fix(kling): 补齐全模态基础视频能力

让 omni_video 模型默认同时注册 video_generate 与 image_to_video，兼容通用视频接口在不显式传 modelType 时的能力推断。新增迁移同步可灵 O1 与 3.0 Omni 的基础目录、平台映射、能力详情和默认快照；集成测试覆盖两个模型的无 modelType 文生视频与图生视频。
2026-07-22 01:41:23 +08:00
parent d0cfd0a385
commit 000ee1bbfd
4 changed files with 262 additions and 0 deletions
@@ -0,0 +1,182 @@
+WITH keling_omni_models(provider_model_name, model_type, capabilities) AS (
+  VALUES
+    (
+      'kling-video-o1',
+      '["video_generate","image_to_video","omni_video"]'::jsonb,
+      '{
+        "video_generate": {
+          "output_resolutions": ["720p", "1080p"],
+          "aspect_ratio_allowed": ["16:9", "1:1", "9:16"],
+          "duration_options": [3, 4, 5, 6, 7, 8, 9, 10],
+          "output_audio": false,
+          "input_audio": false,
+          "prompt_length_limit": {
+            "max": 2500,
+            "count_mode": "non_ascii_weighted",
+            "label": "可灵口径"
+          }
+        },
+        "image_to_video": {
+          "output_resolutions": ["720p", "1080p"],
+          "aspect_ratio_allowed": ["16:9", "1:1", "9:16"],
+          "duration_options": [3, 4, 5, 6, 7, 8, 9, 10],
+          "input_first_frame": true,
+          "input_last_frame": false,
+          "input_first_last_frame": true,
+          "input_reference_generate_single": true,
+          "input_reference_generate_multiple": true,
+          "max_images": 7,
+          "max_images_for_last_frame": 2,
+          "support_video_effect_template": false,
+          "output_audio": false,
+          "input_audio": false,
+          "prompt_length_limit": {
+            "max": 2500,
+            "count_mode": "non_ascii_weighted",
+            "label": "可灵口径"
+          }
+        },
+        "omni_video": {
+          "supported_modes": ["text_to_video", "image_reference", "element_reference", "first_last_frame", "video_reference", "video_edit"],
+          "output_resolutions": ["720p", "1080p"],
+          "aspect_ratio_allowed": ["16:9", "1:1", "9:16"],
+          "duration_options": [3, 4, 5, 6, 7, 8, 9, 10],
+          "output_audio": false,
+          "input_audio": false,
+          "max_videos": 1,
+          "max_audios": 0,
+          "max_images": 7,
+          "max_elements": 7,
+          "max_images_and_elements": 7,
+          "limits_with_video": {"max_images_and_elements": 4},
+          "max_images_for_last_frame": 2,
+          "support_instruction_edit": true,
+          "prompt_length_limit": {
+            "max": 2500,
+            "count_mode": "non_ascii_weighted",
+            "label": "可灵口径"
+          }
+        },
+        "originalTypes": ["video_generate", "image_to_video", "omni_video"]
+      }'::jsonb
+    ),
+    (
+      'kling-v3-omni',
+      '["video_generate","image_to_video","omni_video"]'::jsonb,
+      '{
+        "video_generate": {
+          "output_resolutions": ["720p", "1080p", "2160p"],
+          "aspect_ratio_allowed": ["16:9", "1:1", "9:16"],
+          "duration_options": [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
+          "output_audio": true,
+          "input_audio": false,
+          "prompt_length_limit": {
+            "max": 2500,
+            "count_mode": "non_ascii_weighted",
+            "label": "可灵口径"
+          }
+        },
+        "image_to_video": {
+          "output_resolutions": ["720p", "1080p", "2160p"],
+          "aspect_ratio_allowed": ["16:9", "1:1", "9:16"],
+          "duration_options": [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
+          "input_first_frame": true,
+          "input_last_frame": false,
+          "input_first_last_frame": true,
+          "input_reference_generate_single": true,
+          "input_reference_generate_multiple": true,
+          "max_images": 7,
+          "max_images_for_last_frame": 2,
+          "support_video_effect_template": false,
+          "output_audio": true,
+          "input_audio": false,
+          "prompt_length_limit": {
+            "max": 2500,
+            "count_mode": "non_ascii_weighted",
+            "label": "可灵口径"
+          }
+        },
+        "omni_video": {
+          "supported_modes": ["text_to_video", "image_reference", "element_reference", "first_last_frame", "video_reference", "video_edit", "multi_shot"],
+          "output_resolutions": ["720p", "1080p", "2160p"],
+          "aspect_ratio_allowed": ["16:9", "1:1", "9:16"],
+          "duration_options": [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
+          "output_audio": true,
+          "input_audio": false,
+          "max_videos": 1,
+          "max_audios": 0,
+          "max_images": 7,
+          "max_elements": 7,
+          "max_images_and_elements": 7,
+          "limits_with_video": {
+            "max_images_and_elements": 4,
+            "duration_options": [3, 4, 5, 6, 7, 8, 9, 10]
+          },
+          "max_images_for_last_frame": 2,
+          "support_instruction_edit": true,
+          "prompt_length_limit": {
+            "max": 2500,
+            "count_mode": "non_ascii_weighted",
+            "label": "可灵口径"
+          }
+        },
+        "originalTypes": ["video_generate", "image_to_video", "omni_video"]
+      }'::jsonb
+    )
+)
+UPDATE base_model_catalog model
+SET model_type = defs.model_type,
+    capabilities = defs.capabilities,
+    metadata = COALESCE(model.metadata, '{}'::jsonb)
+      || jsonb_build_object(
+        'originalTypes', defs.model_type,
+        'rawModel', COALESCE(model.metadata->'rawModel', '{}'::jsonb)
+          || jsonb_build_object(
+            'types', defs.model_type,
+            'capabilities', defs.capabilities
+          )
+      ),
+    default_snapshot = CASE
+      WHEN COALESCE(model.default_snapshot, '{}'::jsonb) = '{}'::jsonb THEN model.default_snapshot
+      ELSE jsonb_set(
+        jsonb_set(
+          jsonb_set(model.default_snapshot, '{modelType}', defs.model_type, true),
+          '{capabilities}',
+          defs.capabilities,
+          true
+        ),
+        '{metadata}',
+        COALESCE(model.default_snapshot->'metadata', '{}'::jsonb)
+          || jsonb_build_object(
+            'originalTypes', defs.model_type,
+            'rawModel', COALESCE(model.default_snapshot->'metadata'->'rawModel', '{}'::jsonb)
+              || jsonb_build_object(
+                'types', defs.model_type,
+                'capabilities', defs.capabilities
+              )
+          ),
+        true
+      )
+    END,
+    updated_at = now()
+FROM keling_omni_models defs
+WHERE model.provider_model_name = defs.provider_model_name
+  AND model.model_type @> '["omni_video"]'::jsonb;
+
+WITH keling_omni_models(provider_model_name, model_type, capabilities) AS (
+  SELECT DISTINCT ON (model.provider_model_name)
+    model.provider_model_name,
+    model.model_type,
+    model.capabilities
+  FROM base_model_catalog model
+  WHERE model.provider_model_name IN ('kling-video-o1', 'kling-v3-omni')
+    AND model.model_type @> '["omni_video"]'::jsonb
+  ORDER BY model.provider_model_name, (model.provider_key = 'keling') DESC, model.created_at ASC
+)
+UPDATE platform_models model
+SET model_type = defs.model_type,
+    capabilities = defs.capabilities,
+    updated_at = now()
+FROM keling_omni_models defs
+WHERE COALESCE(NULLIF(model.provider_model_name, ''), model.model_name) = defs.provider_model_name
+  AND model.model_type @> '["omni_video"]'::jsonb;