app.py

import gradio as gr
from transformers import pipeline
import re

# Initialize pipelines
summarizer = pipeline("text2text-generation", model="google/flan-t5-small", tokenizer="google/flan-t5-small")
sentiment_analyzer = pipeline("sentiment-analysis", model="cardiffnlp/twitter-roberta-base-sentiment-latest")
asr_pipeline = pipeline("automatic-speech-recognition", model="openai/whisper-small")

def convert_to_json(transcript_text):
    lines = transcript_text.strip().split("\n")
    session_data = []
    
    therapist_pattern = re.compile(r"^\s*(Therapist|T):", re.IGNORECASE)
    client_pattern = re.compile(r"^\s*(Client|C):", re.IGNORECASE)
    
    current_speaker = None
    current_text = []
    
    for line in lines:
        line = line.strip()
        if therapist_pattern.match(line):
            if current_speaker and current_text:
                session_data.append({"speaker": current_speaker, "text": " ".join(current_text).strip()})
                current_text = []
            current_speaker = "Therapist"
            text_part = therapist_pattern.sub("", line).strip()
            current_text.append(text_part)
        elif client_pattern.match(line):
            if current_speaker and current_text:
                session_data.append({"speaker": current_speaker, "text": " ".join(current_text).strip()})
                current_text = []
            current_speaker = "Client"
            text_part = client_pattern.sub("", line).strip()
            current_text.append(text_part)
        else:
            if current_speaker is None:
                current_speaker = "Unknown"
            current_text.append(line)

    if current_speaker and current_text:
        session_data.append({"speaker": current_speaker, "text": " ".join(current_text).strip()})

    if not session_data:
        session_data = [{"speaker": "Unknown", "text": transcript_text.strip()}]

    json_data = {"session": session_data}
    return json_data

def analyze_session(transcript, custom_instruction, audio):
    # If an audio file is provided, transcribe it
    if audio is not None:
        # 'audio' will be the file path if type="filepath"
        asr_result = asr_pipeline(audio)
        transcript_text = asr_result['text']
    else:
        transcript_text = transcript

    if not transcript_text.strip():
        return "Please provide a transcript or an audio file."

    json_data = convert_to_json(transcript_text)

    prompt = (
        "You are a helpful assistant that summarizes psychotherapy sessions. "
        "The session is provided in JSON format with speaker turns. "
        "Summarize the key themes, emotional shifts, and patterns from this session. "
    )
    if custom_instruction.strip():
        prompt += f" Additionally, {custom_instruction.strip()}"
    prompt += "\n\nJSON data:\n" + str(json_data)

    summary_output = summarizer(prompt, max_length=200, do_sample=False)
    summary = summary_output[0]['generated_text'].strip()

    sentiment_results = sentiment_analyzer(transcript_text)
    main_sentiment = sentiment_results[0]['label']

    words = transcript_text.lower().split()
    keywords_of_interest = ["anxiety", "depression", "relationship", "stress", "fear", "goals", "progress", "cognitive", "behavior"]
    recurring_concerns = list(set([word for word in words if word in keywords_of_interest]))
    if not recurring_concerns:
        recurring_concerns_str = "No specific recurring concerns identified from the predefined list."
    else:
        recurring_concerns_str = "Recurring concerns include: " + ", ".join(recurring_concerns)

    follow_up_suggestions = []
    if "progress" in summary.lower():
        follow_up_suggestions.append("Explore client's perception of progress in more detail.")
    if "relationship" in summary.lower():
        follow_up_suggestions.append("Discuss client's relationship dynamics further.")
    if not follow_up_suggestions:
        follow_up_suggestions.append("Consider following up on the emotional themes identified in the summary.")
    follow_up_suggestions_str = " ".join(follow_up_suggestions)

    final_output = f"**Summary of Session:**\n{summary}\n\n**Overall Sentiment:** {main_sentiment}\n\n**{recurring_concerns_str}**\n\n**Suggested Follow-Up Topics:** {follow_up_suggestions_str}"
    return final_output

description = """# Psychotherapy Session Summarizer

This tool summarizes psychotherapy session transcripts (text or audio) into key themes, emotional shifts, and patterns.

**How to Use:**
- You may upload an audio file of the session or paste the text transcript.
- Optionally provide a custom focus or instruction (e.g., "Focus on how the client talks about their anxiety.").
- Click 'Summarize' to generate a summary along with identified concerns and suggested follow-ups.

**Important:**
-These is a tool for educational or research purpuses, should not be taking as profesional advice. I dont resposabilize for any misuse. 
"""

with gr.Blocks() as demo:
    gr.Markdown(description)
    with gr.Row():
        transcript_input = gr.Textbox(label="Session Transcript (Text)", lines=10, placeholder="Paste the session transcript here...")
        audio_input = gr.Audio(type="filepath", label="Session Audio (Optional)")
    custom_instruction_input = gr.Textbox(label="Custom Instruction (Optional)", placeholder="e.g., Focus on anxiety and coping strategies.")
    summarize_button = gr.Button("Summarize")
    output_box = gr.Markdown()
    
    summarize_button.click(fn=analyze_session, inputs=[transcript_input, custom_instruction_input, audio_input], outputs=output_box)

if __name__ == "__main__":
    demo.launch()