Upload api.py

api.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 from fastapi import FastAPI, UploadFile, File, Form, BackgroundTasks, HTTPException
+from fastapi import Body
 from fastapi.responses import JSONResponse, FileResponse
 from fastapi.middleware.cors import CORSMiddleware
 from pathlib import Path
@@ -36,6 +37,8 @@ TEMP_ROOT = Path("/tmp/temp")
 TEMP_ROOT.mkdir(parents=True, exist_ok=True)
 VIDEOS_ROOT = Path("/tmp/data/videos")
 VIDEOS_ROOT.mkdir(parents=True, exist_ok=True)
+IDENTITIES_ROOT = Path("/tmp/characters")
+IDENTITIES_ROOT.mkdir(parents=True, exist_ok=True)
 
 # Sistema de jobs asíncronos
 class JobStatus(str, Enum):
@@ -389,6 +392,94 @@ async def load_casting(
     n_voices = build_voices_index(Path(voices_dir), client, collection_name="index_voices", drop=drop_collections)
     return {"ok": True, "faces": n_faces, "voices": n_voices}
 
+@app.post("/finalize_casting")
+async def finalize_casting(
+    payload: dict = Body(...),
+):
+    """
+    Consolidate selected face and voice clusters into identities directories and build indices.
+    Expected payload:
+    {
+      "video_name": str,
+      "base_dir": str,  # engine temp base for this video
+      "characters": [
+        {"id": "char1", "name": "Nom", "folder": "/tmp/temp/<video>/char1", "kept_files": ["representative.jpg", ...], "description": "..."}, ...
+      ],
+      "voice_clusters": [
+        {"label": 0, "name": "SPEAKER_00", "clips": ["segment_000.wav", ...]}, ...
+      ]
+    }
+    """
+    import os
+    import shutil
+    from pathlib import Path as _P
+
+    video_name = payload.get("video_name")
+    base_dir = payload.get("base_dir")
+    characters = payload.get("characters", []) or []
+    voice_clusters = payload.get("voice_clusters", []) or []
+
+    if not video_name or not base_dir:
+        raise HTTPException(status_code=400, detail="Missing video_name or base_dir")
+
+    faces_out = IDENTITIES_ROOT / video_name / "faces"
+    voices_out = IDENTITIES_ROOT / video_name / "voices"
+    faces_out.mkdir(parents=True, exist_ok=True)
+    voices_out.mkdir(parents=True, exist_ok=True)
+
+    # Consolidate faces per character name (merge same names)
+    for ch in characters:
+        ch_name = (ch.get("name") or "Unknown").strip() or "Unknown"
+        ch_folder = ch.get("folder")
+        kept = ch.get("kept_files") or []
+        if not ch_folder or not os.path.isdir(ch_folder):
+            continue
+        dst_dir = faces_out / ch_name
+        dst_dir.mkdir(parents=True, exist_ok=True)
+        for fname in kept:
+            src = _P(ch_folder) / fname
+            if src.exists() and src.is_file():
+                try:
+                    shutil.copy2(src, dst_dir / fname)
+                except Exception:
+                    pass
+
+    # Consolidate voices per cluster name
+    clips_dir = _P(base_dir) / "clips"
+    for vc in voice_clusters:
+        v_name = (vc.get("name") or f"SPEAKER_{int(vc.get('label',0)):02d}").strip()
+        dst_dir = voices_out / v_name
+        dst_dir.mkdir(parents=True, exist_ok=True)
+        for wav in (vc.get("clips") or []):
+            src = clips_dir / wav
+            if src.exists() and src.is_file():
+                try:
+                    shutil.copy2(src, dst_dir / wav)
+                except Exception:
+                    pass
+
+    # Build indices using casting_loader helpers
+    db_dir = IDENTITIES_ROOT / video_name / "chroma_db"
+    client = ensure_chroma(db_dir)
+    n_faces = build_faces_index(faces_out, client, collection_name="index_faces", deepface_model='Facenet512', drop=True)
+    n_voices = build_voices_index(voices_out, client, collection_name="index_voices", drop=True)
+
+    # Summary of identities
+    face_identities = sorted([p.name for p in faces_out.iterdir() if p.is_dir()]) if faces_out.exists() else []
+    voice_identities = sorted([p.name for p in voices_out.iterdir() if p.is_dir()]) if voices_out.exists() else []
+
+    return {
+        "ok": True,
+        "video_name": video_name,
+        "faces_dir": str(faces_out),
+        "voices_dir": str(voices_out),
+        "db_dir": str(db_dir),
+        "n_faces_embeddings": n_faces,
+        "n_voices_embeddings": n_voices,
+        "face_identities": face_identities,
+        "voice_identities": voice_identities,
+    }
+
 @app.post("/refine_narration")
 async def refine_narration(
     dialogues_srt: str = Form(...),