Update app.py

app.py

@@ -13,6 +13,12 @@ from langchain.prompts import PromptTemplate, ChatPromptTemplate, HumanMessagePr
 from langchain.llms import OpenAI
 from collections import Counter
 import pandas as pd
+from langchain.document_loaders import TextLoader, YoutubeLoader
+from youtube_transcript_api import YouTubeTranscriptApi
+from langchain.indexes import VectorstoreIndexCreator
+
+
+OPENAI_API_KEY = os.environ['OPENAI_API_KEY']
 
 OPENAI_API_KEY = os.environ['OPENAI_API_KEY']
 GOOGLE_MAPS_API = os.environ['GOOGLE_MAPS_API']
@@ -223,9 +229,84 @@ iface2 = gr.Interface(
 )
 
 
+#### TAB 3 ####
+
+
+
+previous_youtube_url = None
+index = None
+
+def get_video_id(url):
+    video_id = None
+    if 'youtu.be' in url:
+        video_id = url.split('/')[-1]
+    else:
+        video_id = url.split('watch?v=')[-1]
+    return video_id
+
+def get_captions(url):
+    try:
+        video_id = get_video_id(url)
+        transcript_list = YouTubeTranscriptApi.list_transcripts(video_id)
+        transcript = transcript_list.find_transcript(['en'])
+        captions = transcript.fetch()
+
+        formatted_captions = ''
+        for caption in captions:
+            formatted_captions += caption['text'] + ' '
+
+        return formatted_captions
+
+    except Exception as e:
+        print(e)
+        return "Error. Could not fetch captions."
+
+
+
+def answer_question(youtube_url, user_question):
+    # You can implement your logic here to process the video, transcribe it, and answer the user question.
+    # For now, let's return the user question as output.
+    global previous_youtube_url
+    global index
+
+    query = '''
+    You are an expert researcher that can answer any questions from a given text.  Here is the question:
+    {}
+    '''.format(str(user_question))
+
+    if previous_youtube_url == youtube_url:
+        #index = VectorstoreIndexCreator().from_loaders([loader])
+        #query = user_question
+        answer = index.query(llm=OpenAI(model="text-davinci-003"), question = query)
+    else:
+        f= open("temp.txt","w+")
+        f.write(get_captions(youtube_url))
+        f.close() 
+        loader = TextLoader("temp.txt")
+    
+        index = VectorstoreIndexCreator().from_loaders([loader])
+        os.remove("temp.txt")
+
+        #query = user_question
+        answer = index.query(llm=OpenAI(model="text-davinci-003"), question = query)
+
+    return answer
+
+iface3 = gr.Interface(
+    fn=answer_question,
+    inputs=[
+        gr.Textbox(lines=1, placeholder="Enter YouTube URL here..."),
+        gr.Textbox(lines=1, placeholder="Enter your question here...")
+    ],
+    outputs=gr.Textbox(),
+    title="YouTube Smart Q & A",
+    description="Enter a YouTube URL & a question and the app will find the answer from the video captions."
+)
+
+
 
 #tab1 = gr.Tab("AI Market Research", inputs=iface1.inputs, outputs=iface1.outputs)
 #tab2 = gr.Tab("SEO Keyword Extractor", inputs=iface2.inputs, outputs=iface2.outputs)
 
-demo = gr.TabbedInterface([iface2, iface1], ["SEO Keyword Extractor", "AI Market Researcher"])
+demo = gr.TabbedInterface([iface2, iface1, iface3], ["SEO Keyword Extractor", "AI Market Researcher","YouTube Smart Q & A"])
 demo.launch()