production runtime and mock removal

.env.example

@@ -36,10 +36,11 @@ ZHINIAN_WORKER_LOCK_TIMEOUT_MS=300000
 ZHINIAN_WORKER_RETRY_BASE_MS=10000
 ZHINIAN_WORKER_RETRY_MAX_MS=300000
 ZHINIAN_WORKER_REQUEST_TIMEOUT_MS=120000
+ZHINIAN_GO_EMBEDDED_WORKER=true
 
-# Data layer. Use local only for development/single-instance operation.
-# Production must explicitly select postgres; it never falls back to container-local JSON.
-ZHINIAN_DATA_BACKEND=local
+# Data layer. Production and Docker Compose use PostgreSQL; set local only for
+# an explicitly non-production single-process development run.
+ZHINIAN_DATA_BACKEND=postgres
 DATABASE_URL=
 # Migration runner only: PostgreSQL role used by the Go API DATABASE_URL.
 DATABASE_APP_ROLE=
@@ -66,16 +67,12 @@ VOLCENGINE_REGION=cn-north-1
 VOLCENGINE_SERVICE=cv
 VOLCENGINE_VISUAL_ENDPOINT=https://visual.volcengineapi.com
 JIMENG_IMAGE_GENERATE_46_REQ_KEY=jimeng_seedream46_cvtob
-# auto mocks image jobs when Volcengine credentials are missing.
-JIMENG_VISUAL_MOCK=auto
 
 # EvoLink GPT Image 2 relay for image generation.
 EVOLINK_API_KEY=
 EVOLINK_BASE_URL=https://api.evolink.ai
 EVOLINK_IMAGE_MODEL=gpt-image-2
 EVOLINK_IMAGE_QUALITY=medium
-# auto mocks EvoLink image jobs when EVOLINK_API_KEY is missing.
-EVOLINK_MOCK=auto
 
 # Seedance / Volcengine Ark
 SEEDANCE_API_KEY=
@@ -85,8 +82,6 @@ SEEDANCE_RATIO=9:16
 # Seedance 2.0 duration supports integer seconds from 4 to 15, or -1 for model auto.
 SEEDANCE_DURATION=5
 SEEDANCE_RESOLUTION=720p
-# auto mocks video jobs when Seedance credentials are missing.
-SEEDANCE_MOCK=auto
 
 # Aliyun OSS for uploaded/reference assets that Seedance can read.
 ALI_OSS_ENDPOINT=