Open-Source Framework for AI-Powered CI/CD Pipelines
summary
Claimed outcome: model-generated code is only ~10% of the work and the other 90% is structure, validation, cost controls and human gates. Built an open-source Continuous Agent Deployment using GitHub as runtime, a central dispatcher that pulls cards across board columns with WIP limits, station-specific agents on GitHub runners, model configs (4.8 and 4.6 used), three auto-retry test attempts, per-agent branches, cost/observability dashboards, and a plain deterministic deploy that successfully runs in production.
AI-written summary · about
chapters (14)
- 0:00Pipeline With No People
- 1:03Why CAD Matters
- 2:45The Eight Station Board
- 6:40GitHub As Runtime
- 7:38Dispatcher And WIP Pull
- 10:27Configurable Models Costs
- 14:07Station Contracts Prompts
- 16:47Extending With New Stages
- 18:46Reliability Tests Retries
- 22:36Human Gate Deterministic Deploy
- 24:12Observability And Cost Metrics
- 25:20Failure Modes And Limits
- 27:05Why This Is A Framework
- 28:06Wrap Up And Open Source
related talks
21mAI-driven optimization of CI/CD pipelines
44mAgentic Delivery, Guardrailed: AI in CI/CD & Platform Engineering (Open Source Patterns)
40mSecure by Default: Building an AI-Augmented, OSS-Powered Reusable CI/CD Pipeline
36mHow to Safely Deploy AI Agents that Write and Manage your IaC
51mBeyond Code Generation: Governing AI Agents Across the DevSecOps Lifecycle
30mIntelligent Pipelines: Bringing AI/ML into DevSecOps Automation
26mAccelerating DevOps with AI
23mBuilding a Secure By-Design Pipeline with an Open-Source Stack