Repository navigation
Publish images for arm64 too; model server defaults to :edge - #110
Merged
Merged
Conversation
A model server on a Graviton instance could not pull the gateway image ("no
matching manifest for linux/arm64"): images were built for amd64 only. And
the guide's `:latest` moves only on release tags, so it was v0.4.0, from
before `agentfox serve models` existed.
- publish-images builds each image per architecture on a native runner
(ubuntu-24.04-arm for arm64, free for public repos), boot-checks the
gateway on its own architecture, pushes by digest, then one job per image
publishes a single multi-arch tag set and asserts both platforms are in it.
Also publishes `sha-<commit>` for every push, so a machine can pin exactly
what it tested.
- The model server defaults to `:edge` and says how to pin a sha- or release
tag; the guide covers Graviton sizing, disk, docker group, and not sharing
the token.
Verified the arm64 `deps` stage natively: aarch64, torch 2.14.1+cpu, and
`agentfox serve models` present.
Co-Authored-By: Claude <noreply@anthropic.com>
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why
Setting up the model server on a Graviton (arm64) EC2 instance failed in two ways:
docker compose pullfailed withno matching manifest for linux/arm64. Images were built for amd64 only.:latest, which moves only onv*tags. It was still v0.4.0 (Oct 12), from beforeagentfox serve modelsexisted.What
publish-images.yml: builds each image (gateway, dashboard) per architecture on a native runner (ubuntu-24.04-arm, free for public repos). It boot-checks the gateway on its own architecture and pushes by digest. Then onemanifestjob per image publishes a single multi-arch tag set and fails unless bothlinux/amd64andlinux/arm64are in it.sha-<commit>tag on every publish, so a machine can pin exactly what it tested.ci.yml's docker-smoke..env.exampleand compose default to:edge, with how to pin asha-or release tag. The guide adds Graviton sizing (c7g.large), a 30 GB disk minimum, the docker group step, and a note not to share the token.Verified
actionlintis clean.depsstage was built natively on Apple Silicon:aarch64, torch2.14.1+cpu, andagentfox serve models --helpworks.manifestjob asserts both platforms, so a missing architecture fails loudly rather than shipping amd64-only.After merge
ghcr.io/architsharm/agentfox/gateway:edgewill be pullable on arm64. Consider cutting a release (e.g.v0.5.0) so:latestalso includes the model server.🤖 Generated with Claude Code