Readable job names, latest tracks main, and an AGENTS.md

Job names in the Gitea UI rendered as the raw matrix expression
("${{ fromJSON(needs.select.outputs.images) }}"). Gitea resolves a job's matrix
when it PARSES the workflow, before `needs` outputs exist, so matrix.image was
interpolated against an unresolved matrix. There is no way to win with a dynamic
matrix: jobparser.nameWithMatrix interpolates a name containing "${{ }}", and
appends "(<values>)" to one that doesn't — either way the raw expression shows.
See go-gitea/gitea#28207.

So drop the matrix. build.yaml now runs one job, "Build changed images", that
loops over the selected images and emits ::group:: markers, giving a collapsible
section per image. Also: every step has an explicit static name, because Gitea
does not interpolate expressions in names either ("Log in to ${{ env.REGISTRY }}"
rendered literally).

Tag scheme, per review: `latest` now tracks main rather than the newest release,
and `edge` is gone — it's an Alpine/Traefik convention, not a broad standard, and
`main-<sha>` already covers "a specific commit". `latest` has exactly one owner
so a release tag and a main build can't race to define it. Release tags remain
immutable `:vX.Y.Z`/`:vX.Y`/`:vX` for pinning. This also means the ToolHive
manifest's `:latest` resolves as soon as this lands on main, with no release tag
needed first.

hack/docker-tags.sh folded into hack/build-images.sh, which is now the whole
pipeline — tag, build, smoke-test, push — shared by `make build` and CI. It uses
plain `docker build` instead of buildx, dropping setup-buildx-action: it loads
into the local store so the test runs pre-publish, and emits a plain manifest
with no attestations for Gitea's registry.

Verified in a simulated runner (repo in a docker volume, socket mounted, real
runner image): select + build + group markers, all four tag modes, and the
multi-image loop with a scratch second image.

AGENTS.md records the conventions and, importantly, the three Gitea gotchas that
all look fine locally: no bind-mounting the workspace into a sibling container,
no dynamic matrix, no expressions in names.

Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
This commit is contained in:
2026-08-09 12:05:23 -04:00
co-authored by Claude Opus 5
parent 7a81afec65
commit b888f09eb1
8 changed files with 336 additions and 130 deletions
+18 -38
View File
@@ -32,61 +32,41 @@ jobs:
images: ${{ steps.select.outputs.images }} images: ${{ steps.select.outputs.images }}
any: ${{ steps.select.outputs.any }} any: ${{ steps.select.outputs.any }}
steps: steps:
- uses: actions/checkout@v7 - name: Check out repository
uses: actions/checkout@v7
with: with:
fetch-depth: 0 # need history to diff against the base commit fetch-depth: 0 # need history to diff against the base commit
- id: select - name: Work out which images changed
id: select
run: ./hack/select-images.sh >> "$GITHUB_OUTPUT" run: ./hack/select-images.sh >> "$GITHUB_OUTPUT"
env: env:
DISPATCH_IMAGE: ${{ github.event.inputs.image }} DISPATCH_IMAGE: ${{ github.event.inputs.image }}
BASE_SHA: ${{ github.event.pull_request.base.sha || github.event.before }} BASE_SHA: ${{ github.event.pull_request.base.sha || github.event.before }}
# One job that loops, rather than a matrix job per image: Gitea resolves a matrix
# when it parses the workflow, before `needs` outputs exist, so a dynamic matrix
# produces a single job named after the raw expression. Looping keeps the job name
# readable and gives one collapsible log section per image.
build: build:
name: ${{ matrix.image }} name: Build changed images
needs: select needs: select
if: needs.select.outputs.any == 'true' if: needs.select.outputs.any == 'true'
runs-on: ubuntu-latest runs-on: ubuntu-latest
strategy:
fail-fast: false
matrix:
image: ${{ fromJSON(needs.select.outputs.images) }}
steps: steps:
- uses: actions/checkout@v7 - name: Check out repository
uses: actions/checkout@v7
- id: meta # Never on a pull request, which is exactly when nothing gets published.
run: ./hack/docker-tags.sh '${{ matrix.image }}' >> "$GITHUB_OUTPUT" - name: Log in to the Gitea container registry
env: if: github.event_name != 'pull_request'
PR_NUMBER: ${{ github.event.pull_request.number }}
- uses: docker/setup-buildx-action@v4
# Loaded rather than pushed, so the smoke test runs before anything is
# published. provenance is off because Gitea's registry rejects buildkit
# attestation manifests.
- name: Build
uses: docker/build-push-action@v7
with:
context: images/${{ matrix.image }}
load: true
tags: ${{ steps.meta.outputs.tags }}
provenance: false
build-args: |
VERSION=${{ steps.meta.outputs.version }}
REVISION=${{ github.sha }}
CREATED=${{ steps.meta.outputs.created }}
- name: Smoke test
run: make test IMAGE='${{ matrix.image }}' REF='${{ steps.meta.outputs.primary }}'
- name: Log in to ${{ env.REGISTRY }}
if: steps.meta.outputs.push == 'true'
uses: docker/login-action@v4 uses: docker/login-action@v4
with: with:
registry: ${{ env.REGISTRY }} registry: ${{ env.REGISTRY }}
username: ${{ env.DOCKER_USER }} username: ${{ env.DOCKER_USER }}
password: ${{ secrets.API_TOKEN }} password: ${{ secrets.API_TOKEN }}
- name: Push - name: Build, smoke-test and publish
if: steps.meta.outputs.push == 'true' run: ./hack/build-images.sh ${{ needs.select.outputs.images }}
run: printf '%s\n' '${{ steps.meta.outputs.tags }}' | xargs -r -n1 -t docker push env:
PR_NUMBER: ${{ github.event.pull_request.number }}
+3 -1
View File
@@ -8,9 +8,11 @@ on:
jobs: jobs:
lint: lint:
name: Lint
runs-on: ubuntu-latest runs-on: ubuntu-latest
steps: steps:
- uses: actions/checkout@v7 - name: Check out repository
uses: actions/checkout@v7
# Same targets you run locally with `make check lint`. # Same targets you run locally with `make check lint`.
- name: Repository layout - name: Repository layout
+156
View File
@@ -0,0 +1,156 @@
# AGENTS.md
Guidance for coding agents when working in this repository.
## Overview
A base image registry. Each image is self-contained in its own directory under
`images/` and is published to the **public** `rmcguire` namespace on Gitea at
`gitea.libretechconsulting.com/rmcguire/<image>`. Anonymous pulls work, so
consumers need no `imagePullSecret`.
```
images/<name>/Dockerfile the image
images/<name>/README.md required — `make check` fails without it
images/<name>/test.sh optional smoke test; must be executable to run
template/ skeleton to copy when adding an image
hack/ the three scripts CI and the Makefile share
Makefile single entry point for local work and CI
.gitea/workflows/ lint.yaml, build.yaml
```
## Core principle: do not pin, rebuild instead
Base image tags, apk package versions and tool releases are deliberately
unpinned. Rebuilding is how upstream updates land; the **published image tag** is
what pins things for consumers. Do not add `=version` to `apk add`, and do not
pin `FROM` to a digest.
Corollary: prefer an Alpine package over an upstream download when one exists —
one install mechanism beats two. Only fetch from upstream when Alpine does not
package the tool at all (`yq` is the sole example today). Where a major version
must be held back, constrain the major only and still track latest within it.
## Adding an image
```sh
cp -r template images/my-image
$EDITOR images/my-image/Dockerfile images/my-image/README.md images/my-image/test.sh
make build IMAGE=my-image # builds and smoke-tests
make check lint
```
No workflow changes are ever needed — `hack/select-images.sh` discovers any
directory under `images/` containing a `Dockerfile`.
Dockerfile conventions, all visible in `images/node-agent/Dockerfile`:
- Keep the `ARG VERSION/REVISION/CREATED` + `LABEL` block **last**, so changing
build metadata does not invalidate the layers above it.
- `# hadolint ignore=DL3018` above `apk add`, since versions are unpinned on
purpose. State the reason in a comment.
- Set `SHELL ["/bin/bash", "-o", "pipefail", "-c"]` before any `RUN` containing a
pipe (only after bash is installed).
- For a tool fetched from upstream, run it in the same layer (`yq --version`).
That is the check that matters: it catches a truncated download or an HTML
error page, which is the realistic failure mode.
## Tagging and releases
Tagging is derived from the git ref by `hack/build-images.sh`, identically for
every image in a run:
| Ref | Tags | Published |
| --- | --- | --- |
| push to `main` | `:latest`, `:main-<sha>` | yes |
| tag `<image>/vX.Y.Z` | `:vX.Y.Z`, `:vX.Y`, `:vX` | yes |
| pull request | `:pr-<n>` | no |
| anything else / local | `:dev` | no |
`latest` has exactly one owner — `main` — so a release tag and a `main` build can
never race to define it. Release tags are per image: `git tag node-agent/v1.0.0`
builds and publishes that image alone.
## Smoke tests
`images/<name>/test.sh` takes an image ref and must exit non-zero on failure. Two
conventions worth keeping, both in `images/node-agent/test.sh`:
- **Assert the deployment contract, not just tool presence.** node-agent runs
the checks under `--user 1000:1000 --read-only --tmpfs /tmp --cap-drop ALL`,
mirroring the securityContext it is actually deployed with, so CI catches
read-only or non-root breakage.
- **Do not `set -e` inside the container script.** Use a `fail` counter so every
check reports, then exit with it.
## Gitea Actions gotchas
These cost real debugging time. All three are Gitea behaviours, not bugs in this
repo, and all three will look fine locally.
**Never bind-mount the workspace into a sibling container.** The job itself runs
in a container and the checkout lives in a docker volume, not on the host. A
`docker run -v "$PWD":/repo` started through the host's docker socket resolves
that path against the **host**, where it does not exist — Docker silently creates
an empty directory, and the tool reports every file as missing. Stream files in
on stdin instead; `hack/lint.sh` does this for hadolint and shellcheck. Note
`docker build` and `docker run` without `-v` are unaffected, because the build
context is streamed over the API.
**Do not use a dynamic `fromJSON(needs...)` matrix.** Gitea resolves a job's
matrix when it *parses* the workflow, before `needs` outputs exist, so you get a
single job whose name is the raw uninterpolated expression. See
`nameWithMatrix` in Gitea's `modules/actions/jobparser` and
[go-gitea/gitea#28207](https://github.com/go-gitea/gitea/issues/28207). That is
why `build.yaml` loops over images inside one job and emits `::group::` markers
per image instead.
**Do not put `${{ }}` in a job or step `name:`.** Gitea does not interpolate
expression contexts in names — `name: Log in to ${{ env.REGISTRY }}` renders
literally. Give every step an explicit static `name:`, or it displays as the raw
`run:` command.
## Verifying
```sh
make all # check + lint + build + smoke-test; what CI does
make help # all targets
```
To reproduce the CI container environment locally — worth it before touching
anything in `hack/` or the workflows, since bind-mount and path assumptions only
break there:
```sh
docker volume create ci-sim
docker run --rm -v ci-sim:/w -v "$PWD":/src:ro alpine sh -c 'cp -a /src/. /w/'
docker run --rm -v ci-sim:/workspace/rmcguire/images \
-v /var/run/docker.sock:/var/run/docker.sock \
-w /workspace/rmcguire/images -e GITHUB_ACTIONS=true \
--entrypoint bash gitea/runner-images:ubuntu-latest -c 'make check lint'
docker volume rm ci-sim
```
The runner is a local Docker container (`gitea-gitea_runner-1`), config at
`~/work/docker/gitea/data/act_runner/config/config.yaml`. Its cached
`gitea/runner-images:ubuntu-latest` can go stale; current action majors all
declare `using: node24`, so refresh it with
`docker pull gitea/runner-images:ubuntu-latest` if JS actions misbehave.
When bumping action versions, check the real latest major rather than copying
from a sibling repo — several were a major behind.
## Downstream consumers
`node-agent` backs the shell MCP server in the **`50W/kube-manifests`** repo. When
its toolset changes, update all three of these or agents will not know what they
can run:
- `toolhive/mcpserver-shell.yaml` — the `WHAT'S IN THIS CONTAINER` comment block.
- `toolhive/mcptoolconfig-shell.yaml``toolsOverride.run_process.description`,
the tool description models actually read.
- `ai/kagent/remotemcpserver-shell.yaml``spec.description`.
That pod runs with no Kubernetes credentials, HTTPS-only egress, a read-only root
filesystem and all capabilities dropped, which is what `node-agent`'s smoke test
mirrors. Do not add a pull secret there — the registry is public.
+4 -7
View File
@@ -43,13 +43,10 @@ shellcheck: ## Lint every shell script
lint: hadolint shellcheck ## Run all linters lint: hadolint shellcheck ## Run all linters
build: ## Build IMAGE, or every image when IMAGE is unset build: ## Build and smoke-test IMAGE, or every image when IMAGE is unset
@for i in $(TARGETS); do \ @hack/build-images.sh $(TARGETS)
echo "==> building $$i"; \
docker build -t $(REGISTRY)/$(NAMESPACE)/$$i:dev images/$$i || exit 1; \
done
test: ## Smoke-test IMAGE, or every image when IMAGE is unset test: ## Smoke-test an already-built IMAGE, or every image when IMAGE is unset
@for i in $(TARGETS); do \ @for i in $(TARGETS); do \
if [ -x images/$$i/test.sh ]; then \ if [ -x images/$$i/test.sh ]; then \
images/$$i/test.sh "$(if $(REF),$(REF),$(REGISTRY)/$(NAMESPACE)/$$i:dev)" || exit 1; \ images/$$i/test.sh "$(if $(REF),$(REF),$(REGISTRY)/$(NAMESPACE)/$$i:dev)" || exit 1; \
@@ -58,4 +55,4 @@ test: ## Smoke-test IMAGE, or every image when IMAGE is unset
fi; \ fi; \
done done
all: check lint build test ## Everything CI does, locally all: check lint build ## Everything CI does, locally (build smoke-tests too)
+17 -12
View File
@@ -23,12 +23,15 @@ The registry is public, so no pull secret is required.
## Tags ## Tags
| Tag | Points at | | Tag | Points at | Set by |
| --- | --- | | --- | --- | --- |
| `vX.Y.Z` / `vX.Y` / `vX` | a released build | | `latest` | the current tip of `main` | push to `main` |
| `latest` | the most recent release | | `main-<sha>` | one specific commit on `main` | push to `main` |
| `edge` | the current tip of `main` | | `vX.Y.Z` / `vX.Y` / `vX` | an immutable released build | tag `<image>/vX.Y.Z` |
| `main-<sha>` | one specific commit on `main` | | `pr-<n>` | a pull request build — built, never published | pull request |
`latest` has exactly one owner — `main` — so a release tag and a `main` build can
never race to define it. Pin to `vX.Y.Z` when you want a build that will not move.
Releases are cut per image by pushing a tag named `<image>/vX.Y.Z`, which builds Releases are cut per image by pushing a tag named `<image>/vX.Y.Z`, which builds
and publishes that image alone: and publishes that image alone:
@@ -51,23 +54,25 @@ cp -r template images/my-image
here (`FROM gitea.libretechconsulting.com/rmcguire/node-agent:latest`). here (`FROM gitea.libretechconsulting.com/rmcguire/node-agent:latest`).
2. Fill in `images/my-image/README.md`. 2. Fill in `images/my-image/README.md`.
3. Extend `images/my-image/test.sh` to assert whatever the image promises. 3. Extend `images/my-image/test.sh` to assert whatever the image promises.
4. `make build test IMAGE=my-image` 4. `make build IMAGE=my-image`
CI needs no changes — it discovers every directory under `images/` that contains CI needs no changes — it discovers every directory under `images/` that contains
a `Dockerfile`. a `Dockerfile`. See [AGENTS.md](AGENTS.md) for the full conventions.
## Local development ## Local development
```sh ```sh
make help # list targets make help # list targets
make all # everything CI does make all # everything CI does
make build test IMAGE=node-agent # one image make build IMAGE=node-agent # build and smoke-test one image
``` ```
## CI ## CI
Two Gitea workflows, both of which just call the `make` targets above: Two Gitea workflows, both of which just call the `make` and `hack/` entry points
above, so a green `make all` locally means a green pipeline:
- **`lint.yaml`** — repository layout, `hadolint`, `shellcheck`. - **`lint.yaml`** — repository layout, `hadolint`, `shellcheck`.
- **`build.yaml`** — builds only the images whose files changed, smoke-tests each - **`build.yaml`** — works out which images changed, then builds, smoke-tests and
one before anything is published, then pushes on `main` or a release tag. (on `main` or a release tag) publishes each one. Nothing is pushed until its
smoke test passes.
+121
View File
@@ -0,0 +1,121 @@
#!/usr/bin/env bash
# Build, smoke-test and — in CI, on a publishing ref — push images.
#
# Usage: build-images.sh <image>...
#
# How an image gets tagged is a property of the git ref, not of the image, so one
# run treats every image it builds the same way:
#
# tag <image>/vX.Y.Z -> :vX.Y.Z :vX.Y :vX published release
# push to main -> :latest :main-<sha> latest tracks main
# pull request -> :pr-<n> built, never pushed
# anywhere else -> :dev built, never pushed (local default)
#
# `latest` has exactly one owner (main), so a release tag and a main build can never
# race to define it.
#
# Uses plain `docker build` rather than buildx with a container driver: it loads
# straight into the local image store so the smoke test can run before anything is
# published, and it produces a plain manifest with no attestations, which Gitea's
# registry rejects.
set -euo pipefail
cd "$(dirname "$0")/.."
REGISTRY=${REGISTRY:-gitea.libretechconsulting.com}
NAMESPACE=${NAMESPACE:-rmcguire}
[ $# -gt 0 ] || {
echo "usage: ${0##*/} <image>..." >&2
exit 2
}
sha=$(git rev-parse --short=7 HEAD)
revision=$(git rev-parse HEAD)
created=$(date -u +%Y-%m-%dT%H:%M:%SZ)
push=false
if [ "${GITHUB_REF_TYPE:-}" = tag ]; then
mode="release"
version=${GITHUB_REF_NAME##*/}
[[ $version =~ ^v[0-9]+\.[0-9]+\.[0-9]+$ ]] || {
echo "tag '${GITHUB_REF_NAME}' is not of the form <image>/vX.Y.Z" >&2
exit 1
}
push=true
elif [ "${GITHUB_EVENT_NAME:-}" = pull_request ]; then
mode="pr"
version="pr-${PR_NUMBER:-0}"
elif [ "${GITHUB_REF_NAME:-}" = main ]; then
mode="main"
version="main-$sha"
push=true
else
mode="local"
version=dev
fi
tags_for() {
local repo="$REGISTRY/$NAMESPACE/$1" n
case $mode in
release)
n=${version#v}
printf '%s\n' "$repo:$version" "$repo:v${n%.*}" "$repo:v${n%%.*}"
;;
main) printf '%s\n' "$repo:latest" "$repo:$version" ;;
*) printf '%s\n' "$repo:$version" ;;
esac
}
# Collapsible sections in the Actions log; plain headers when run locally.
group() {
if [ -n "${GITHUB_ACTIONS:-}" ]; then
echo "::group::$*"
else
echo "==> $*"
fi
}
endgroup() {
if [ -n "${GITHUB_ACTIONS:-}" ]; then
echo "::endgroup::"
fi
}
echo "mode=$mode version=$version push=$push images=$*"
for image in "$@"; do
[ -f "images/$image/Dockerfile" ] || {
echo "no such image: images/$image/Dockerfile does not exist" >&2
exit 1
}
mapfile -t tags < <(tags_for "$image")
tag_args=()
for tag in "${tags[@]}"; do tag_args+=(--tag "$tag"); done
group "build $image"
docker build "${tag_args[@]}" \
--build-arg "VERSION=$version" \
--build-arg "REVISION=$revision" \
--build-arg "CREATED=$created" \
"images/$image"
endgroup
if [ -x "images/$image/test.sh" ]; then
group "test $image"
"images/$image/test.sh" "${tags[0]}"
endgroup
else
echo "note: $image has no test.sh, nothing to smoke-test"
fi
if [ "$push" = true ]; then
group "push $image"
for tag in "${tags[@]}"; do
echo "pushing $tag"
docker push "$tag"
done
endgroup
fi
done
echo "done: $* ($mode)"
-54
View File
@@ -1,54 +0,0 @@
#!/usr/bin/env bash
# Work out how one image should be tagged, and whether it should be published.
# Prints GITHUB_OUTPUT lines:
#
# tags=<multiline> every ref to build, one per line
# primary=<ref> the ref the smoke test runs against
# version=<string> org.opencontainers.image.version
# created=<rfc3339> org.opencontainers.image.created
# push=true|false whether these refs get published
#
# Tag scheme:
# tag <image>/vX.Y.Z -> :vX.Y.Z :vX.Y :vX :latest published release
# push to main -> :edge :main-<sha> tip of main
# pull request -> :pr-<n> built, never published
# anything else -> :dev-<sha> built, never published
set -euo pipefail
image=${1:?usage: docker-tags.sh <image>}
repo="${REGISTRY:?REGISTRY is not set}/${NAMESPACE:?NAMESPACE is not set}/${image}"
sha=$(git rev-parse --short=7 HEAD)
created=$(date -u +%Y-%m-%dT%H:%M:%SZ)
push=false
if [ "${GITHUB_REF_TYPE:-}" = tag ]; then
version=${GITHUB_REF_NAME##*/}
[[ $version =~ ^v[0-9]+\.[0-9]+\.[0-9]+$ ]] || {
echo "tag '${GITHUB_REF_NAME}' is not of the form <image>/vX.Y.Z" >&2
exit 1
}
n=${version#v}
tags=("$repo:$version" "$repo:v${n%.*}" "$repo:v${n%%.*}" "$repo:latest")
push=true
elif [ "${GITHUB_EVENT_NAME:-}" = pull_request ]; then
version="pr-${PR_NUMBER:-0}"
tags=("$repo:$version")
elif [ "${GITHUB_REF_NAME:-}" = main ]; then
version="edge-$sha"
tags=("$repo:edge" "$repo:main-$sha")
push=true
else
version="dev-$sha"
tags=("$repo:$version")
fi
printf 'version %s, push %s, tags: %s\n' "$version" "$push" "${tags[*]}" >&2
echo "tags<<__TAGS__"
printf '%s\n' "${tags[@]}"
echo "__TAGS__"
echo "primary=${tags[0]}"
echo "version=$version"
echo "created=$created"
echo "push=$push"
+14 -15
View File
@@ -1,9 +1,16 @@
#!/usr/bin/env bash #!/usr/bin/env bash
# Decide which images CI should build. Prints GITHUB_OUTPUT lines: # Decide which images CI should build. Prints GITHUB_OUTPUT lines:
# #
# images=["node-agent",...] JSON array consumed by the build job's matrix # images=node-agent other-image space-separated, fed straight to build-images.sh
# any=true|false whether there is anything to build at all # any=true|false whether there is anything to build at all
# #
# Space-separated rather than a JSON matrix on purpose. Gitea resolves a job's matrix
# when it parses the workflow — before `needs` outputs exist — so a dynamic
# `fromJSON(needs...)` matrix yields one job whose name is the raw, uninterpolated
# expression. See jobparser.nameWithMatrix in the Gitea source, and
# https://github.com/go-gitea/gitea/issues/28207. Looping inside one job gives an
# honest job name and collapsible per-image log sections instead.
#
# Rules, first match wins: # Rules, first match wins:
# 1. workflow_dispatch naming one image -> that image # 1. workflow_dispatch naming one image -> that image
# 2. tag push (<image>/vX.Y.Z) -> the image named in the tag # 2. tag push (<image>/vX.Y.Z) -> the image named in the tag
@@ -27,21 +34,13 @@ all_images() {
done done
} }
# Names on stdin -> ["a","b"]. Built by hand so the runner needs no jq.
as_json() {
local out='' name
while IFS= read -r name; do
[ -n "$name" ] || continue
out="${out:+$out,}\"$name\""
done
printf '[%s]' "$out"
}
emit() { emit() {
local names=$1 reason=$2 any=false local names=$1 reason=$2 any=false flat
[ -n "$names" ] && any=true # Collapse the newline-separated list onto one line for the workflow output.
echo "selected (${reason}): ${names:-<none>}" >&2 flat=$(printf '%s\n' "$names" | tr '\n' ' ' | sed -e 's/ */ /g' -e 's/^ //' -e 's/ $//')
printf 'images=%s\n' "$(printf '%s\n' "$names" | as_json)" [ -n "$flat" ] && any=true
echo "selected (${reason}): ${flat:-<none>}" >&2
printf 'images=%s\n' "$flat"
printf 'any=%s\n' "$any" printf 'any=%s\n' "$any"
exit 0 exit 0
} }