refactor stage 1

core/schema/models/detect_jobs.py

@@ -72,49 +72,58 @@ class DetectJob:
 
 
 @dataclass
-class StageCheckpoint:
+class Timeline:
     """
-    A checkpoint saved after a pipeline stage completes.
+    The frame sequence from a source video.
 
-    Binary data (frame images, crops) goes to S3/MinIO.
-    Everything else (structured state) lives here in Postgres.
+    Independent of stages — exists before any stage runs.
+    Stages annotate the timeline, they don't own it.
+    Frames are stored in MinIO as JPEGs.
     """
 
     id: UUID
-    job_id: UUID
-    stage: str
-    stage_index: int  # position in NODES list (0-7)
+    source_asset_id: Optional[UUID] = None
+    source_video: str = ""
+    profile_name: str = ""
+    fps: float = 2.0
 
-    # S3 reference for binary data only
-    frames_prefix: str = ""  # s3 prefix: checkpoints/{job_id}/frames/
-
-    # Frame metadata (non-image fields)
+    # Frame metadata (images in MinIO, metadata here)
+    frames_prefix: str = ""  # s3: timelines/{id}/frames/
     frames_manifest: Dict[int, str] = field(default_factory=dict)  # seq → s3 key
-    frames_meta: List[Dict[str, Any]] = field(default_factory=list)  # sequence, chunk_id, timestamp, hash
-    filtered_frame_sequences: List[int] = field(default_factory=list)
+    frames_meta: List[Dict[str, Any]] = field(default_factory=list)
 
-    # Stage output — stored as blob in MinIO: checkpoints/{job_id}/stages/{stage}.bson
-    # Each stage's serialize_fn/deserialize_fn owns the format.
-    # Postgres only stores the S3 key, not the data itself.
-    stage_output_key: str = ""  # s3 key to the serialized stage output
+    created_at: Optional[datetime] = None
 
-    # Pipeline state (small, stays in Postgres)
-    stats: Dict[str, Any] = field(default_factory=dict)
-    config_snapshot: Dict[str, Any] = field(default_factory=dict)
+
+@dataclass
+class Checkpoint:
+    """
+    A snapshot of pipeline state on a timeline.
+
+    Stage outputs stored as JSONB — each stage serializes to JSON,
+    the checkpoint stores it without knowing the shape.
+
+    parent_id forms a tree: multiple children from the same parent
+    = different config tries from the same starting point.
+    """
+
+    id: UUID
+    timeline_id: UUID
+    parent_id: Optional[UUID] = None  # null = root checkpoint
+
+    # Stage outputs — JSONB per stage, opaque to the checkpoint layer
+    stage_outputs: Dict[str, Any] = field(default_factory=dict)
+
+    # Config that produced this checkpoint
     config_overrides: Dict[str, Any] = field(default_factory=dict)
 
-    # Input refs (for replay)
-    video_path: str = ""
-    profile_name: str = ""
+    # Pipeline state
+    stats: Dict[str, Any] = field(default_factory=dict)
 
-    # Scenario — a checkpoint bookmarked for the editor workflow.
-    # Created by seeders (manual scripts that populate state from real footage)
-    # or captured from a running pipeline. Loaded via URL:
-    #   /detection/?job=<job_id>#/editor/<stage>
+    # Scenario bookmark
     is_scenario: bool = False
-    scenario_label: str = ""  # human-readable name, e.g. "chelsea_edges_lowcanny"
+    scenario_label: str = ""
 
-    # Timestamps
     created_at: Optional[datetime] = None