From e7ae82e51d6c9396b5fd83f0759099680d59fb3a Mon Sep 17 00:00:00 2001
From: hailin <hailin.zhao@gdzx.xyz>
Date: Fri, 20 Feb 2026 00:10:38 -0800
Subject: [PATCH] feat: add healthcheck to all services in docker-compose

NestJS services use wget to check API endpoints.
voice-service uses curl to check FastAPI /docs endpoint.
web-admin uses wget to check Next.js root.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---
 deploy/docker/docker-compose.yml | 54 ++++++++++++++++++++++++++++++++
 1 file changed, 54 insertions(+)

diff --git a/deploy/docker/docker-compose.yml b/deploy/docker/docker-compose.yml
index d92aaf0..100ada9 100644
--- a/deploy/docker/docker-compose.yml
+++ b/deploy/docker/docker-compose.yml
@@ -85,6 +85,12 @@ services:
       - JWT_SECRET=${JWT_SECRET:-dev-jwt-secret}
       - JWT_REFRESH_SECRET=${JWT_REFRESH_SECRET:-dev-jwt-refresh-secret}
       - AUTH_SERVICE_PORT=3001
+    healthcheck:
+      test: ["CMD-SHELL", "wget -qO- http://localhost:3001/api/v1/auth || exit 1"]
+      interval: 30s
+      timeout: 5s
+      retries: 3
+      start_period: 15s
     depends_on:
       postgres:
         condition: service_healthy
@@ -113,6 +119,12 @@ services:
       - ANTHROPIC_API_KEY=${ANTHROPIC_API_KEY}
       - AGENT_ENGINE_TYPE=claude_code_cli
       - AGENT_SERVICE_PORT=3002
+    healthcheck:
+      test: ["CMD-SHELL", "wget -qO- http://localhost:3002/api/v1/agent || exit 1"]
+      interval: 30s
+      timeout: 5s
+      retries: 3
+      start_period: 15s
     depends_on:
       postgres:
         condition: service_healthy
@@ -139,6 +151,12 @@ services:
       - DB_DATABASE=${POSTGRES_DB:-it0}
       - REDIS_URL=redis://redis:6379
       - OPS_SERVICE_PORT=3003
+    healthcheck:
+      test: ["CMD-SHELL", "wget -qO- http://localhost:3003/api/v1/ops || exit 1"]
+      interval: 30s
+      timeout: 5s
+      retries: 3
+      start_period: 15s
     depends_on:
       postgres:
         condition: service_healthy
@@ -166,6 +184,12 @@ services:
       - REDIS_URL=redis://redis:6379
       - VAULT_MASTER_KEY=${VAULT_MASTER_KEY:-dev-vault-key}
       - INVENTORY_SERVICE_PORT=3004
+    healthcheck:
+      test: ["CMD-SHELL", "wget -qO- http://localhost:3004/api/v1/inventory || exit 1"]
+      interval: 30s
+      timeout: 5s
+      retries: 3
+      start_period: 15s
     depends_on:
       postgres:
         condition: service_healthy
@@ -190,6 +214,12 @@ services:
       - DB_DATABASE=${POSTGRES_DB:-it0}
       - REDIS_URL=redis://redis:6379
       - MONITOR_SERVICE_PORT=3005
+    healthcheck:
+      test: ["CMD-SHELL", "wget -qO- http://localhost:3005/api/v1/monitor || exit 1"]
+      interval: 30s
+      timeout: 5s
+      retries: 3
+      start_period: 15s
     depends_on:
       postgres:
         condition: service_healthy
@@ -217,6 +247,12 @@ services:
       - TWILIO_AUTH_TOKEN=${TWILIO_AUTH_TOKEN}
       - TWILIO_PHONE_NUMBER=${TWILIO_PHONE_NUMBER}
       - COMM_SERVICE_PORT=3006
+    healthcheck:
+      test: ["CMD-SHELL", "wget -qO- http://localhost:3006/api/v1/comm || exit 1"]
+      interval: 30s
+      timeout: 5s
+      retries: 3
+      start_period: 15s
     depends_on:
       postgres:
         condition: service_healthy
@@ -243,6 +279,12 @@ services:
       - DB_DATABASE=${POSTGRES_DB:-it0}
       - REDIS_URL=redis://redis:6379
       - AUDIT_SERVICE_PORT=3007
+    healthcheck:
+      test: ["CMD-SHELL", "wget -qO- http://localhost:3007/api/v1/audit || exit 1"]
+      interval: 30s
+      timeout: 5s
+      retries: 3
+      start_period: 15s
     depends_on:
       postgres:
         condition: service_healthy
@@ -261,6 +303,12 @@ services:
       - WHISPER_MODEL=${WHISPER_MODEL:-large-v3}
       - KOKORO_MODEL=${KOKORO_MODEL:-kokoro-82m}
       - DEVICE=${VOICE_DEVICE:-cpu}
+    healthcheck:
+      test: ["CMD-SHELL", "curl -sf http://localhost:3008/docs || exit 1"]
+      interval: 30s
+      timeout: 5s
+      retries: 3
+      start_period: 30s
     depends_on:
       - agent-service
     networks:
@@ -277,6 +325,12 @@ services:
       - API_BASE_URL=http://api-gateway:8000
       - NEXT_PUBLIC_API_BASE_URL=/api/proxy
       - NEXT_PUBLIC_WS_URL=ws://localhost:18000
+    healthcheck:
+      test: ["CMD-SHELL", "wget -qO- http://localhost:3000/ || exit 1"]
+      interval: 30s
+      timeout: 5s
+      retries: 3
+      start_period: 15s
     depends_on:
       - api-gateway
     networks: