app.py

import gradio as gr
from gradio_client import Client
import re


def call_api(transcript):
    client = Client(
        "https://huggingface-projects-llama-2-13b-chat.hf.space/--replicas/dknmg/")
    prompt = """
    INSTRUCTIONS:
    Convert this Transcript into Segment Cards.

    IMPORTANT:

    - use the same language from the transcription to write the segment cards.
    - make sure that the duration is correct.
    - dont use special characters or emojis in the description
    - write as many segment cards as the transcription has
    - only output the segment cards
    - dont answer with additional text
    - Add a <card> tag before and after every segment card
    - be precise and make sure every segment card is complete
    - Your Answer should be like the Output Format

    CONTENT OF CARD:
    {
    Title: (max. 50 Characters)
    Description: (summary of content max. 140 Characters)
    Rating: 1-5 (How Relevant is this content in relation to the transript and how good is this topic for a youtube short video)
    Duration: [00:00.000 - 00:00.000]
    }
    TRANSCRIPT: 
    """
    transcript = str(transcript)
    result = client.predict(
        transcript,
        prompt,
        1024,
        0.6,
        0.9,
        50,
        1.2,
        api_name="/chat"
    )
    return result


def transcription_to_segments(segments):
    segment_cards_array = []
    card_pattern = r'<card>.*?</card>'
    for segment in segments:
        result = call_api(segment)
        card_matches = re.findall(card_pattern, result, re.DOTALL)
        for card_match in card_matches:
            card_object = {
                "Title": re.search(r'Title: (.*?)\n', card_match).group(1),
                "Description": re.search(r'Description: (.*?)\n', card_match).group(1),
                "Rating": re.search(r'Rating: (.*?)\n', card_match).group(1),
                "Duration": re.search(r'Duration: (.*?)\n', card_match).group(1)
            }
            segment_cards_array.append(card_object)
    return segment_cards_array


# Define the Gradio interface for transcription and segmentation
interface = gr.Interface(
    fn=transcription_to_segments,
    inputs="text",
    outputs="json"
)

# Launch the Gradio interface
interface.launch()