Update app.py

app.py

@@ -1,20 +1,17 @@
 import gradio as gr
-#import urllib.request
 import requests
-#import bs4
-#import lxml
 import os
-#import subprocess
 from huggingface_hub import InferenceClient,HfApi
 import random
 import json
 import datetime
-#from pypdf import PdfReader
 import uuid
-#from query import tasks
+import yt_dlp
+import cv2
+import whisper
+
 from agent import (
     PREFIX,
-    SAVE_MEMORY,
     COMPRESS_DATA_PROMPT,
     COMPRESS_DATA_PROMPT_SMALL,
     LOG_PROMPT,
@@ -28,12 +25,59 @@ save_data=f'https://huggingface.co/datasets/{reponame}/raw/main/'
 #token_self = os.environ['HF_TOKEN']
 #api=HfApi(token=token_self)
 
-
+sizes = list(whisper._MODELS.keys())
+langs = ["none"] + sorted(list(whisper.tokenizer.LANGUAGES.values()))
+current_size = "base"
+loaded_model = whisper.load_model(current_size)
 
 VERBOSE = True
 MAX_HISTORY = 100
 MAX_DATA = 20000
 
+def dl(inp,img):
+    fps="Error"
+    out = None
+    out_file=[]
+    if img == None and inp !="":
+        try:
+            inp_out=inp.replace("https://","")
+            inp_out=inp_out.replace("/","_").replace(".","_").replace("=","_").replace("?","_")
+            if "twitter" in inp:
+                os.system(f'yt-dlp "{inp}" --extractor-arg "twitter:api=syndication" --trim-filenames 160 -o "{uid}/{inp_out}.mp4" -S res,mp4 --recode mp4')  
+            else:
+                os.system(f'yt-dlp "{inp}" --trim-filenames 160 -o "{uid}/{inp_out}.mp4" -S res,mp4 --recode mp4')  
+                
+            out = f"{uid}/{inp_out}.mp4"
+            capture = cv2.VideoCapture(out)
+            fps = capture.get(cv2.CAP_PROP_FPS)
+            capture.release()
+        except Exception as e:
+            print(e)
+            out = None
+    elif img !=None and inp == "":
+        capture = cv2.VideoCapture(img)
+        fps = capture.get(cv2.CAP_PROP_FPS)
+        capture.release()
+        out = f"{img}"
+    return out
+  
+def csv(segments):
+    output = ""
+    for segment in segments:
+      output += f"{segment['start']},{segment['end']},{segment['text']}\n"
+    return output
+def transcribe(path,lang,size):
+    if size != current_size:
+      loaded_model = whisper.load_model(size)
+      current_size = size
+    results = loaded_model.transcribe(path, language=lang)
+    subs = ".csv"
+    if subs == "None":
+      return results["text"]
+    elif subs == ".csv":
+      return csv(results["segments"])
+
+
 def format_prompt(message, history):
   prompt = "<s>"
   for user_prompt, bot_response in history:
@@ -231,7 +275,16 @@ def clear_fn():
 
 with gr.Blocks() as app:
     gr.HTML("""<center><h1>Mixtral 8x7B TLDR Summarizer + Web</h1><h3>Summarize Data of unlimited length</h3>""")
+    with gr.Row():
+        with gr.Column():
+            with gr.Row():
+                inp_url = gr.Textbox(label="Video URL")
+                url_btn = gr.Button("Load Video")
+            vid = gr.Video()
+        trans_btn=gr.Button("Transcribe")
+        trans = gr.Textbox()
     chatbot = gr.Chatbot(label="Mixtral 8x7B Chatbot",show_copy_button=True)
+    
     with gr.Row():
         with gr.Column(scale=3):
             prompt=gr.Textbox(label = "Instructions (optional)")
@@ -245,21 +298,16 @@ with gr.Blocks() as app:
         stop_button=gr.Button("Stop")
         clear_btn = gr.Button("Clear")
     with gr.Row():
-        with gr.Tab("Text"):
-            data=gr.Textbox(label="Input Data (paste text)", lines=6)
-        with gr.Tab("File"):
-            file=gr.Files(label="Input File(s) (.pdf .txt)")
-        with gr.Tab("Raw HTML"):
-            url = gr.Textbox(label="URL")
-        with gr.Tab("PDF URL"):
-            pdf_url = gr.Textbox(label="PDF URL")       
-        with gr.Tab("PDF Batch"):
-            pdf_batch = gr.Textbox(label="PDF URL Batch (comma separated)")
+        sz = gr.Dropdown(label="Model Size", choices=sizes, value='base')
+        lang = gr.Dropdown(label="Language (Optional)", choices=langs, value="none")
     json_out=gr.JSON()
     e_box=gr.Textbox()
     #text=gr.JSON()
     #inp_query.change(search_models,inp_query,models_dd)
+
+    url_btn.click(dl,[inp_url,vid],vid)
+    trans_btn.click(transcribe,[vid,lang,sz],trans)
     clear_btn.click(clear_fn,None,[prompt,chatbot])
-    go=button.click(summarize,[prompt,chatbot,report_check,sum_mem_check,data,file,url,pdf_url,pdf_batch],[prompt,chatbot,e_box,json_out])
+    go=button.click(summarize,[prompt,chatbot,report_check,sum_mem_check],[prompt,chatbot,e_box,json_out])
     stop_button.click(None,None,None,cancels=[go])
 app.queue(default_concurrency_limit=20).launch(show_api=False)