102 lines
2.9 KiB
Python
102 lines
2.9 KiB
Python
|
|
# To parse video ids
|
|
import re
|
|
|
|
# Youtube Transcript stuff import
|
|
import youtube_transcript_api._errors
|
|
from youtube_transcript_api import YouTubeTranscriptApi
|
|
from youtube_transcript_api.formatters import TextFormatter
|
|
|
|
# OpenAI API stuff import
|
|
from openai import AssistantEventHandler
|
|
from openai import OpenAI
|
|
# For streaming
|
|
from typing_extensions import override
|
|
|
|
import threading
|
|
|
|
# Output buffer and thread lock
|
|
output_buffer = []
|
|
output_lock = threading.Lock()
|
|
|
|
# To get the env var
|
|
from dotenv import load_dotenv
|
|
import os
|
|
|
|
load_dotenv()
|
|
|
|
### OpenAI Config
|
|
|
|
# Setting up OpenAI Client with API Key
|
|
api_key = os.getenv("OPENAI_API_KEY")
|
|
client = OpenAI(
|
|
organization='org-7ANUFsqOVIXLLNju8Rvmxu3h',
|
|
project="proj_NGz8Kux8CSka7DRJucAlDCz6",
|
|
api_key=api_key
|
|
)
|
|
|
|
# screw bardo assistant that is configured to make notes and 5Q&A based on any given YouTube Transcript
|
|
asst_screw_bardo_id = "asst_JGFaX6uOIotqy5mIJnu3Yyp7"
|
|
|
|
# This is copy and pasted straight up from the quickstart guide, just appending to an output buffer instead of directly printing:
|
|
class EventHandler(AssistantEventHandler):
|
|
@override
|
|
def on_text_created(self, text) -> None:
|
|
with output_lock:
|
|
output_buffer.append(f"\nassistant > {text}")
|
|
|
|
@override
|
|
def on_text_delta(self, delta, snapshot):
|
|
with output_lock:
|
|
output_buffer.append(delta.value)
|
|
|
|
def on_tool_call_created(self, tool_call):
|
|
with output_lock:
|
|
output_buffer.append(f"\nassistant > {tool_call.type}\n")
|
|
|
|
def create_and_stream(transcript):
|
|
with client.beta.threads.create_and_run_stream(
|
|
assistant_id=asst_screw_bardo_id,
|
|
thread={
|
|
"messages": [{"role": "user", "content": transcript}]
|
|
},
|
|
event_handler=EventHandler()
|
|
) as stream:
|
|
stream.until_done()
|
|
|
|
def get_video_id(url):
|
|
youtu_be = r'(?<=youtu.be/)([A-Za-z0-9_-]{11})'
|
|
youtube_com = r'(?<=youtube\.com\/watch\?v=)([A-Za-z0-9_-]{11})'
|
|
|
|
id = re.search(youtu_be, url)
|
|
if not id:
|
|
id = re.search(youtube_com, url)
|
|
|
|
if not id:
|
|
print("Couldn't parse video ID from URL")
|
|
return None
|
|
|
|
return id.group(1)
|
|
|
|
# Takes the transcript and formats it in basic text before writing it to auto-transcript.txt
|
|
def get_auto_transcript(video_id):
|
|
trans_api_errors = youtube_transcript_api._errors
|
|
try:
|
|
transcript = YouTubeTranscriptApi.get_transcript(video_id, languages=['en'], proxies=None, cookies=None, preserve_formatting=False)
|
|
except trans_api_errors.TranscriptsDisabled:
|
|
return None
|
|
|
|
formatter = TextFormatter() # Ensure that you create an instance of TextFormatter
|
|
|
|
txt_transcript = formatter.format_transcript(transcript)
|
|
return txt_transcript
|
|
|
|
|
|
# Stores the video id imputted by the user
|
|
"""
|
|
video_id = get_video_id()
|
|
|
|
transcript = get_auto_transcript(video_id)
|
|
|
|
create_and_stream(transcript)
|
|
""" |