initial update, video snapshotting reduced

utils.py

@@ -10,6 +10,7 @@ from uuid import uuid4
 from langchain_core.documents  import Document
 from text_extractor import TextExtractor
 import os
+import math
 import json
 from groq import Groq
 import re
@@ -213,10 +214,10 @@ def split_audio_by_duration(audio_file_path, chunk_duration_minutes, print_outpu
     chunk_paths = []
 
     if audio_duration_ms > chunk_length_ms:
-        # Calculate the number of chunks
-        num_chunks = audio_duration_ms // chunk_length_ms + (1 if audio_duration_ms % chunk_length_ms != 0 else 0)
+        # Calculate the number of chunks, using math.ceil to ensure rounding up
+        num_chunks = math.ceil(audio_duration_ms / chunk_length_ms)
 
-        for i in range(num_chunks):
+        for i in range(int(num_chunks)):
             start_ms = i * chunk_length_ms
             end_ms = min(start_ms + chunk_length_ms, audio_duration_ms)
             chunk = audio[start_ms:end_ms]
@@ -299,7 +300,7 @@ def preprocess_video_data(video_path: str, time_interval: int):
     video_name = os.path.splitext(os.path.basename(video_path))[0]
 
     # Create a directory for snapshots using the video name
-    snapshot_dir = os.path.join(os.path.dirname(video_path), f"{video_name}_snapshots")
+    snapshot_dir = os.path.join(os.path.dirname(video_path), f"{video_name}")
     os.makedirs(snapshot_dir, exist_ok=True)
 
     # Get the duration of the video using ffmpeg
@@ -308,15 +309,19 @@ def preprocess_video_data(video_path: str, time_interval: int):
 
     # Loop through the video and take snapshots at 0s, 3min, 6min, etc.
     for i in range(0, int(duration), time_interval):
-        # Calculate the time for the current frame
-        frame_time = i
+        start_time = i
+        end_time = min(i + time_interval, int(duration))
+        
+        # Format the interval as 'start-end'
+        interval_str = f"{start_time}-{end_time}"
+        
         # Save the snapshot as an image file in the created folder
-        frame_img = os.path.join(snapshot_dir, f"frame_at_{frame_time//60}min.png")
+        frame_img = os.path.join(snapshot_dir, f"{interval_str}s.png")
         
         # Extract the frame using ffmpeg
         (
             ffmpeg
-            .input(video_path, ss=frame_time)
+            .input(video_path, ss=start_time)
             .output(frame_img, vframes=1)
             .run()
         )
@@ -326,6 +331,9 @@ def preprocess_video_data(video_path: str, time_interval: int):
     
     # now creating document from the audio file
     documents = create_audio_document(audio_path, chunk_duration_minutes=0.5, file_type='video')
+    
+    # deleting the audio file
+    os.remove(audio_path)
     return documents
 
 
@@ -333,11 +341,11 @@ def preprocess_video_data(video_path: str, time_interval: int):
 #-----------------------------------------------------OTHERS--------------------------------------------------------------
 
 def save_embedded_data(embeddings, key="data"):
-  embeddings.save_local(f"vec-db/index/faiss_index_{key}")
+  embeddings.save_local(f"index/faiss_index_{key}")
   print("Embeddings saved")
 
 def load_embedded_data(embeddings=embeddings, key="data"):
-  embed_db = FAISS.load_local(f"vec-db/index/faiss_index_{key}", embeddings, allow_dangerous_deserialization=True)
+  embed_db = FAISS.load_local(f"index/faiss_index_{key}", embeddings, allow_dangerous_deserialization=True)
   return embed_db
 
 
@@ -361,7 +369,7 @@ def load_documents_from_directory(directory_path: str):
         # updating the path
         path = os.path.join(directory_path, file)
         # getting the file extension and doc name
-        doc_name, extension = file.split('.')[0] , file.split('.')[-1]
+        doc_name, extension = path.split('/')[-1].split('.')[0] , file.split('.')[-1]
         # checking if the file is a text document
         if extension in text_doc:
             # loading the document