YACWC

VectorService/vector_service.py

@@ -1,11 +1,25 @@
+import uvicorn.protocols.http.httptools_impl
+
+# orig_data_received = uvicorn.protocols.http.httptools_impl.HttpToolsProtocol.data_received
+# def debug_data_received(self, data):
+#     try:
+#         print("RAW INVALID DATA RECEIVED: %r", data)
+#     except Exception:
+#         pass
+#     return orig_data_received(self, data)
+# uvicorn.protocols.http.httptools_impl.HttpToolsProtocol.data_received = debug_data_received
 
 from typing import Union, Optional, List
+from fastapi.exceptions import RequestValidationError
 from pydantic import BaseModel
 from fastapi import FastAPI, Request, Depends
 from CommonCode.settings import get_logger
+from CommonCode.video_meta import FTPVideo
 import logging
-from fastapi.responses import StreamingResponse
+from fastapi.responses import JSONResponse, StreamingResponse
 import os
+import numpy as np
+from fastapi import FastAPI, Request, status
 import sys
 import json
 import time
@@ -20,15 +34,54 @@ session_manager = SessionManager(
     interface=RedisSessionInterface(redis.from_url("redis://localhost"))
 )
 
-logger = get_logger(__name__,'/var/log/vector_search_logs/main_embed_scores', stdout=True, systemd=False, level = logging.INFO)
-r = redis.Redis(host='localhost', port=6379, db=15)
+logger = get_logger(
+    __name__,
+    "/var/log/vector_search_logs/main_embed_scores",
+    stdout=True,
+    systemd=False,
+    level=logging.INFO,
+)
+r = redis.Redis(host="localhost", port=6379, db=15)
+
+
+@app.exception_handler(RequestValidationError)
+async def validation_exception_handler(request: Request, exc: RequestValidationError):
+    exc_str = f"{exc}".replace("\n", " ").replace("   ", " ")
+    logging.error(f"{request}: {exc_str}")
+    content = {"status_code": 10422, "message": exc_str, "data": None}
+    return JSONResponse(
+        content=content, status_code=status.HTTP_422_UNPROCESSABLE_ENTITY
+    )
+
+
+@app.get("/video_embeddings")
+def get_video_embeddings(video_path: str):
+    llvec = ES.get_embeddings_for_a_file(video_path)
+    if llvec is not None:
+        llvec = dict(llvec)
+        llvec["embeds"] = llvec["embeds"].tolist()
+        llvec["frame_numbers"] = llvec["frame_numbers"].tolist()
+
+    return llvec
+
+
+@app.get("/prompt_embedding")
+def get_prompt_embedding(prompt: str):
+    return ES.get_query_vector(prompt)
+
+
+@app.get("/match_scores")
+def get_clip_scores(video_path: str, prompt: str):
+    results = ES.get_clip_scores(video_path, prompt)
+    return results.T.tolist()
 
 
 class VideosPostRequest(BaseModel):
     query: str = "A cat and a human"
     threshold: float = 0.10
     c_dirs: Optional[List[str]] = None
-    task_id: str = 'compute_log'
+    task_id: str = "compute_log"
+
 
 @app.post("/videos.json")
 async def videos_json(
@@ -58,30 +111,27 @@ async def videos_json(
             "/srv/ftp_tcc/leopards1/2025/09/08",
             "/srv/ftp_tcc/leopards1/2025/09/09",
             "/srv/ftp_tcc/leopards1/2025/09/10",
-            "/srv/ftp_tcc/leopards1/2025/09/11",            
+            "/srv/ftp_tcc/leopards1/2025/09/11",
         ]
-    
 
     # print(','.join([str(x) for x in c_dirs]))
     # message = {'task':'SCHEDULED','when':[str(x) for x in c_dirs], 'time':time.time()}
     # r.rpush(task_id, json.dumps(message))?
 
-
     for x in c_dirs:
-        message = {'task':'QUEUEING', 'when': str(x), 'time': time.time()}
+        message = {"task": "QUEUEING", "when": str(x), "time": time.time()}
         r.rpush(task_id, json.dumps(message))
 
     folder_scores = ES.calculate_embedding_score_in_folders(
-        tuple(c_dirs), threshold=threshold, query=query, redis_key = task_id)
+        tuple(c_dirs), threshold=threshold, query=query, redis_key=task_id
+    )
     # if p_hits != ES.calculate_embedding_score_in_folders.cache_info().hits:
     #     logger.info("FROM CACHE")
     # else:pp
     #     logger.info("COMPUTED FROM SCRATCH")
 
-
-
     folder_scores["breaks"] = ES.add_breaks_between_videos(folder_scores)
-    folder_scores['videos'] = ES.collapse_scores_to_maxmin_avg(folder_scores)
+    folder_scores["videos"] = ES.collapse_scores_to_maxmin_avg(folder_scores)
     return folder_scores