Harbor Maintainers Talk - Role of Harbor in AI, IOT Edge and the Future
summary
Harbor now stores OCI-packaged AI models as a first-class artifact (released in v2.13) using Model Pack (CNAI model spec) and the modctl build/push/pull workflow; the UI shows model family, tags, README and an internal filesystem view. Edge distribution uses Harbor Satellite, a single-binary embedded registry (~5–10 MB binary, ~10–15 MB idle RAM in demo) that pulls a central registry on a configurable interval (demo 10s) and reconciles deletions/updates via Ground Control state/config artifacts for eventual consistency. An experimental MCP server exposes agent-style management.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:51What is Harbor
- 2:54Harbor in AI ecosystem
- 3:35AI model storage
- 4:38Demo: AI model storage
- 7:07Harbor Satellite for IoT Edge
- 8:55Satellite architecture
- 11:25Demo: Harbor Satellite
- 13:51Satellite design decisions
- 14:56Harbor MCP server
- 15:40Demo: MCP server
- 16:44Q&A
related talks
27mHarbor Project - The Maintainers Session
28mHarbor - Enterprise Cloud Native Artifact Registry
33mHarbor - Enterprise Cloud Native Artifact Registry
43mDeep Dive: Harbor - Enterprise Cloud-native Artifact Registry
07mProject Lightning Talk: Harbor Project Update: Gitless GitOps, Satellite at the E...
36mHarbor
21mSponsored Demo: Cloud Native AI: Model Management with Harbor & Velero
34mDeep Dive: Harbor