From bea144c91e06d785fdf347f040c525834550a743 Mon Sep 17 00:00:00 2001
From: Mikei386 <44135113+Mikei386@users.noreply.github.com>
Date: Fri, 21 Aug 2026 14:03:32 +0200
Subject: [PATCH] Make fresh router startup reliable

---
 compose.yaml | 15 ++++++++++++---
 install.sh   | 15 ++++++++++++++-
 2 files changed, 26 insertions(+), 4 deletions(-)

diff --git a/compose.yaml b/compose.yaml
index 1e5998d..a74cbd0 100644
--- a/compose.yaml
+++ b/compose.yaml
@@ -323,8 +323,15 @@ services:
     cap_drop: [ALL]
     # The entrypoint fixes ownership of fresh named volumes and immediately
     # drops to uid/gid 10002 via gosu before starting the router. Without this
-    # narrowly scoped capability a clean installation loops before startup.
-    cap_add: [CHOWN]
+    # narrowly scoped capabilities a clean installation cannot initialize the
+    # volumes and then switch to its unprivileged runtime identity.
+    cap_add: [CHOWN, SETUID, SETGID]
+    healthcheck:
+      test: [CMD, python, -c, "import urllib.request; urllib.request.urlopen('http://127.0.0.1:8081/health', timeout=2)"]
+      interval: 5s
+      timeout: 3s
+      retries: 24
+      start_period: 5s
     depends_on:
       profile-controller:
         condition: service_healthy
@@ -351,7 +358,9 @@ services:
       - "${AI_BIND_ADDRESS:-127.0.0.1}:8080:8080"
     dns: ["${AI_DNS:-1.1.1.1}"]
     networks: [frontend, tools]
-    depends_on: [router]
+    depends_on:
+      router:
+        condition: service_healthy
     security_opt: ["no-new-privileges:true"]
 
 networks:
diff --git a/install.sh b/install.sh
index 47150c3..a6bb943 100755
--- a/install.sh
+++ b/install.sh
@@ -319,8 +319,21 @@ build_and_start() {
   docker compose --env-file "$SECRETS_DIR/stack.env" up -d --build \
     profile-controller router open-webui
 
+  log "Auf gesunden Router warten"
+  local deadline=$((SECONDS + 180))
+  until [ "$(docker inspect --format '{{if .State.Health}}{{.State.Health.Status}}{{else}}{{.State.Status}}{{end}}' mike-ai-router 2>/dev/null || true)" = "healthy" ]; do
+    if (( SECONDS >= deadline )); then
+      docker logs --tail 80 mike-ai-router >&2 || true
+      die "Router wurde nicht rechtzeitig gesund"
+    fi
+    sleep 2
+  done
+
   log "Fast-Profil aktivieren und Readiness prüfen"
-  docker compose --env-file "$SECRETS_DIR/stack.env" exec -T router python - <<'PY'
+  # Use docker exec directly here. Some Compose/Docker combinations return a
+  # transient HTTP 409 while upgrading the exec stream immediately after a
+  # freshly built service has been recreated.
+  docker exec -i mike-ai-router python - <<'PY'
 import json, os, time, urllib.request
 key = os.environ["ROUTER_API_KEY"]
 request = urllib.request.Request(