Empowering LLMs with Backstage: Broader Insights Driven by the Developer Portal
summary
Backstage hosts an LLM-backed agent that calls existing plugins as tools to fetch live catalog, Kubernetes and securityHub data, reasons over a high-severity finding (container not using readOnlyRootFilesystem), and then opens a GitHub issue — demo uses LangGraph, Helm-aware prompts, and a prototype securityHub plugin to get real-time findings. Architecture centers on tool use/function calling and guided flows; code is being open-sourced and the demo works end-to-end, creating issues from live data.
AI-written summary · about
chapters (23)
- 0:00Introduction
- 0:36Industry Direction
- 1:23The Shift
- 2:10Use Cases
- 3:11AWS and BMW
- 3:51What Backstage gives us
- 4:59Plugin Backchat
- 5:33Getting more value out of Backstage
- 6:14Building a chat assistant in Backstage
- 7:00Using a prototype Backstage plugin
- 7:44Resolving the big red text error
- 8:09Basic architecture
- 8:32The question
- 9:31The answer
- 10:00Rag AI
- 10:30Nexus System
- 11:07Tool Use
- 12:37Evolving Plugins
- 13:42Live Demo
- 20:13Dedicated Agent
- 22:00Conclusion
- 23:43Open sourcing the plugin
- 24:49Wrap up
related talks
21mLeveraging GenAI for Seamless Developer Experiences in Backstage
28mBuilding Open Developer Platforms with Backstage
08mLightning Talk: Maximising Developer Experience with Backstage: An Exploratio...
27mBuilding an Internal Developer Platform with Backstage Plugins
25mLLM Powered Agents with Kubernetes
30mBuilding Robust, Secure LLM and Agentic AI Workflows
27mSecuring The Gateway And Mitigating Risks In LLM API Integration
19mBackstage Plugins: Another way to communicate your data