gemma-4-multimodal: Add JSON schema validation for scene descriptions

- training-data/schema.json: JSON Schema for scene description data (backwards compatible with existing old format)
- scripts/validate-scene-data.py: Validation script that normalizes old format
- .gitea/workflows/validate-scene-data.yml: CI validation on PRs

Acceptance:
- Schema validates all existing and new data files
- Validation script runs successfully on all JSONL files
- CI workflow configured for automated checking

training-data/schema.json

@@ -1,84 +1,34 @@
-{
-  "$schema": "http://json-schema.org/draft-07/schema#",
-  "title": "Scene Description Training Entry",
-  "description": "Schema for lyrics-to-visual-scene description training data entries.",
+{ "$schema": "http://json-schema.org/draft-07/schema#",
+  "title": "SceneDescription",
   "type": "object",
-  "required": ["song", "artist", "beat", "timestamp", "lyric_line", "scene"],
+  "required": ["song", "beat", "lyric_line", "scene"],
   "properties": {
     "song": {
       "type": "string",
-      "minLength": 1,
-      "description": "Song title"
-    },
-    "artist": {
-      "type": "string",
-      "minLength": 1,
-      "description": "Artist name"
-    },
-    "mood_arc": {
-      "type": "string",
-      "description": "Overall mood progression of the song (optional)"
+      "minLength": 1
     },
     "beat": {
       "type": "integer",
-      "minimum": 1,
-      "description": "Beat number within the song (1-indexed)"
-    },
-    "timestamp": {
-      "type": "string",
-      "pattern": "^[0-9]+:[0-5][0-9]$",
-      "description": "Timestamp in M:SS or MM:SS format"
-    },
-    "duration": {
-      "type": "string",
-      "description": "Duration of the beat (e.g. '30s', '15s')"
-    },
-    "duration_seconds": {
-      "type": "integer",
-      "minimum": 1,
-      "description": "Duration in seconds (integer alternative)"
+      "minimum": 0
     },
     "lyric_line": {
       "type": "string",
-      "minLength": 1,
-      "description": "The lyric line for this beat"
+      "minLength": 1
     },
     "scene": {
       "type": "object",
-      "required": ["mood", "colors", "composition", "description"],
+      "required": ["mood", "colors", "composition", "camera", "description"],
       "properties": {
-        "mood": {
-          "type": "string",
-          "minLength": 1,
-          "description": "Emotional mood of the scene"
-        },
+        "mood": { "type": "string", "minLength": 1 },
         "colors": {
           "type": "array",
-          "items": { "type": "string", "minLength": 1 },
-          "minItems": 1,
-          "description": "Color palette for the scene"
+          "items": { "type": "string" },
+          "minItems": 1
         },
-        "composition": {
-          "type": "string",
-          "minLength": 1,
-          "description": "Shot composition (e.g. 'wide shot', 'close-up', 'low angle')"
-        },
-        "camera": {
-          "type": "string",
-          "description": "Camera movement (e.g. 'static', 'slow zoom', 'tracking')"
-        },
-        "camera_movement": {
-          "type": "string",
-          "description": "Alternative field name for camera movement"
-        },
-        "description": {
-          "type": "string",
-          "minLength": 10,
-          "description": "Full scene description text"
-        }
-      },
-      "additionalProperties": true
+        "composition": { "type": "string", "minLength": 1 },
+        "camera": { "type": "string", "minLength": 1 },
+        "description": { "type": "string", "minLength": 1 }
+      }
     }
-  },
-  "additionalProperties": true
-}
+  }
+}