Spaces:
Runtime error
Runtime error
joseluhf11
commited on
Commit
•
1fc91f5
1
Parent(s):
04a1ef7
Upload app.py
Browse files
app.py
ADDED
@@ -0,0 +1,72 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Importamos la librería
|
2 |
+
from youtube_transcript_api import YouTubeTranscriptApi
|
3 |
+
import re
|
4 |
+
from langchain.vectorstores import FAISS
|
5 |
+
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
6 |
+
from langchain.embeddings import AwaEmbeddings
|
7 |
+
import os
|
8 |
+
import openai
|
9 |
+
import gradio as gr
|
10 |
+
|
11 |
+
|
12 |
+
def get_transcript(url):
|
13 |
+
video_id = re.search(r"(?<=v=)([^&#]+)", url)
|
14 |
+
video_id = video_id.group(0)
|
15 |
+
|
16 |
+
# retrieve the available transcripts
|
17 |
+
transcript_list = YouTubeTranscriptApi.list_transcripts(video_id)
|
18 |
+
|
19 |
+
# iterate over all available transcripts
|
20 |
+
for transcript in transcript_list:
|
21 |
+
subtitles = transcript.translate('en').fetch()
|
22 |
+
|
23 |
+
# Imprimimos los transcript
|
24 |
+
text = ''
|
25 |
+
for sub in subtitles:
|
26 |
+
text = text + ' ' + sub['text']
|
27 |
+
return text
|
28 |
+
|
29 |
+
|
30 |
+
embeddings = AwaEmbeddings()
|
31 |
+
|
32 |
+
text_splitter = RecursiveCharacterTextSplitter(
|
33 |
+
# Set a really small chunk size, just to show.
|
34 |
+
chunk_size = 1500,
|
35 |
+
chunk_overlap = 100,
|
36 |
+
length_function = len,
|
37 |
+
is_separator_regex = False,
|
38 |
+
)
|
39 |
+
|
40 |
+
def chat(url, question, api_key):
|
41 |
+
os.environ["OPENAI_API_KEY"] = api_key
|
42 |
+
openai.api_key = api_key
|
43 |
+
|
44 |
+
info = get_transcript(url)
|
45 |
+
texts = text_splitter.create_documents([info])
|
46 |
+
db = FAISS.from_documents(texts, embeddings)
|
47 |
+
docs = db.similarity_search(question)
|
48 |
+
|
49 |
+
prompt = [
|
50 |
+
{"role": "system", "content": """You are my Youtube Asisstant. I will pass you texts from a Youtube Video Transcrip and I need you to use them to answer my question from the Youtube Video.
|
51 |
+
Please do not invent any information, and I am asking about information in the Youtube Video."""},
|
52 |
+
{"role":"user", "content": f"Context:{docs}"},
|
53 |
+
{"role":"user", "content": f"Question:{question}"},
|
54 |
+
]
|
55 |
+
response = openai.ChatCompletion.create(
|
56 |
+
model="gpt-3.5-turbo-0613",
|
57 |
+
messages=prompt,
|
58 |
+
temperature = 0
|
59 |
+
)
|
60 |
+
|
61 |
+
return response["choices"][0]["message"]["content"]
|
62 |
+
|
63 |
+
|
64 |
+
# Create a gradio interface with two inputs and one output
|
65 |
+
demo = gr.Interface(
|
66 |
+
fn=chat, # The function to call
|
67 |
+
inputs=[gr.Textbox(label="Youtube URL"), gr.Textbox(label="Question"), gr.Textbox(label="API_KEY")], # The input components
|
68 |
+
outputs=gr.Textbox(label="Answer") # The output component
|
69 |
+
)
|
70 |
+
|
71 |
+
# Launch the interface
|
72 |
+
demo.launch(share=True)
|