recommendations, and feedback in CockroachDB as persistent memory. • Uses vector-based semantic search to find similar past incidents. • Uses historical solutions as context to improve new recommendations. • Engineer feedback helps the system continuously improve. • Interactive dashboard shows incident history, analytics, and system health. • Builds a continuously growing knowledge base that becomes smarter with every incident.

How we built it

• Frontend: React + Tailwind CSS + React Router for a fast, responsive dashboard. • API & UI: Axios for backend communication, Recharts for analytics, and Framer Motion for interactive UI. • Backend: Node.js + Express.js with REST APIs for incident management and AI communication. • Persistent Memory: CockroachDB Cloud stores incidents, AI summaries, root causes, feedback, and vector embeddings. • AI Layer: Amazon Bedrock analyzes incidents, retrieves historical context, and generates recommendations. • Cloud Storage: AWS S3 handles additional cloud storage requirements. • Core Workflow: Incident → AI Analysis → Store Memory → Semantic Search → Recommendation → User Feedback → Continuous Learning

Challenges we ran into

• Making memory the core feature instead of using the database only for storage. • Combining vector search with AI reasoning while maintaining fast response times. • Ensuring relevant incident retrieval instead of returning unrelated historical incidents. • Scaling AI, database, and cloud services without affecting system performance. • Designing a simple dashboard that clearly shows AI reasoning and memory without overwhelming users.

Accomplishments that we're proud of

• Built an AI-powered DevOps platform with persistent memory at its core. • Successfully used CockroachDB as long-term AI memory. • Implemented AI-driven incident analysis with context-aware recommendations. • Created a professional enterprise-style dashboard for real-world DevOps use. • Built a scalable and modular architecture for future expansion. • Demonstrated that historical memory makes AI recommendations more effective than stateless tools.

What we learned

• Persistent memory makes AI agents more effective in real-world production environments. • Historical context improves AI reasoning and makes recommendations more relevant. • Full-stack integration requires careful coordination between frontend, backend, database, and cloud. • Semantic search & vector embeddings are powerful for connecting similar incidents. • Engineer feedback is essential for improving AI recommendations over time. • AI + Memory + Human Feedback creates a stronger continuous-learning system.

What's next

• Advanced AI Agents: Integrate Amazon Bedrock Agents + AWS Lambda for automated agentic workflows. • Automatic Incident Ingestion: Connect CloudWatch, Grafana, Prometheus, and Datadog for real-time incident collection. • Scalable Vector Search: Expand semantic search using CockroachDB Distributed Vector Indexing. • Predictive Incident Detection: Detect and alert teams about potential failures before they happen. • Automated RCA & Reports: Generate root-cause analysis and post-incident reports automatically. • Collaboration Integrations: Connect with Slack, Microsoft Teams, GitHub, and Jira. • Enterprise Security: Add role-based access control and enterprise-grade security. • Long-Term Vision: Evolve into a production-ready AI Operations Center that continuously learns from incidents and helps engineers stay ahead of failures.