Featuer/gitops extra command (#47)

Co-authored-by: moilanik <niko.moilanen@tietoevry.com>
Reviewed-on: #47

.gitea/workflows/ci-container-build-push.yml

@@ -23,7 +23,7 @@ on:
 
 jobs:
   build-push:
-    runs-on: ubuntu-latest
+    runs-on: docker
     steps:
       - uses: actions/checkout@v4
 

.gitea/workflows/docker-build-push.yml

@@ -32,7 +32,7 @@ concurrency:
 
 jobs:
   build-push:
-    runs-on: ubuntu-latest
+    runs-on: docker
     steps:
       - uses: actions/checkout@v4
       - uses: actions/checkout@v4
@@ -91,7 +91,7 @@ jobs:
           fi
 
   tag-commit:
-    runs-on: ubuntu-latest
+    runs-on: docker
     needs: [build-push]
     steps:
       - uses: actions/checkout@v4

.gitea/workflows/gitops-dispatch.yml

@@ -17,6 +17,9 @@ on:
       GITOPS_REPO:
         required: true
         type: string
+      GITOPS_EXTRA_CMD:
+        required: false
+        type: string
     secrets:
       GITOPS_DISPATCH_TOKEN:
         required: true
@@ -34,6 +37,7 @@ env:
   GITOPS_SOURCE_COMMIT: ${{ github.sha }}
   GITEA_API_URL: ${{ fromJson(inputs.env_json).GITEA_API_URL }}
   GITOPS_TAG_PREFIX: ${{ fromJson(inputs.env_json).GIT_TAG_PREFIX || '' }}
+  GITOPS_EXTRA_CMD: ${{ inputs.GITOPS_EXTRA_CMD || '' }}
   GITOPS_WORKFLOW: gitops-service.yaml
   GITOPS_DISPATCH_TIMEOUT: 30
 
@@ -49,19 +53,28 @@ jobs:
           ID=$(date +%s | md5sum | head -c 8)
           echo "dispatch_id=$ID" >> "$GITHUB_OUTPUT"
 
+      - name: Resolve commit author
+        id: author
+        run: |
+          echo "name=$(git log -1 --format='%an')" >> "$GITHUB_OUTPUT"
+          echo "email=$(git log -1 --format='%ae')" >> "$GITHUB_OUTPUT"
+
       - name: Dispatch to GitOps repo
         env:
           GITEA_TOKEN: ${{ secrets.GITOPS_DISPATCH_TOKEN }}
         run: |
           INPUTS=$(jq -nc \
             --arg dispatch_id "${{ steps.gen.outputs.dispatch_id }}" \
+            --arg author_name "${{ steps.author.outputs.name }}" \
+            --arg author_email "${{ steps.author.outputs.email }}" \
             --arg file "$GITOPS_FILE" \
             --arg yq_tpl "$GITOPS_YQ_TPL" \
             --arg version "$GITOPS_VERSION" \
             --arg source_repo "$GITOPS_SOURCE_REPO" \
             --arg source_commit "$GITOPS_SOURCE_COMMIT" \
             --arg git_tag_prefix "${GITOPS_TAG_PREFIX:-}" \
-            '{dispatch_id: $dispatch_id, file: $file, yq_tpl: $yq_tpl, version: $version, source_repo: $source_repo, source_commit: $source_commit, git_tag_prefix: $git_tag_prefix}')
+            --arg extra_cmd "${GITOPS_EXTRA_CMD:-}" \
+            '{dispatch_id: $dispatch_id, author_name: $author_name, author_email: $author_email, file: $file, yq_tpl: $yq_tpl, version: $version, source_repo: $source_repo, source_commit: $source_commit, git_tag_prefix: $git_tag_prefix, extra_cmd: $extra_cmd}')
           curl -s -X POST \
             "${GITEA_API_URL}/api/v1/repos/${GITOPS_REPO}/actions/workflows/${GITOPS_WORKFLOW}/dispatches" \
             -H "Authorization: token $GITEA_TOKEN" \

README.md

@@ -151,58 +151,102 @@ ja sitä kautta Gitea ei tarvitse päivityksessä mitään temppuja.
 Päivityksen jälkeen muista tappaa pod (käynnistyy automaattisesti uudelleen), että lataa varmasti kaikki uudesta. Sillä ConfigMap tms eivät lataudu
 mikäli pod jatkaa ajamista.
 
+Klusterissa on kaksi StatefulSetiä, joilla on eri labelit:
+- **`act-runner`** — yleisrunneri (label `ubuntu-latest`). DinD-sidecar on olemassa mutta idle — `require_docker: false` estää runneria käyttämästä sitä.
+- **`act-runner-docker`** — Docker-buildien runneri (label `docker`). DinD on aktiivinen, `require_docker: true`.
+
+Vain `docker-build-push.yml` ja `ci-container-build-push.yml` käyttävät `docker`-labelia.
+Kaikki muut workflowt (testit, lintit, helm-publish, gitops-dispatch) ajetaan `ubuntu-latest`-runnerilla.
+Steppien suoritus ei mene Docker Daemonin läpi, joten konekielisiä kontteja luodaan suoraan K8s-runtimella.
+
 ```bash
 helm repo add gitea https://dl.gitea.com/charts
 helm repo update
 
+# 1. Yleisrunneri (DinD idle)
 helm upgrade --install act-runner gitea/actions \
   --set enabled=true \
   --set giteaRootURL="$GITEA_URL" \
   --set existingSecret=act-runner-token \
   --set existingSecretKey=token \
-  --set statefulset.replicas=3 \
+  --set statefulset.replicas=2 \
   --set statefulset.runner.tag=1.0.8 \
   --set statefulset.dind.tag=29.5.2-dind \
+  --set statefulset.dind.resources.requests.memory=250Mi \
+  --set statefulset.dind.resources.limits.memory=750Mi \
   --set-string 'statefulset.runner.config=log:
   level: info
 cache:
-  enabled: false
+  enabled: true
 container:
-  require_docker: true
+  require_docker: false' \
-  docker_timeout: 300s' \
   --namespace "$GITEA_ACTIONS_NAMESPACE" \
   --create-namespace
+
+# 2. Docker-runner (DinD)
+helm upgrade --install act-runner-docker gitea/actions \
+  --set enabled=true \
+  --set giteaRootURL="$GITEA_URL" \
+  --set existingSecret=act-runner-token \
+  --set existingSecretKey=token \
+  --set statefulset.replicas=1 \
+  --set statefulset.runner.tag=1.0.8 \
+  --set statefulset.dind.tag=29.5.2-dind \
+  --set statefulset.dind.resources.requests.memory=250Mi \
+  --set statefulset.dind.resources.limits.memory=750Mi \
+  --set-string 'statefulset.runner.config=log:
+  level: info
+cache:
+  enabled: true
+container:
+  require_docker: true
+  docker_timeout: 300s
+runner:
+  labels:
+    - "docker:docker://catthehacker/ubuntu:act-latest"' \
+  --namespace "$GITEA_ACTIONS_NAMESPACE"
 ``` 
 
-path escapes from parent -bugi korjattiin Docker 29.5.2:ssa. Tämän teko aikana default on 29.5.1 — juuri tämän alle jäävä versio.
+`path escapes from parent` -bugi korjattiin Docker 29.5.2:ssa. Tämän teko aikana default on 29.5.1 — juuri tämän alle jäävä versio.
 
 Oletus-lokitaso on `debug` — suositeltu `info`. Näkee jobien aloitukset ja valmistumiset ilman konttikerrosten purkua (Downloading/Extracting-spämmiä). `debug` on tarpeen vain vianselvityksessä.
 
+`cache.enabled: true` nappaa image-cachen käyttöön — ilman sitä jokainen ajo lataa konttikuvat uudestaan.
+
 #### Docker (DinD)
 
-Helm chart deployaa DinD:n init-sidecarina (`docker:dind` samassa podissa).
+> **Huomio:** Gitea 1.26.x ei tue vielä label-pohjaista runner-valintaa.
-`require_docker: true` kytkee jobit siihen — erillistä DinD-asennusta ei tarvita.
+> `runs-on: ubuntu-latest`-jobi saattaa päätyä `docker`-labeliselle runnerille.
+> Bugi on tunnettu Gitean FAQ:ssa — korjaus tulossa myöhemmässä versiossa.
+> Katso: [docs.gitea.com/usage/actions/faq](https://docs.gitea.com/usage/actions/faq)
 
-**DinD-tag pinottu:** `29.5.2-dind` (ei chart-oletusta). Docker 29.5.1 aiheuttaa act-runnerissa
+Helm chart deployaa DinD:n init-sidecarina (`docker:dind` samassa podissa).
-`path escapes from parent` -virheen job-kontin käynnistyksessä.
+Molemmissa StatefulSetissä on DinD-sidecar, mutta:
+- **`act-runner`**: `require_docker: false` → runner ei käytä DinD:tä lainkaan, steppien suoritus menee suoraan K8s-runtimella
+- **`act-runner-docker`**: `require_docker: true` → runner luo steppikontit DinD:n kautta (tarvitaan `docker build` -komentoja varten)
+
+**DinD-tag pinottu:** `29.5.2-dind` molemmissa (ei chart-oletusta `29.5.1-dind`). Docker 29.5.1 aiheuttaa act-runnerissa
+`path escapes from parent` -virheen job-kontin käynnistyksessä sekä `mkdirat var/run: file exists` -virheen tiedostojen kopioinnissa.
 
 Maven/npm-ajot käyttävät vain workflow'n `container:`-imagea; DinD tarvitaan vasta Docker-buildissä.
 
-### 3. Varmista
+### 5. Varmista
 
 ```bash
 kubectl get pods -n gitea-actions
 # → act-runner-runner-0  Running
+# → act-runner-docker-runner-0  Running
 
-kubectl exec -n gitea-actions act-runner-runner-0 -c dind -- docker version
+kubectl exec -n gitea-actions act-runner-docker-runner-0 -c dind -- docker version
 # → Server Version: 29.5.2 (tai uudempi)
 ```
 
-Gitean puolella runner ilmestyy Active-tilaan pienellä viiveellä:
+Gitean puolella runnerit ilmestyvät Active-tilaan pienellä viiveellä:
 
 ```
 Site Admin → Actions → Runners  (tai Org → Settings → Actions → Runners)
 # → act-runner-runner-0  Active  ubuntu-latest
+# → act-runner-docker-runner-0  Active  docker
 ```
 
 Tämän jälkeen `.gitea/workflows/ci.yml` triggeröityy automaattisesti pushista.
@@ -279,4 +323,7 @@ Tarkka asennus: [skills/gitops-update/SKILL.md](skills/gitops-update/SKILL.md)
 | `existingSecret` | Kubernetes secretin nimi, jossa token |
 | `existingSecretKey` | Avain secretin sisällä |
 | `statefulset.dind.tag` | DinD-image tag (`29.5.2-dind` minimi) |
+| `statefulset.dind.resources.requests.memory` | DinD muistirequest (suositus `250Mi`) |
+| `statefulset.dind.resources.limits.memory` | DinD muistilimitti (suositus `750Mi`) |
 | `statefulset.runner.labels` | Mukautetut labelit |
+

git-pages/README.md

@@ -149,3 +149,42 @@ curl -X PATCH https://ci-reports.helm-dev.keskikuja.site/owner/repo/commit/sha8/
 - `git-pages-publish-token` = plaintext (luetaan Giteaan viedessä)
 
 Tarkemmat secret-ohjeet: [docs/secrets.md](docs/secrets.md).
+
+---
+
+## Testaus
+
+Retention-logiikalle on unit-testit, jotka testaa funktiot ja Phase 3 -säännöt
+erikseen ilman ulkoisia riippuvuuksia.
+
+```bash
+cd git-pages
+bats tests/retention.bats
+```
+
+Testit käyttävät `<root>/files/retention-lib.sh` -jaettua kirjastoa, jota myös
+`retention-cleanup.sh` sourceaa. Uutta testiä kirjoittaessa:
+
+1. Luo config `write_config`-helperilla
+2. Täytä `KEEP`-array testidatalla (muoto: `dir|owner|repo|branch|days`)
+3. Kutsu `apply_retention "$CONFIG"`
+4. Tarkista `TO_DELETE`-array ja `$output`
+
+**Vaatimukset:** `bats`, `jq`, `date` (GNU date tai BSD date ISO 8601 -tuella).
+
+---
+
+## Retention
+
+Ylläpitoscripti, joka poistaa vanhat raportit git-pagesista retention‑sääntöjen mukaan.
+Ajetaan sidecar‑ tai cronjob‑tilassa Kubernetesissa.
+
+### Air gap -yhteensopimattomuus
+
+Retention‑kontti asentaa tarvitsemansa työkalut (`curl`, `jq`) ajon aikana
+packagemanagerilla (`apt-get` / `apk`). Tämä **ei toimi air gap -ympäristössä**,
+jossa konttirekisteriin tai pakettivarastoihin ei ole verkkoyhteyttä.
+
+**TODO:** Rakenna custom Docker‑image, jossa deps on valmiina:
+`FROM alpine:latest && apk add --no-cache curl jq`.
+Pushaa omaan rekisteriin ja päivitä `values.yaml`:n `retention.image`.

git-pages/files/retention-cleanup.sh

@@ -13,75 +13,11 @@ curl_with_host() {
 
 [ -f "$CONFIG" ] || { echo "ERROR: config missing: $CONFIG" >&2; exit 1; }
 
-declare -A BRANCH_CACHE
+declare -A REPO_BRANCHES_CACHE
-branch_exists() {
+declare -A REPO_STATUS
-  local owner="$1" repo="$2" branch="$3" key="${owner}/${repo}/${branch}"
-  local status attempt
 
-  [ -z "$GITEA_API_URL" ] && return 0
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-  [ -z "$GITEA_TOKEN" ] && return 0
+source "$SCRIPT_DIR/retention-lib.sh"
-  
-  if [ "${BRANCH_CACHE[$key]:-}" = "1" ]; then
-    return 0
-  fi
-  
-  # Retry up to 2 times on API errors (hardcoded)
-  for attempt in 1 2 3; do
-    status=$(curl -sS -o /dev/null -w "%{http_code}" \
-      -H "Authorization: token ${GITEA_TOKEN}" \
-      "${GITEA_API_URL}/api/v1/repos/${owner}/${repo}/branches/${branch}" 2>/dev/null || echo "000")
-    
-    if [ "$status" = "200" ]; then
-      BRANCH_CACHE[$key]=1
-      return 0
-    fi
-    
-    if [ "$status" = "404" ]; then
-      return 1
-    fi
-    
-    # API error - retry if not last attempt
-    if [ "$attempt" -lt 3 ]; then
-      sleep 10
-      continue
-    fi
-  done
-  
-  # All retries failed - keep report (fail-safe)
-  echo "  WARN: Gitea API error for ${owner}/${repo}/${branch} (status ${status}) after 3 attempts - KEEPING report"
-  BRANCH_CACHE[$key]=1
-  return 0
-}
-
-default_max_age=$(jq -r '.branches.default.maxAgeDays // 90' "$CONFIG")
-default_keep_min=$(jq -r '.branches.default.keepMin // 5' "$CONFIG")
-
-rule_max_age() {
-  local branch="$1" v
-  v=$(jq -r --arg b "$branch" '.branches[$b].maxAgeDays // empty' "$CONFIG")
-  [ -n "$v" ] && echo "$v" || echo "$default_max_age"
-}
-
-rule_keep_min() {
-  local branch="$1" v
-  v=$(jq -r --arg b "$branch" '.branches[$b].keepMin // empty' "$CONFIG")
-  [ -n "$v" ] && echo "$v" || echo "$default_keep_min"
-}
-
-age_days() {
-  local published="$1" epoch_pub now
-  epoch_pub=$(date -u -d "$published" +%s 2>/dev/null || echo 0)
-  [ "$epoch_pub" -eq 0 ] && echo 99999 && return
-  now=$(date -u +%s)
-  echo $(( (now - epoch_pub) / 86400 ))
-}
-
-parse_path() {
-  local rel="$1"
-  OWNER="${rel%%/*}"
-  rest="${rel#*/}"
-  REPO="${rest%%/*}"
-}
 
 echo "Fetching manifest from ${PAGES_URL}/.git-pages/manifest.json"
 MANIFEST=$(curl_with_host "${PAGES_URL}/.git-pages/manifest.json")
@@ -97,6 +33,7 @@ fi
 echo ""
 echo "=== Phase 1: collect reports ==="
 declare -A SEEN_REPORTS
+declare -A SEEN_ECHO_COMMITS
 declare -a REPORTS
 while IFS= read -r meta_path; do
   report_dir=$(dirname "$meta_path")
@@ -117,58 +54,105 @@ while IFS= read -r meta_path; do
 
   days=$(age_days "$published")
   REPORTS+=("${report_dir}|${OWNER}|${REPO}|${branch}|${days}")
-  echo "  ${OWNER}/${REPO} branch=${branch} age=${days}d"
+
+  commit_dir=$(dirname "$report_dir")
+  if [ -z "${SEEN_ECHO_COMMITS[$commit_dir]:-}" ]; then
+    SEEN_ECHO_COMMITS[$commit_dir]=1
+    echo "  ${commit_dir} branch=${branch} age=${days}d"
+  fi
 done <<< "$META_PATHS"
 
 [ "${#REPORTS[@]}" -eq 0 ] && { echo "No actionable reports"; exit 0; }
 
 echo ""
-echo "=== Phase 2: check branches in Gitea ==="
+echo "=== Phase 2: check branches/repos in Gitea ==="
+if [ -z "$GITEA_API_URL" ] || [ -z "$GITEA_TOKEN" ]; then
+  echo "ERROR: GITEA_API_URL and GITEA_TOKEN must be set" >&2
+  exit 1
+fi
 declare -a TO_DELETE
 declare -a KEEP
+declare -A SEEN_ECHO_BRANCHES
+declare -A SEEN_ECHO_REPO_DELETED
+declare -A UNIQUE_BRANCHES
+declare -A REASON_MAP
+declare -A COMMIT_BRANCH_MAP
+
+# Build commit→branch mapping
+for entry in "${REPORTS[@]}"; do
+  IFS='|' read -r dir _ _ branch _ <<< "$entry"
+  commit_dir=$(dirname "$dir")
+  [ -n "${COMMIT_BRANCH_MAP[$commit_dir]:-}" ] || COMMIT_BRANCH_MAP["$commit_dir"]=$branch
+done
+for entry in "${REPORTS[@]}"; do
+  IFS='|' read -r _ owner repo branch _ <<< "$entry"
+  UNIQUE_BRANCHES["${owner}/${repo}/${branch}"]=1
+done
+TOTAL_BRANCHES=${#UNIQUE_BRANCHES[@]}
+BRANCHES_EXISTING=0
+BRANCH_DELETED_COUNT=0
+REPO_DELETED_COUNT=0
+MAXAGE_DELETED=0
+KEEPMIN_DELETED=0
 for entry in "${REPORTS[@]}"; do
   IFS='|' read -r dir owner repo branch days <<< "$entry"
 
-  if [ -n "$GITEA_API_URL" ] && [ -n "$GITEA_TOKEN" ]; then
+  branch_key="${owner}/${repo}/${branch}"
-    if branch_exists "$owner" "$repo" "$branch"; then
+  if branch_exists "$owner" "$repo" "$branch"; then
-      echo "  BRANCH EXISTS: ${owner}/${repo}/${branch}"
+    if [ -z "${SEEN_ECHO_BRANCHES[$branch_key]:-}" ]; then
-      KEEP+=("${dir}|${owner}|${repo}|${branch}|${days}")
+      SEEN_ECHO_BRANCHES[$branch_key]=1
-    else
+      BRANCHES_EXISTING=$((BRANCHES_EXISTING + 1))
-      echo "  BRANCH DELETED: ${owner}/${repo}/${branch} -> DELETE"
+      echo "  BRANCH EXISTS: ${branch_key}"
-      TO_DELETE+=("$dir")
     fi
-  else
     KEEP+=("${dir}|${owner}|${repo}|${branch}|${days}")
+  else
+    if [ -z "${SEEN_ECHO_BRANCHES[$branch_key]:-}" ]; then
+      SEEN_ECHO_BRANCHES[$branch_key]=1
+      repo_key="${owner}/${repo}"
+      if [ "${REPO_STATUS[$repo_key]:-}" = "deleted" ]; then
+        REPO_DELETED_COUNT=$((REPO_DELETED_COUNT + 1))
+        if [ -z "${SEEN_ECHO_REPO_DELETED[$repo_key]:-}" ]; then
+          SEEN_ECHO_REPO_DELETED[$repo_key]=1
+          echo "  REPO DELETED: ${repo_key} -> DELETE ALL"
+        fi
+        reason="repo deleted"
+      else
+        BRANCH_DELETED_COUNT=$((BRANCH_DELETED_COUNT + 1))
+        echo "  BRANCH DELETED: ${branch_key} -> DELETE"
+        reason="branch deleted"
+      fi
+    fi
+    REASON_MAP["$dir"]="$reason"
+    TO_DELETE+=("$dir")
   fi
 done
 
 echo ""
 echo "=== Phase 3: apply retention rules to remaining reports ==="
-declare -A BRANCH_COUNTS
+PHASE2_DELETED=${#TO_DELETE[@]}
-if [ "${#KEEP[@]}" -gt 0 ]; then
+apply_retention "$CONFIG"
-  IFS=$'\n'
+PHASE3_DELETED=$(( ${#TO_DELETE[@]} - PHASE2_DELETED ))
-  for entry in $(printf '%s\n' "${KEEP[@]}" | sort -t'|' -k4,4 -k5,5rn); do
-    IFS='|' read -r dir owner repo branch days <<< "$entry"
-    max_age=$(rule_max_age "$branch")
-    keep_min=$(rule_keep_min "$branch")
 
-    if [ "$days" -gt "$max_age" ]; then
+fmt_num() {
-      echo "  DELETE: ${dir} (age ${days}d > maxAge ${max_age}d, branch ${branch})"
+  local n="$1" out=""
-      TO_DELETE+=("$dir")
+  [ -z "$n" ] && { echo "?"; return; }
-      continue
+  n="${n##0}"  # strip leading zeros
-    fi
+  while [ "${#n}" -gt 3 ]; do
-
+    out=" ${n: -3}$out"
-    key="${branch}"
+    n="${n:0:${#n}-3}"
-    count="${BRANCH_COUNTS[$key]:-0}"
-    count=$((count + 1))
-    BRANCH_COUNTS["$key"]=$count
-    if [ "$count" -gt "$keep_min" ]; then
-      echo "  DELETE: ${dir} (kept ${keep_min}/${count}, exceeds keepMin, branch ${branch})"
-      TO_DELETE+=("$dir")
-    fi
   done
-  unset IFS
+  echo "${n}${out}"
-fi
+}
+
+echo ""
+echo "=== Summary ==="
+echo "  Branches:"
+echo "    existing:          $(fmt_num $BRANCHES_EXISTING)"
+echo "    deleted:           $(fmt_num $BRANCH_DELETED_COUNT)"
+echo "    repo gone:         $(fmt_num $REPO_DELETED_COUNT)"
+echo "  Commits:"
+echo "    deleted by maxAge: $(fmt_num $MAXAGE_DELETED)"
+echo "    deleted by keepMin:$(fmt_num $KEEPMIN_DELETED)"
 
 if [ "${#TO_DELETE[@]}" -eq 0 ]; then
   echo "Nothing to delete"
@@ -193,14 +177,37 @@ echo "Downloading archive.tar..."
 HTTP_CODE=$(curl_with_host -o "$ARCHIVE_FILE" -w "%{http_code}" -sS "${PAGES_URL}/.git-pages/archive.tar")
 
 if [ "$HTTP_CODE" = "200" ] && tar -tf "$ARCHIVE_FILE" >/dev/null 2>&1; then
-  echo "Extracting archive..."
+  OLD_KB=$(du -sk "$ARCHIVE_FILE" 2>/dev/null | awk '{print $1}')
+  echo "Extracting archive (${OLD_KB}kB)..."
   tar -xf "$ARCHIVE_FILE" -C "$SITE_DIR"
 
-  for dir in "${TO_DELETE[@]}"; do
+  declare -A GROUP_SEEN
-    if [ -d "$SITE_DIR/$dir" ]; then
+  declare -A GROUP_LINES
-      echo "  Removing: $dir"
+  for del in "${TO_DELETE[@]}"; do
-      rm -rf "$SITE_DIR/$dir"
+    if [ ! -d "$SITE_DIR/$del" ]; then
+      continue
     fi
+
+    commit_dir=$(dirname "$del")
+    branch="${COMMIT_BRANCH_MAP[$commit_dir]:-?}"
+    reason="${REASON_MAP[$del]:-?}"
+    repo_path="${del%%/reports/*}"
+    commit_hash="${commit_dir##*/}"
+    key="${repo_path}/${branch} | Reason: ${reason}"
+    seen_key="${key}|${commit_hash}"
+
+    if [ -z "${GROUP_SEEN[$seen_key]:-}" ]; then
+      GROUP_SEEN[$seen_key]=1
+      GROUP_LINES["$key"]="${GROUP_LINES[$key]:-} $commit_hash"
+    fi
+
+    rm -rf "$SITE_DIR/$del"
+  done
+  for key in "${!GROUP_LINES[@]}"; do
+    echo "  Removing: ${key}"
+    for hash in ${GROUP_LINES[$key]}; do
+      echo "    commit: ${hash}"
+    done
   done
 else
   echo "archive.tar failed (HTTP ${HTTP_CODE}) - falling back to manifest-based rebuild"
@@ -248,6 +255,7 @@ if [ -z "$(ls -A "$SITE_DIR" 2>/dev/null)" ]; then
 fi
 
 tar -cf "$NEW_TAR" -C "$SITE_DIR" .
+NEW_KB=$(du -sk "$NEW_TAR" 2>/dev/null | awk '{print $1}')
 
 echo "PUT: replacing site contents..."
 HTTP_CODE=$(curl_with_host -X PUT "${PAGES_URL}/" \
@@ -259,6 +267,9 @@ HTTP_CODE=$(curl_with_host -X PUT "${PAGES_URL}/" \
 echo "HTTP ${HTTP_CODE}"
 if [ "$HTTP_CODE" = "200" ] || [ "$HTTP_CODE" = "201" ] || [ "$HTTP_CODE" = "204" ]; then
   echo "Site rebuild completed."
+  if [ -n "${OLD_KB:-}" ]; then
+    echo "  archive size: $(fmt_num $OLD_KB)kB → $(fmt_num $NEW_KB)kB"
+  fi
 else
   echo "ERROR: PUT HTTP ${HTTP_CODE}" >&2
   exit 1

git-pages/files/retention-lib.sh

@@ -0,0 +1,184 @@
+#!/usr/bin/env bash
+# Shared functions for retention-cleanup.sh
+# Can be sourced by tests for unit testing
+
+age_days() {
+  local published="$1" epoch_pub now
+  epoch_pub=$(date -d "$published" +%s 2>/dev/null || date -j -f "%Y-%m-%dT%H:%M:%SZ" "$published" +%s 2>/dev/null || echo 0)
+  [ "$epoch_pub" -eq 0 ] && echo 99999 && return
+  now=$(date -u +%s)
+  echo $(( (now - epoch_pub) / 86400 ))
+}
+
+parse_path() {
+  local rel="$1"
+  OWNER="${rel%%/*}"
+  rest="${rel#*/}"
+  REPO="${rest%%/*}"
+}
+
+read_rule() {
+  local config="$1" branch="$2" key="$3" default="$4"
+  v=$(jq -r --arg b "$branch" --arg k "$key" '.branches[$b][$k] // empty' "$config")
+  [ -n "$v" ] && echo "$v" || echo "$default"
+}
+
+# ---------------------------------------------------------------------------
+# Gitea branch/repo checking via git ls-remote
+# Uses global: GITEA_API_URL, GITEA_TOKEN
+# Sets global: REPO_BRANCHES_CACHE, REPO_STATUS
+# ---------------------------------------------------------------------------
+
+# Fetch all branches for a repo (one git ls-remote call per repo).
+# Sets REPO_STATUS[owner/repo].
+# Echos branch list on success.
+# Returns: 0=ok, 1=deleted, 2=cert_error, 3=error (fail-safe keep)
+repo_branches() {
+  local owner="$1" repo="$2" key="${owner}/${repo}"
+  local attempt output
+
+  [ -z "$GITEA_API_URL" ] && return 0
+  [ -z "$GITEA_TOKEN" ] && return 0
+
+  # Check cached status first (avoids re-running git on every report)
+  case "${REPO_STATUS[$key]:-}" in
+    deleted)    return 1 ;;
+    cert_error) return 2 ;;
+    error)      echo "${REPO_BRANCHES_CACHE[$key]:-}"; return 3 ;;
+  esac
+
+  # Cache hit (success with branch list)
+  [ -n "${REPO_BRANCHES_CACHE[$key]:-}" ] && { echo "${REPO_BRANCHES_CACHE[$key]}"; return 0; }
+
+  local git_host
+  git_host=$(echo "$GITEA_API_URL" | sed -E 's|^https?://||' | sed 's|/.*$||')
+  local git_url="https://token:${GITEA_TOKEN}@${git_host}/${owner}/${repo}.git"
+
+  for attempt in 1 2 3; do
+    output=$(git ls-remote --heads "$git_url" 2>&1) && {
+      local branches
+      branches=$(echo "$output" | sed -n 's|.*refs/heads/||p')
+      REPO_BRANCHES_CACHE[$key]="$branches"
+      REPO_STATUS[$key]="ok"
+      echo "$branches"
+      return 0
+    }
+
+    # Repo deleted → no retry
+    if echo "$output" | grep -qiE "fatal:.*(not found|repository.*not|could not read)"; then
+      REPO_BRANCHES_CACHE[$key]="__REPO_DELETED__"
+      REPO_STATUS[$key]="deleted"
+      echo "  REPO DELETED: ${owner}/${repo}" >&2
+      return 1
+    fi
+
+    [ "$attempt" -lt 3 ] && sleep 10
+  done
+
+  # Certificate verification failure → configuration error, stop
+  if echo "$output" | grep -qi "server certificate verification failed"; then
+    REPO_STATUS[$key]="cert_error"
+    echo "[ERROR] git-pages.retention: certificate verification failed for ${owner}/${repo}" >&2
+    echo "[ERROR] git-pages.retention:   check CA certificates or set GIT_SSL_NO_VERIFY=1" >&2
+    echo "[ERROR] git-pages.retention:   git output:" >&2
+    echo "$output" >&2
+    return 2
+  fi
+
+  # Other network errors → fail-safe keep, continue
+  REPO_BRANCHES_CACHE[$key]="__REPO_ERROR__"
+  REPO_STATUS[$key]="error"
+  echo "[WARN] git-pages.retention: cannot reach Gitea for ${owner}/${repo} — keeping all reports" >&2
+  echo "[WARN] git-pages.retention:   git output:" >&2
+  echo "$output" >&2
+  return 3
+}
+
+# Check if a specific branch exists in a repo.
+# Returns 0 (exists), 1 (not found/deleted).
+# Returns 2 (cert error), 3 (network error).
+branch_exists() {
+  local owner="$1" repo="$2" branch="$3"
+  local branches rc
+
+  [ -z "$GITEA_API_URL" ] && return 0
+  [ -z "$GITEA_TOKEN" ] && return 0
+
+  branches=$(repo_branches "$owner" "$repo")
+  rc=$?
+
+  # Return codes from repo_branches propagate through $() subshell:
+  # 0=ok, 1=deleted, 2=cert_error, 3=error
+  case $rc in
+    2) echo "[FATAL] git-pages.retention: cannot reach Gitea (${owner}/${repo}) — check configuration" >&2
+       exit 1 ;;
+    3) return 0 ;;  # network error → fail-safe keep
+    1) return 1 ;;  # repo/branch gone
+  esac
+
+  echo "$branches" | grep -qxF "$branch"
+}
+
+# Phase 3: apply retention rules to KEEP array, populate TO_DELETE
+# Reads from global KEEP array
+# Populates global TO_DELETE array
+# Usage: apply_retention <config_path>
+apply_retention() {
+  local config="$1"
+  local default_max_age default_keep_min
+  local max_age keep_min key count seen_key commit_dir
+  local entry dir owner repo branch days
+
+  default_max_age=$(jq -r '.branches.default.maxAgeDays // 90' "$config")
+  default_keep_min=$(jq -r '.branches.default.keepMin // 5' "$config")
+
+  declare -A BRANCH_COUNTS
+  declare -A SEEN_COMMITS
+  declare -A DELETED_COMMITS
+
+  if [ "${#KEEP[@]}" -eq 0 ]; then
+    return
+  fi
+
+  IFS=$'\n'
+  for entry in $(printf '%s\n' "${KEEP[@]}" | sort -t'|' -k4,4 -k5,5n); do
+    IFS='|' read -r dir owner repo branch days <<< "$entry"
+
+    max_age=$(read_rule "$config" "$branch" "maxAgeDays" "$default_max_age")
+    keep_min=$(read_rule "$config" "$branch" "keepMin" "$default_keep_min")
+
+    # Age check — per-report-type deletion
+    if [ "$days" -gt "$max_age" ]; then
+      echo "  DELETE: ${dir} (age ${days}d > maxAge ${max_age}d, branch ${branch})"
+      TO_DELETE+=("$dir")
+      REASON_MAP["$dir"]="maxAgeDays exceed"
+      MAXAGE_DELETED=$((MAXAGE_DELETED + 1))
+      continue
+    fi
+
+    # keepMin — per-commit counting
+    commit_dir=$(dirname "$dir")
+    key="$branch"
+    seen_key="${key}|${commit_dir}"
+
+    if [ -z "${SEEN_COMMITS[$seen_key]:-}" ]; then
+      SEEN_COMMITS["$seen_key"]=1
+      count="${BRANCH_COUNTS[$key]:-0}"
+      count=$((count + 1))
+      BRANCH_COUNTS["$key"]=$count
+    else
+      count="${BRANCH_COUNTS[$key]:-0}"
+    fi
+
+    if [ "$count" -gt "$keep_min" ]; then
+      if [ -z "${DELETED_COMMITS[$commit_dir]:-}" ]; then
+        DELETED_COMMITS[$commit_dir]=1
+        echo "  DELETE: ${commit_dir} (kept ${keep_min}/${count} commits, exceeds keepMin, branch ${branch})"
+        TO_DELETE+=("$commit_dir")
+        REASON_MAP["$commit_dir"]="keepMin exceed"
+        KEEPMIN_DELETED=$((KEEPMIN_DELETED + 1))
+      fi
+    fi
+  done
+  unset IFS
+}

git-pages/templates/deployment.yaml

@@ -70,8 +70,15 @@ spec:
               set -euo pipefail
               echo "Retention sidecar: installing deps..."
               apt-get update -qq
-              apt-get install -y --no-install-recommends curl jq python3 >/dev/null
+              apt-get install -y --no-install-recommends curl jq git ca-certificates >/dev/null
               echo "Retention sidecar: ready"
+              # Sleep until 01:00 so retention runs at night
+              now_epoch=$(date +%s)
+              target_epoch=$(date -d "today 01:00:00" +%s)
+              [ "$target_epoch" -le "$now_epoch" ] && target_epoch=$((target_epoch + 86400))
+              sleep_sec=$((target_epoch - now_epoch))
+              echo "Retention sidecar: next run in $((sleep_sec / 3600))h (at 01:00)"
+              sleep $sleep_sec
               while true; do
                 /scripts/retention-cleanup.sh
                 echo "Retention sidecar: next run in 24h"

git-pages/templates/retention-configmap.yaml

@@ -8,6 +8,8 @@ metadata:
 data:
   retention.json: |
     {{- .Values.retention.rules | toJson | nindent 4 }}
+  retention-lib.sh: |
+    {{- .Files.Get "files/retention-lib.sh" | nindent 4 }}
   retention-cleanup.sh: |
     {{- .Files.Get "files/retention-cleanup.sh" | nindent 4 }}
   retention-run.sh: |

git-pages/templates/retention-cronjob.yaml

@@ -33,7 +33,7 @@ spec:
                 - |
                   set -euo pipefail
                   apt-get update -qq
-                  apt-get install -y --no-install-recommends curl jq >/dev/null
+                  apt-get install -y --no-install-recommends curl jq git >/dev/null
                   chmod +x /scripts/retention-run.sh /scripts/retention-cleanup.sh
                   /scripts/retention-run.sh
               env:

git-pages/tests/retention.bats

@@ -0,0 +1,623 @@
+#!/usr/bin/env bats
+
+setup() {
+  source "$(dirname "$BATS_TEST_DIRNAME")/files/retention-lib.sh"
+  declare -gA REPO_BRANCHES_CACHE
+  declare -gA REPO_STATUS
+  declare -gA REASON_MAP
+  MAXAGE_DELETED=0
+  KEEPMIN_DELETED=0
+  CONFIG=$(mktemp)
+}
+
+teardown() {
+  rm -f "$CONFIG"
+}
+
+write_config() {
+  cat > "$CONFIG"
+}
+
+# ---------------------------------------------------------------------------
+# read_rule
+# ---------------------------------------------------------------------------
+
+@test "read_rule returns default when branch has no override" {
+  write_config <<'EOF'
+{"branches":{"default":{"maxAgeDays":90,"keepMin":5}}}
+EOF
+  result=$(read_rule "$CONFIG" "nonexistent" "maxAgeDays" 90)
+  [ "$result" = "90" ]
+}
+
+@test "read_rule returns branch-specific value" {
+  write_config <<'EOF'
+{"branches":{"default":{"maxAgeDays":90,"keepMin":5},"main":{"maxAgeDays":365,"keepMin":20}}}
+EOF
+  result=$(read_rule "$CONFIG" "main" "keepMin" 5)
+  [ "$result" = "20" ]
+}
+
+@test "read_rule returns default for undefined key even if branch exists" {
+  write_config <<'EOF'
+{"branches":{"default":{"maxAgeDays":90,"keepMin":5},"main":{"maxAgeDays":365}}}
+EOF
+  result=$(read_rule "$CONFIG" "main" "keepMin" 5)
+  [ "$result" = "5" ]
+}
+
+# ---------------------------------------------------------------------------
+# parse_path
+# ---------------------------------------------------------------------------
+
+@test "parse_path extracts owner and repo" {
+  parse_path "my-owner/my-repo/reports/abc123/go-test-unit"
+  [ "$OWNER" = "my-owner" ]
+  [ "$REPO" = "my-repo" ]
+}
+
+@test "parse_path handles owner with hyphens" {
+  parse_path "niko/agent-platform/reports/abc1234/go-test-bdd"
+  [ "$OWNER" = "niko" ]
+  [ "$REPO" = "agent-platform" ]
+}
+
+# ---------------------------------------------------------------------------
+# apply_retention — keepMin per commit
+# ---------------------------------------------------------------------------
+
+@test "keepMin: 6 commits × 4 types, keepMin=10 → all kept (6 commits < 10)" {
+  # With old per-file counting, 24 files > 10 keepMin would delete 14.
+  # With per-commit counting, 6 commits < 10 keepMin keeps everything.
+  write_config <<'EOF'
+{"branches":{"default":{"maxAgeDays":365,"keepMin":10}}}
+EOF
+
+  KEEP=()
+  local -a commits=(c1 c2 c3 c4 c5 c6)
+  local -a ages=(100 80 60 40 20 5)
+  local -a types=(go-test-bdd go-test-unit helm-lint helm-kubeconform)
+  for i in "${!commits[@]}"; do
+    for t in "${types[@]}"; do
+      KEEP+=("niko/agent-platform/reports/${commits[$i]}/$t|niko|agent-platform|main|${ages[$i]}")
+    done
+  done
+
+  TO_DELETE=()
+  apply_retention "$CONFIG"
+
+  [ "${#TO_DELETE[@]}" -eq 0 ]
+}
+
+@test "keepMin: 8 commits × 1 type, keepMin=5 → deletes 3 oldest" {
+  write_config <<'EOF'
+{"branches":{"default":{"maxAgeDays":365,"keepMin":5}}}
+EOF
+
+  KEEP=()
+  local -a ages=(80 70 60 50 40 30 20 10)
+  for i in "${!ages[@]}"; do
+    KEEP+=("niko/r/reports/c$((i+1))/test|niko|r|main|${ages[$i]}")
+  done
+
+  TO_DELETE=()
+  apply_retention "$CONFIG"
+
+  # 5 newest (c8-c4) kept, 3 oldest (c3,c2,c1) deleted
+  [ "${#TO_DELETE[@]}" -eq 3 ]
+  [[ "${TO_DELETE[0]}" == "niko/r/reports/c3" ]]
+  [[ "${TO_DELETE[1]}" == "niko/r/reports/c2" ]]
+  [[ "${TO_DELETE[2]}" == "niko/r/reports/c1" ]]
+}
+
+@test "keepMin: 12 commits × 1 type, keepMin=5 → keeps 5 newest, deletes 7 oldest" {
+  write_config <<'EOF'
+{"branches":{"default":{"maxAgeDays":90,"keepMin":5}}}
+EOF
+
+  KEEP=()
+  for i in $(seq 1 12); do
+    KEEP+=("niko/r/reports/c${i}/test|niko|r|feature/foo|$(( 13 - i ))")
+  done
+
+  TO_DELETE=()
+  apply_retention "$CONFIG"
+
+  # 7 oldest commits deleted (12 - 5 = 7)
+  [ "${#TO_DELETE[@]}" -eq 7 ]
+  # Oldest 7 should be c1..c7 (highest days = oldest = processed last after sort)
+  # Sort is ascending by days, so processed as c12(1d), c11(2d), ..., c1(12d)
+  # keepMin=5: c12-c8 kept, c7-c1 deleted
+  [[ "${TO_DELETE[0]}" == "niko/r/reports/c7" ]]
+  [[ "${TO_DELETE[6]}" == "niko/r/reports/c1" ]]
+}
+
+@test "keepMin: 2 commits × 3 types, keepMin=5 → all kept (2 < 5)" {
+  write_config <<'EOF'
+{"branches":{"default":{"maxAgeDays":90,"keepMin":5}}}
+EOF
+
+  KEEP=()
+  KEEP+=("niko/r/reports/c1/test-a|niko|r|main|30")
+  KEEP+=("niko/r/reports/c1/test-b|niko|r|main|30")
+  KEEP+=("niko/r/reports/c1/test-c|niko|r|main|30")
+  KEEP+=("niko/r/reports/c2/test-a|niko|r|main|10")
+  KEEP+=("niko/r/reports/c2/test-b|niko|r|main|10")
+  KEEP+=("niko/r/reports/c2/test-c|niko|r|main|10")
+
+  TO_DELETE=()
+  apply_retention "$CONFIG"
+
+  [ "${#TO_DELETE[@]}" -eq 0 ]
+}
+
+@test "keepMin: 6 commits × 2 types, keepMin=3 → keeps 3 newest, deletes 3 oldest" {
+  write_config <<'EOF'
+{"branches":{"default":{"maxAgeDays":365,"keepMin":3}}}
+EOF
+
+  KEEP=()
+  local -a commits=(c1 c2 c3 c4 c5 c6)
+  local -a ages=(60 50 40 30 20 10)
+  local -a types=(jest pytest)
+  for i in "${!commits[@]}"; do
+    for t in "${types[@]}"; do
+      KEEP+=("niko/r/reports/${commits[$i]}/$t|niko|r|feature/x|${ages[$i]}")
+    done
+  done
+
+  TO_DELETE=()
+  apply_retention "$CONFIG"
+
+  # Sort by days ascending: c6(10d), c5(20d), c4(30d), c3(40d), c2(50d), c1(60d)
+  # keepMin=3: c6,c5,c4 kept; c3,c2,c1 deleted
+  [ "${#TO_DELETE[@]}" -eq 3 ]
+  [[ "${TO_DELETE[0]}" == "niko/r/reports/c3" ]]
+  [[ "${TO_DELETE[1]}" == "niko/r/reports/c2" ]]
+  [[ "${TO_DELETE[2]}" == "niko/r/reports/c1" ]]
+}
+
+# ---------------------------------------------------------------------------
+# apply_retention — maxAge
+# ---------------------------------------------------------------------------
+
+@test "maxAge: report exceeding maxAge is deleted" {
+  write_config <<'EOF'
+{"branches":{"default":{"maxAgeDays":90,"keepMin":5}}}
+EOF
+
+  KEEP=(
+    "niko/r/reports/c1/test|niko|r|main|100"
+    "niko/r/reports/c2/test|niko|r|main|50"
+  )
+
+  TO_DELETE=()
+  apply_retention "$CONFIG"
+
+  # c1 (100d) > 90, deleted; c2 (50d) < 90, kept
+  [ "${#TO_DELETE[@]}" -eq 1 ]
+  [[ "${TO_DELETE[0]}" == "niko/r/reports/c1/test" ]]
+}
+
+@test "maxAge deletes report-level dir, not commit-level" {
+  write_config <<'EOF'
+{"branches":{"default":{"maxAgeDays":30,"keepMin":5}}}
+EOF
+
+  KEEP=(
+    "niko/r/reports/c1/test-a|niko|r|main|100"
+    "niko/r/reports/c1/test-b|niko|r|main|20"
+    "niko/r/reports/c2/test-a|niko|r|main|10"
+  )
+
+  TO_DELETE=()
+  apply_retention "$CONFIG"
+
+  # Only test-a for c1 is old; test-b for c1 is young, c2 is young
+  [ "${#TO_DELETE[@]}" -eq 1 ]
+  [[ "${TO_DELETE[0]}" == "niko/r/reports/c1/test-a" ]]
+}
+
+# ---------------------------------------------------------------------------
+# apply_retention — maxAge + keepMin interaction
+# ---------------------------------------------------------------------------
+
+@test "maxAge takes precedence over keepMin — aged report deleted, not counted in keepMin" {
+  write_config <<'EOF'
+{"branches":{"default":{"maxAgeDays":30,"keepMin":2}}}
+EOF
+
+  # 3 commits, 1 type each. c1 is old (100d), c2 and c3 are young.
+  # With keepMin=2: c1 should be deleted by maxAge, c2 and c3 kept.
+  # Without the continue after maxAge check, c1 would consume a keepMin slot.
+  KEEP=(
+    "niko/r/reports/c1/test|niko|r|main|100"
+    "niko/r/reports/c2/test|niko|r|main|10"
+    "niko/r/reports/c3/test|niko|r|main|5"
+  )
+
+  TO_DELETE=()
+  apply_retention "$CONFIG"
+
+  # c1 deleted by maxAge, c2 and c3 within keepMin=2
+  [ "${#TO_DELETE[@]}" -eq 1 ]
+}
+
+# ---------------------------------------------------------------------------
+# apply_retention — sorting (newest first)
+# ---------------------------------------------------------------------------
+
+@test "sort order: newest commits processed first within same branch" {
+  write_config <<'EOF'
+{"branches":{"default":{"maxAgeDays":365,"keepMin":2}}}
+EOF
+
+  KEEP=(
+    "niko/r/reports/c1/test|niko|r|main|100"
+    "niko/r/reports/c2/test|niko|r|main|50"
+    "niko/r/reports/c3/test|niko|r|main|10"
+  )
+
+  TO_DELETE=()
+  apply_retention "$CONFIG"
+
+  # Sort by days ascending: c3(10d) 1st, c2(50d) 2nd, c1(100d) 3rd
+  # keepMin=2: c3 and c2 kept, c1 deleted
+  [ "${#TO_DELETE[@]}" -eq 1 ]
+  [[ "${TO_DELETE[0]}" == "niko/r/reports/c1" ]]
+}
+
+@test "sort order: branches sorted alphabetically" {
+  write_config <<'EOF'
+{"branches":{"default":{"maxAgeDays":365,"keepMin":1}}}
+EOF
+
+  KEEP=(
+    "niko/r/reports/c1/test|niko|r|z-branch|50"
+    "niko/r/reports/c2/test|niko|r|a-branch|60"
+    "niko/r/reports/c3/test|niko|r|m-branch|10"
+  )
+
+  TO_DELETE=()
+  apply_retention "$CONFIG"
+
+  # Alphabetical: a-branch, m-branch, z-branch
+  # Each has 1 commit, keepMin=1 → nothing deleted
+  [ "${#TO_DELETE[@]}" -eq 0 ]
+}
+
+@test "multi-branch: each branch has own keepMin counter" {
+  write_config <<'EOF'
+{"branches":{"default":{"maxAgeDays":90,"keepMin":2}}}
+EOF
+
+  KEEP=(
+    "niko/r/reports/c1/test|niko|r|branch-a|30"
+    "niko/r/reports/c2/test|niko|r|branch-a|20"
+    "niko/r/reports/c3/test|niko|r|branch-a|10"
+    "niko/r/reports/c4/test|niko|r|branch-b|60"
+    "niko/r/reports/c5/test|niko|r|branch-b|50"
+    "niko/r/reports/c6/test|niko|r|branch-b|40"
+    "niko/r/reports/c7/test|niko|r|branch-b|30"
+  )
+
+  TO_DELETE=()
+  apply_retention "$CONFIG"
+ 
+  # branch-a: 3 reports → keep 2 newest (c2,c3), delete 1 oldest (c1)
+  # branch-b: 4 reports → keep 2 newest (c6,c7), delete 2 oldest (c4,c5)
+  # Actually: Sort is by branch, then by days ascending
+  # branch-a processed first: c3(10d) 1st, c2(20d) 2nd (keep), c1(30d) 3rd (delete)
+  # branch-b processed next: c7(30d) 1st, c6(40d) 2nd (keep), c5(50d) 3rd (delete), c4(60d) 4th (delete)
+  [ "${#TO_DELETE[@]}" -eq 3 ]
+  [[ "${TO_DELETE[0]}" == "niko/r/reports/c1" ]]
+  [[ "${TO_DELETE[1]}" == "niko/r/reports/c5" ]]
+  [[ "${TO_DELETE[2]}" == "niko/r/reports/c4" ]]
+}
+
+# ---------------------------------------------------------------------------
+# apply_retention — empty / edge cases
+# ---------------------------------------------------------------------------
+
+@test "empty KEEP array → nothing deleted" {
+  write_config <<'EOF'
+{"branches":{"default":{"maxAgeDays":90,"keepMin":5}}}
+EOF
+
+  KEEP=()
+  TO_DELETE=()
+  apply_retention "$CONFIG"
+  [ "${#TO_DELETE[@]}" -eq 0 ]
+}
+
+@test "TO_DELETE preserves Phase 2 entries after apply_retention (no new deletions)" {
+  write_config <<'EOF'
+{"branches":{"default":{"maxAgeDays":365,"keepMin":10}}}
+EOF
+
+  KEEP=(
+    "niko/r/reports/c1/test|niko|r|main|10"
+    "niko/r/reports/c2/test|niko|r|main|5"
+  )
+  TO_DELETE=(
+    "niko/r/reports/abc/branch-gone"
+    "niko/r/reports/def/repo-gone"
+  )
+
+  apply_retention "$CONFIG"
+
+  [ "${#TO_DELETE[@]}" -eq 2 ]
+  [[ "${TO_DELETE[0]}" == "niko/r/reports/abc/branch-gone" ]]
+  [[ "${TO_DELETE[1]}" == "niko/r/reports/def/repo-gone" ]]
+}
+
+@test "TO_DELETE preserves Phase 2 entries AND adds retention deletions" {
+  write_config <<'EOF'
+{"branches":{"default":{"maxAgeDays":365,"keepMin":3}}}
+EOF
+
+  KEEP=(
+    "niko/r/reports/c1/test|niko|r|main|40"
+    "niko/r/reports/c2/test|niko|r|main|30"
+    "niko/r/reports/c3/test|niko|r|main|20"
+    "niko/r/reports/c4/test|niko|r|main|10"
+  )
+  TO_DELETE=(
+    "niko/r/reports/abc/branch-gone"
+  )
+
+  apply_retention "$CONFIG"
+
+  # 1 pre-existing + 1 commit deleted (c1, oldest of 4, keepMin=3)
+  [ "${#TO_DELETE[@]}" -eq 2 ]
+  [[ "${TO_DELETE[0]}" == "niko/r/reports/abc/branch-gone" ]]
+}
+
+# ---------------------------------------------------------------------------
+# branch_exists — mocking REPO_STATUS / REPO_BRANCHES_CACHE
+# ---------------------------------------------------------------------------
+
+@test "branch_exists: branch in list → return 0" {
+  GITEA_API_URL="https://gitea.example.com"
+  GITEA_TOKEN="test-token"
+
+  REPO_BRANCHES_CACHE["owner/repo"]=$'main\nfeature/x'
+  REPO_STATUS["owner/repo"]="ok"
+
+  run branch_exists "owner" "repo" "main"
+  [ "$status" -eq 0 ]
+}
+
+@test "branch_exists: branch not in list → return 1" {
+  GITEA_API_URL="https://gitea.example.com"
+  GITEA_TOKEN="test-token"
+
+  REPO_BRANCHES_CACHE["owner/repo"]=$'main\nfeature/x'
+  REPO_STATUS["owner/repo"]="ok"
+
+  run branch_exists "owner" "repo" "nonexistent"
+  [ "$status" -eq 1 ]
+}
+
+@test "branch_exists: cert error → exit 1 with [FATAL]" {
+  GITEA_API_URL="https://gitea.example.com"
+  GITEA_TOKEN="test"
+
+  REPO_STATUS["owner/repo"]="cert_error"
+
+  run branch_exists "owner" "repo" "any-branch"
+  [ "$status" -eq 1 ]
+  [[ "$output" == *"[FATAL]"* ]]
+}
+
+@test "branch_exists: repo deleted → return 1" {
+  GITEA_API_URL="https://gitea.example.com"
+  GITEA_TOKEN="test-token"
+
+  REPO_BRANCHES_CACHE["owner/repo"]="__REPO_DELETED__"
+  REPO_STATUS["owner/repo"]="deleted"
+
+  run branch_exists "owner" "repo" "any-branch"
+  [ "$status" -eq 1 ]
+}
+
+@test "branch_exists: network error → return 0 (fail-safe keep)" {
+  GITEA_API_URL="https://gitea.example.com"
+  GITEA_TOKEN="test-token"
+
+  REPO_BRANCHES_CACHE["owner/repo"]="__REPO_ERROR__"
+  REPO_STATUS["owner/repo"]="error"
+
+  run branch_exists "owner" "repo" "any-branch"
+  [ "$status" -eq 0 ]
+}
+
+@test "branch_exists: empty GITEA_API_URL → return 0 (skip)" {
+  GITEA_API_URL=""
+  GITEA_TOKEN="test-token"
+
+  run branch_exists "owner" "repo" "any-branch"
+  [ "$status" -eq 0 ]
+}
+
+@test "branch_exists: empty GITEA_TOKEN → return 0 (skip)" {
+  GITEA_API_URL="https://gitea.example.com"
+  GITEA_TOKEN=""
+
+  run branch_exists "owner" "repo" "any-branch"
+  [ "$status" -eq 0 ]
+}
+
+# ---------------------------------------------------------------------------
+# repo_branches — git ls-remote error detection patterns
+# ---------------------------------------------------------------------------
+
+@test "error detection: 'command not found' does NOT trigger repo deleted" {
+  # This must NOT match — "bash: git: command not found" is NOT a repo deletion
+  local msg="bash: git: command not found"
+  run grep -qiE "fatal:.*(not found|repository.*not|could not read)" <<< "$msg"
+  [ "$status" -eq 1 ]
+}
+
+@test "error detection: 'fatal: repo not found' triggers repo deleted" {
+  # This MUST match — genuine git error for deleted/missing repo
+  local msg="fatal: repository 'https://gitea.app/owner/repo.git' not found"
+  run grep -qiE "fatal:.*(not found|repository.*not|could not read)" <<< "$msg"
+  [ "$status" -eq 0 ]
+}
+
+@test "error detection: 'could not read from remote' triggers repo deleted" {
+  local msg="fatal: could not read from remote repository"
+  run grep -qiE "fatal:.*(not found|repository.*not|could not read)" <<< "$msg"
+  [ "$status" -eq 0 ]
+}
+
+# ---------------------------------------------------------------------------
+# git ls-remote integration (real git, temp repo)
+# ---------------------------------------------------------------------------
+
+@test "git ls-remote parsing: lists branches correctly" {
+  local tmpdir=$(mktemp -d)
+
+  git -C "$tmpdir" init -b main source >/dev/null 2>&1
+  git -C "$tmpdir/source" config user.email "test@test"
+  git -C "$tmpdir/source" config user.name "test"
+  git -C "$tmpdir/source" commit --allow-empty -m "init" >/dev/null 2>&1
+  git -C "$tmpdir/source" branch feature/x >/dev/null 2>&1
+  git clone --bare "$tmpdir/source" "$tmpdir/repo.git" >/dev/null 2>&1
+
+  local url="file://$tmpdir/repo.git"
+  local output
+  output=$(git ls-remote --heads "$url" 2>&1)
+  local branches
+  branches=$(echo "$output" | sed -n 's|.*refs/heads/||p')
+
+  echo "$branches" | grep -qxF "main"
+  [ "$?" -eq 0 ]
+
+  echo "$branches" | grep -qxF "feature/x"
+  [ "$?" -eq 0 ]
+
+  ! echo "$branches" | grep -qxF "nonexistent"
+
+  rm -rf "$tmpdir"
+}
+
+# ---------------------------------------------------------------------------
+# repo_branches — retry + error output (using git mock)
+# ---------------------------------------------------------------------------
+
+@test "repo_branches: success returns branches immediately" {
+  local mockdir=$(mktemp -d)
+  cat > "$mockdir/git" << 'SCRIPT'
+#!/usr/bin/env bash
+echo "abc123	refs/heads/main"
+echo "def456	refs/heads/feature/x"
+SCRIPT
+  chmod +x "$mockdir/git"
+  local save_PATH="$PATH"
+  export PATH="$mockdir:$PATH"
+  GITEA_API_URL="https://gitea.example.com"
+  GITEA_TOKEN="test"
+  REPO_BRANCHES_CACHE=()
+  REPO_STATUS=()
+
+  run repo_branches "owner" "repo"
+
+  [ "$status" -eq 0 ]
+  [[ "$output" == *"main"* ]]
+  [[ "$output" == *"feature/x"* ]]
+
+  export PATH="$save_PATH"
+  rm -rf "$mockdir"
+}
+
+@test "repo_branches: retries 3 times on transient error" {
+  local mockdir=$(mktemp -d)
+  cat > "$mockdir/git" << 'SCRIPT'
+#!/usr/bin/env bash
+echo "call" >> "$MOCKDIR/count"
+echo "fatal: unable to access 'https://...'" >&2
+exit 1
+SCRIPT
+  chmod +x "$mockdir/git"
+  # Inject mockdir path into mock script via env var
+  sed -i '' "s|\$MOCKDIR|$mockdir|g" "$mockdir/git"
+  local save_PATH="$PATH"
+  export PATH="$mockdir:$PATH"
+  GITEA_API_URL="https://gitea.example.com"
+  GITEA_TOKEN="test"
+  REPO_BRANCHES_CACHE=()
+  REPO_STATUS=()
+
+  local start=$SECONDS
+  run repo_branches "owner" "repo"
+
+  [ "$status" -eq 3 ]
+  [[ "$output" == *"[WARN] git-pages.retention"* ]]
+  [[ "$output" == *"keeping all reports"* ]]
+  [[ "$output" == *"git output:"* ]]
+  [[ "$output" == *"unable to access"* ]]
+  [ $(cat "$mockdir/count" | wc -l) -eq 3 ]
+  [ $(( SECONDS - start )) -ge 18 ]
+
+  export PATH="$save_PATH"
+  rm -rf "$mockdir"
+}
+
+@test "repo_branches: certificate error → [ERROR] + return 1" {
+  local mockdir=$(mktemp -d)
+  cat > "$mockdir/git" << 'SCRIPT'
+#!/usr/bin/env bash
+echo "call" >> "$MOCKDIR/count"
+echo "fatal: unable to access 'https://gitea.app/owner/repo.git/': server certificate verification failed. CAfile: none CRLfile: none" >&2
+exit 1
+SCRIPT
+  chmod +x "$mockdir/git"
+  sed -i '' "s|\$MOCKDIR|$mockdir|g" "$mockdir/git"
+  local save_PATH="$PATH"
+  export PATH="$mockdir:$PATH"
+  GITEA_API_URL="https://gitea.example.com"
+  GITEA_TOKEN="test"
+  REPO_BRANCHES_CACHE=()
+  REPO_STATUS=()
+
+  run repo_branches "owner" "repo"
+
+  [ "$status" -eq 2 ]
+  [[ "$output" == *"[ERROR]"* ]]
+  [[ "$output" == *"certificate verification"* ]]
+  [[ "$output" == *"git output:"* ]]
+  [[ "$output" == *"unable to access"* ]]
+
+  export PATH="$save_PATH"
+  rm -rf "$mockdir"
+}
+
+@test "repo_branches: repo not found returns immediately (no retry)" {
+  local mockdir=$(mktemp -d)
+  cat > "$mockdir/git" << 'SCRIPT'
+#!/usr/bin/env bash
+echo "call" >> "$MOCKDIR/count"
+echo "fatal: repository 'https://gitea.app/owner/repo.git' not found" >&2
+exit 1
+SCRIPT
+  chmod +x "$mockdir/git"
+  sed -i '' "s|\$MOCKDIR|$mockdir|g" "$mockdir/git"
+  local save_PATH="$PATH"
+  export PATH="$mockdir:$PATH"
+  GITEA_API_URL="https://gitea.example.com"
+  GITEA_TOKEN="test"
+  REPO_BRANCHES_CACHE=()
+  REPO_STATUS=()
+
+  run repo_branches "owner" "repo"
+
+  [ "$status" -eq 1 ]
+  [[ "$output" == *"REPO DELETED"* ]]
+  [[ "$output" != *"[WARN]"* ]]
+  [ $(cat "$mockdir/count" | wc -l) -eq 1 ]
+
+  export PATH="$save_PATH"
+  rm -rf "$mockdir"
+}

scripts/gitops-dispatch.sh

@@ -22,7 +22,10 @@ INPUTS=$(jq -nc \
   --arg source_repo "$GITOPS_SOURCE_REPO" \
   --arg source_commit "$GITOPS_SOURCE_COMMIT" \
   --arg git_tag_prefix "${GITOPS_TAG_PREFIX:-}" \
-  '{file: $file, yq_tpl: $yq_tpl, version: $version, source_repo: $source_repo, source_commit: $source_commit, git_tag_prefix: $git_tag_prefix}')
+  --arg extra_cmd "${GITOPS_EXTRA_CMD:-}" \
+  --arg author_name "${GIT_USER_NAME:-}" \
+  --arg author_email "${GIT_USER_EMAIL:-}" \
+  '{file: $file, yq_tpl: $yq_tpl, version: $version, source_repo: $source_repo, source_commit: $source_commit, git_tag_prefix: $git_tag_prefix, extra_cmd: $extra_cmd, author_name: $author_name, author_email: $author_email}')
 
 DIR="$(cd "$(dirname "$0")" && pwd)"
 echo "gitops-dispatch: dispatching to $GITOPS_REPO/$GITOPS_WORKFLOW..."

scripts/gitops-update.sh

@@ -84,7 +84,12 @@ _gitops_update() {
 
   cd "${CLONE_DIR}" || _gitops_fail "Failed to enter clone directory"
   yq eval -i "${YQ_EXPR}" "${INPUT_FILE}" || _gitops_fail "Failed to update ${INPUT_FILE}"
-  git add "${INPUT_FILE}" || _gitops_fail "Failed to stage ${INPUT_FILE}"
+  if [ -n "${GITOPS_EXTRA_CMD:-}" ]; then
+    eval "${GITOPS_EXTRA_CMD}" || _gitops_fail "Extra command failed: ${GITOPS_EXTRA_CMD}"
+    git add -A || _gitops_fail "Failed to stage all changes"
+  else
+    git add "${INPUT_FILE}" || _gitops_fail "Failed to stage ${INPUT_FILE}"
+  fi
 
   if git diff --cached --quiet; then
     echo "No changes — ${INPUT_FILE} already at ${VERSION}"
@@ -93,8 +98,8 @@ _gitops_update() {
     exit 0
   fi
 
-  git -c user.name="gitea-ci-bot" \
+  git -c user.name="${GIT_USER_NAME:-gitea-ci-bot}" \
-      -c user.email="ci@keskikuja.site" \
+      -c user.email="${GIT_USER_EMAIL:-ci@keskikuja.site}" \
       commit -m "[skip ci] gitops: update version to ${VERSION}" || _gitops_fail "Failed to commit"
   GITOPS_SHA="$(git rev-parse HEAD)"
   git push || _gitops_fail "Failed to push"

skills/gitops-update/SKILL.md

@@ -56,6 +56,15 @@ on:
       git_tag_prefix:
         required: false
         type: string
+      extra_cmd:
+        required: false
+        type: string
+      author_name:
+        required: false
+        type: string
+      author_email:
+        required: false
+        type: string
 
 env:
   INPUT_FILE: ${{ inputs.file }}
@@ -66,6 +75,9 @@ env:
   GITOPS_REPO: ${{ github.repository }}
   GITEA_API_URL: ${{ gitea.server_url }}
   GIT_TAG_PREFIX: ${{ inputs.git_tag_prefix || '' }}
+  GITOPS_EXTRA_CMD: ${{ inputs.extra_cmd || '' }}
+  GIT_USER_NAME: ${{ inputs.author_name || '' }}
+  GIT_USER_EMAIL: ${{ inputs.author_email || '' }}
 
 jobs:
   update:
@@ -130,15 +142,37 @@ gitops-update:
 
 This single job handles: dispatch → poll → find commit SHA → set commit-status on your commit → produce `GITOPS_SUMMARY` output.
 
+To run extra commands (e.g. `helm dependency update`) after the version bump and before the commit:
+
+```yaml
+gitops-update:
+  needs: [load-config, check-version, helm-build-push]
+  if: success()
+  uses: niko/gitea-ci-library/.gitea/workflows/gitops-dispatch.yml@v1
+  secrets: inherit
+  with:
+    env_json: ${{ needs.load-config.outputs.env_json }}
+    version: ${{ needs.check-version.outputs.version }}
+    GITOPS_FILE: Chart.yaml
+    GITOPS_YQ_TPL: '(.dependencies[] | select(.name == "agent-platform-helm") | .version) = "{{VERSION}}"'
+    GITOPS_REPO: niko/agent-platform-gitops
+    GITOPS_EXTRA_CMD: helm dependency update
+```
+
+When `GITOPS_EXTRA_CMD` is set, the script runs it after `yq` and stages all changes (`git add -A`) instead of only the input file — so any files generated by the extra command (e.g. `Chart.lock`, `charts/`) are included in the commit.
+
+By default the GitOps commit is made as `gitea-ci-bot`. To use the original commit author instead, the dispatch workflow resolves it automatically from the consumer repo — no extra config needed. Just ensure the GitOps repo's `gitops-service.yaml` template has the `author_name` and `author_email` inputs and env mappings.
+
 ### 2.3 Parameters
 
 | Input | Required | Description |
-|---|---|---|
+|---|---|---|---|
 | `env_json` | Yes | Config JSON with `GITEA_API_URL`, optional `GIT_TAG_PREFIX` (for multi-component repos) |
 | `version` | Yes | Version to write (e.g. `0.2.3`) |
 | `GITOPS_FILE` | Yes | Path in GitOps repo (e.g. `dev/Chart.yaml`) |
 | `GITOPS_YQ_TPL` | Yes | yq expression, `{{VERSION}}` is replaced at runtime |
 | `GITOPS_REPO` | Yes | GitOps repo slug (e.g. `niko/agent-platform-gitops`) |
+| `GITOPS_EXTRA_CMD` | No | Shell command to run after yq update, before git commit (e.g. `helm dependency update`) |
 
 ### 2.4 Output
 
@@ -177,7 +211,7 @@ report-summary:
 ## 4. What happens at runtime
 
 1. Consumer's `gitops-dispatch.yml` generates a unique `dispatch_id` and POSTs it to the GitOps repo
-2. GitOps workflow clones its own repo, applies `yq`, commits + pushes
+2. GitOps workflow clones its own repo, applies `yq`, runs `GITOPS_EXTRA_CMD` if set, then commits + pushes
 3. Consumer polls the GitOps repo's runs until the workflow completes
 4. Consumer lists recent commits and finds the matching one by commit message `"gitops: update version to X.Y.Z"`
 5. Consumer sets commit-status `gitops/{repo}[/{prefix}]` on its own commit with a link to the exact GitOps commit