foreverpyrite/screw-bardo
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Streaming Update (NEEDS OPTIMIZATIONS AND FAILSAFES!!!!!!!!!!!!!!!!!!!!!!)

Browse Source
This commit is contained in:
ForeverPyrite
2024-10-05 16:57:03 -04:00
parent 5601e8a874
commit 3ea5681f86
5 changed files with 81 additions and 77 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
.gitignore vendored
View File

@@ -1,4 +1,3 @@
thread-killer.py
log.txt
.env

40
app.py
View File

@@ -1,23 +1,12 @@
from flask import Flask, render_template, Response, request
from main import get_auto_transcript, get_video_id, create_and_stream, EventHandler
from main import get_auto_transcript, get_video_id, create_and_stream, output_buffer, output_lock
from datetime import datetime
import sys
import io
import threading
import pytz
import time
app = Flask(__name__, static_folder="website/static", template_folder="website")
class StreamToLogger(io.StringIO):
def __init__(self):
super().__init__()
def write(self, message):
# I could probably log stuff here
print(message, end='') # Print to standard output (console)
# You could also log this message or handle it differently.
@app.route('/')
def home():
return render_template('index.html')
@@ -33,9 +22,6 @@ def streaming():
@app.route('/process_url', methods=['POST'])
def process_url():
old_stdout = sys.stdout
new_stdout = StreamToLogger()
sys.stdout = new_stdout
# Opens a file to log the video id and the assistants respone to see if I can further improve instructions:
#log = open("log.txt", "at", 1)
url = request.form['url']
@@ -50,11 +36,31 @@ def process_url():
if (not transcript):
return "Successfully parsed video ID from URL, however the ID was either invalid, the transcript was disabled by the video owner, or some other error was raised because of YouTube."
# Start processing in a separate thread
threading.Thread(target=create_and_stream, args=(transcript,)).start()
# Process the transcript and stream the result.
# response = create_and_stream(transcript)
# log.write(f"\n\n\n### New Entry at {datetime.now(pytz.timezone('America/New_York')).strftime('%Y-%m-%d %H:%M:%S')}\n\n URL: {url}\n Video ID: {video_id}\n\nAssistant Response: \n{response}")
# Return a response
return Response(create_and_stream(transcript), content_type="text/plain", status=200, direct_passthrough=True) # Add more detailed output if needed
return Response("Processing started. Check /stream_output for updates.", status=200) # Add more detailed output if needed
@app.route('/stream_output')
def stream_output():
def yoink():
while True:
with output_lock:
if output_buffer:
message = output_buffer.pop(0)
yield f"{message}"
time.sleep(0.005) # Adjust as necessary for your application
return Response(yoink(), content_type='text/plain')
if __name__ == '__main__': # Change this line to properly check for main
app.run(debug=True)

77
main.py
View File

@@ -13,6 +13,12 @@ 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
@@ -21,36 +27,6 @@ load_dotenv()
### OpenAI Config
# This is copy and pasted straight up from the quickstart guide:
class EventHandler(AssistantEventHandler):
@override
def on_text_created(self, text) -> None:
print(f"\nassistant > ", end="", flush=True)
@override
def on_text_delta(self, delta, snapshot):
print(delta.value, end="", flush=True)
def on_tool_call_created(self, tool_call):
print(f"\nassistant > {tool_call.type}\n", flush=True)
def on_tool_call_delta(self, delta, snapshot):
if delta.type == 'code_interpreter':
if delta.code_interpreter.input:
print(delta.code_interpreter.input, end="", flush=True)
if delta.code_interpreter.outputs:
print(f"\n\noutput >", flush=True)
for output in delta.code_interpreter.outputs:
if output.type == "logs":
print(f"\n{output.logs}", flush=True)
# Setting up OpenAI Client with API Key
api_key = os.getenv("OPENAI_API_KEY")
client = OpenAI(
@@ -62,23 +38,32 @@ client = OpenAI(
# screw bardo assistant that is configured to make notes and 5Q&A based on any given YouTube Transcript
asst_screw_bardo_id = "asst_JGFaX6uOIotqy5mIJnu3Yyp7"
# uhh no we need a new thread each time tf
# make sure to call the function after the transcript is confirmed to work, it would be very stupid to call the function and make a new thread this early
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()
messages = stream.get_final_messages()
return messages[0].content[0].text.value
# 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})'

2
website/index.html
View File

@@ -17,7 +17,7 @@
<body>
<div class="content">
<p id="response-area">Response will appear here.</p>
<pre id="response-area">Response will appear here. </pre>
<div class="form_box">
<input id="url_box" placeholder="Paste the lecture URL here." autofocus></input>
<input id="submit" type="submit" onclick=""></input>

38
website/static/script.js
View File

@@ -1,13 +1,14 @@
document.addEventListener("DOMContentLoaded", (event) => {
const response_area = document.getElementById('response-area');
document.getElementById('submit').addEventListener('click', function() {
var url = document.getElementById('url_box').value;
const response_area = document.getElementById('response-area');
if (!url) {
response_area.innerText = 'Please enter a URL.';
return;
}
// First, process the URL
fetch('/process_url', {
method: 'POST',
headers: {
@@ -15,6 +16,26 @@ document.addEventListener("DOMContentLoaded", (event) => {
},
body: new URLSearchParams({ url: url })
})
.then(response => {
if (!response.ok) {
throw new Error('Network response was not ok');
}
// Start streaming once processing is started
streamOutput(response_area);
})
.catch(error => {
console.error('Error processing URL:', error);
response_area.innerText = 'Error processing URL: ' + error.message;
});
});
});
function streamOutput(response_area) {
// Fetch the streaming output
const streamResponsePromise = fetch('/stream_output');
response_area.innerHTML = ""
streamResponsePromise
.then(response => {
const reader = response.body.getReader();
const decoder = new TextDecoder("utf-8");
@@ -28,14 +49,7 @@ document.addEventListener("DOMContentLoaded", (event) => {
// Decode and process the chunk
const chunk = decoder.decode(value, { stream: true });
// Split the received chunk by new line to handle multiple lines (if any)
chunk.split('\n').forEach(data => {
if (data.trim()) { // Avoid empty strings
// Update the inner HTML of the output div
response_area.innerHTML += `<p>${data}</p>`;
}
});
response_area.innerHTML += chunk;
// Continue reading
readStream();
@@ -47,6 +61,6 @@ document.addEventListener("DOMContentLoaded", (event) => {
})
.catch(error => {
console.error('Error fetching stream:', error);
response_area.innerText = 'Error fetching stream: ' + error.message;
});
});
});
}