added classes to make it easier

_integrations/youtube-summarizer/AudioTranscriber.py

@@ -0,0 +1,37 @@
+import io
+import time
+from pydub import AudioSegment
+import openai
+
+
+class AudioTranscriber:
+    def __init__(self, api_key, model='whisper-1', chunk_duration=20):
+        self.client = openai.OpenAI(api_key=api_key)
+        self.model = model
+        self.chunk_duration_ms = chunk_duration * 60 * 1000
+
+    def transcribe_audio(self, filename):
+        audio = AudioSegment.from_file(filename)
+        transcription = ""
+        for split_number, i in enumerate(range(0, len(audio), self.chunk_duration_ms)):
+            chunk = audio[i: i + self.chunk_duration_ms]
+            file_obj = io.BytesIO(chunk.export(format="mp3").read())
+            file_obj.name = f"part_{split_number}.mp3"
+            transcription += self.transcribe_chunk(file_obj, split_number)
+        return transcription
+
+    def transcribe_chunk(self, file_obj, split_number):
+        print(f"Transcribing part {split_number + 1}!")
+        attempts = 0
+        while attempts < 3:
+            try:
+                transcript = self.client.audio.transcriptions.create(
+                    model=self.model, file=file_obj
+                )
+                return transcript.text
+            except Exception as e:
+                attempts += 1
+                print(f"Attempt {attempts} failed. Exception: {str(e)}")
+                time.sleep(5)
+        print("Failed to transcribe after 3 attempts.")
+        return ""

_integrations/youtube-summarizer/AutoGenInteraction.py

@@ -0,0 +1,52 @@
+import autogen
+
+
+class AutoGenInteraction:
+    def __init__(self, api_key, transcription):
+        self.llm_config = {
+            "config_list": [{
+                "model": "gpt-3.5-turbo",
+                "api_key": api_key
+            }],
+            "temperature": 0,
+            "cache_seed": None
+        }
+        self.transcription = transcription
+
+    def initiate_chat(self):
+        user = autogen.UserProxyAgent(
+            "user",
+            system_message="You are an administrator.",
+            human_input_mode="NEVER",
+            is_termination_msg=lambda msg: "TERMINATE" in msg["content"],
+            code_execution_config=False,
+        )
+
+        agent_summarizer = autogen.AssistantAgent(
+            "agent_summarizer",
+            system_message="You are an assistant agent. When you are done, reply with TERMINATE",
+            llm_config=self.llm_config
+        )
+
+        agent_translator = autogen.AssistantAgent(
+            "ai_agent_translator",
+            system_message="You are an assistant agent. When you are done, reply with TERMINATE",
+            llm_config=self.llm_config
+        )
+
+        user.initiate_chats(
+            [
+                {
+                    "recipient": agent_summarizer,
+                    "message": f"Can you summarize: {self.transcription} this nicely with the title and then bullet points?",
+                    "clear_history": True,
+                    "silent": False,
+                    "summary_method": "last_msg",
+                },
+                {
+                    "recipient": agent_translator,
+                    "message": "Can you translate this into Dutch?",
+                    "summary_method": "last_msg",
+                },
+            ]
+        )

_integrations/youtube-summarizer/YouTubeAudioDownloader.py

@@ -0,0 +1,29 @@
+import yt_dlp
+import os
+
+
+class YouTubeAudioDownloader:
+    def __init__(self, url, output_template='%(title)s [%(id)s].%(ext)s'):
+        self.url = url
+        self.ydl_opts = {
+            'format': 'bestaudio/best',
+            'outtmpl': output_template,
+            'postprocessors': [{
+                'key': 'FFmpegExtractAudio',
+                'preferredcodec': 'm4a',
+            }]
+        }
+
+    def download_audio(self):
+        with yt_dlp.YoutubeDL(self.ydl_opts) as ydl:
+            info_dict = ydl.extract_info(self.url, download=True)
+            filename = ydl.prepare_filename(info_dict)
+        return self.find_file_in_directory(filename)
+
+    @staticmethod
+    def find_file_in_directory(base_filename):
+        base_name, _ = os.path.splitext(base_filename)
+        for file in os.listdir('.'):
+            if file.startswith(base_name):
+                return file
+        raise FileNotFoundError(f"The file {base_filename} does not exist in the current directory.")

_integrations/youtube-summarizer/main.py

@@ -0,0 +1,32 @@
+import os
+from dotenv import load_dotenv
+
+from YouTubeAudioDownloader import YouTubeAudioDownloader
+from AudioTranscriber import AudioTranscriber
+from AutoGenInteraction import AutoGenInteraction
+
+load_dotenv()
+
+
+def main():
+    url = 'https://www.youtube.com/watch?v=WqIzUopTPvU'
+    openai_api_key = os.getenv("OPENAI_API_KEY")
+
+    # gives you the filename from yt video
+    downloader = YouTubeAudioDownloader(url)
+    filename = downloader.download_audio()
+    print(f"Downloaded and saved audio (m4a) as: {filename}")
+
+    # transcribes the audio
+    transcriber = AudioTranscriber(api_key=openai_api_key)
+    transcription = transcriber.transcribe_audio(filename)
+    print("Transcription completed.")
+
+    # ai agents summarize/translate
+    autogen_interaction = AutoGenInteraction(api_key=openai_api_key, transcription=transcription)
+    autogen_interaction.initiate_chat()
+    print("Chat completed.")
+
+
+if __name__ == "__main__":
+    main()