Upload app.py

app.py

@@ -0,0 +1,160 @@
+import streamlit as st
+import whisper
+import requests
+import json
+from datetime import datetime
+import os
+from dotenv import load_dotenv
+import numpy as np
+from io import BytesIO
+import tempfile
+
+# Load environment variables - try both .env and system environment
+load_dotenv()
+HUGGINGFACE_TOKEN = os.getenv('HUGGINGFACE_TOKEN')
+
+if not HUGGINGFACE_TOKEN:
+    st.error("Please set the HUGGINGFACE_TOKEN in your Space's secrets.")
+    st.stop()
+
+# Initialize Whisper model
+@st.cache_resource
+def load_whisper_model():
+    return whisper.load_model("base")
+
+def process_audio(audio_file):
+    """Process audio file and generate transcription and clinical notes"""
+    try:
+        with st.spinner("Transcribing audio..."):
+            # Transcribe audio
+            model = load_whisper_model()
+            result = model.transcribe(audio_file)
+            st.session_state.transcription = result["text"]
+        
+        with st.spinner("Generating clinical notes..."):
+            # Generate clinical notes
+            st.session_state.clinical_notes = get_clinical_notes(st.session_state.transcription)
+            
+    except Exception as e:
+        st.error(f"Error processing audio: {str(e)}")
+
+# Mixtral API call function
+def get_clinical_notes(transcription):
+    API_URL = "https://api-inference.huggingface.co/models/mistralai/Mixtral-8x7B-Instruct-v0.1"
+    headers = {"Authorization": f"Bearer {HUGGINGFACE_TOKEN}"}
+    
+    messages = [
+        {"role": "system", "content": "You are a medical assistant helping to generate clinical notes from doctor-patient conversations. Format the notes in a clear, professional structure with the following sections: Chief Complaint, History of Present Illness, Review of Systems, Physical Examination, Assessment, and Plan."},
+        {"role": "user", "content": f"Generate clinical notes from this doctor-patient conversation: {transcription}"}
+    ]
+    
+    payload = {
+        "model": "mistralai/Mixtral-8x7B-Instruct-v0.1",
+        "messages": messages,
+        "max_tokens": 500,
+        "stream": False
+    }
+    
+    try:
+        response = requests.post(API_URL, headers=headers, json=payload)
+        response.raise_for_status()
+        return response.json()['choices'][0]['message']['content']
+    except Exception as e:
+        st.error(f"Error generating clinical notes: {str(e)}")
+        return None
+
+# Main app
+st.set_page_config(
+    page_title="AI Clinical Notes",
+    page_icon="🦀",
+    layout="wide"
+)
+
+st.title("🦀 AI Clinical Notes")
+st.markdown("### Created by Dr. Fernando Ly")
+st.markdown("This application helps medical professionals automatically generate clinical notes from patient conversations.")
+
+# Create columns for better layout
+col1, col2 = st.columns(2)
+
+with col1:
+    st.subheader("Audio Input")
+    
+    # Create tabs for different input methods
+    tab1, tab2 = st.tabs(["🎤 Record", "📁 Upload"])
+    
+    with tab1:
+        st.write("Record your conversation directly:")
+        st.write("⚠️ Please allow microphone access when prompted")
+        
+        if "audio_recorder_key" not in st.session_state:
+            st.session_state["audio_recorder_key"] = 0
+            
+        audio_bytes = st.audio_recorder(
+            text="Click to record",
+            recording_color="#e87676",
+            neutral_color="#6aa36f",
+            key=f"audio_recorder_{st.session_state.audio_recorder_key}"
+        )
+        
+        if st.button("Clear Recording"):
+            st.session_state.audio_recorder_key += 1
+            st.experimental_rerun()
+        
+        if audio_bytes:
+            # Display the recorded audio
+            st.audio(audio_bytes, format="audio/wav")
+            
+            # Save audio temporarily and process
+            with tempfile.NamedTemporaryFile(delete=False, suffix='.wav') as tmp_file:
+                tmp_file.write(audio_bytes)
+                process_audio(tmp_file.name)
+                # Clean up
+                os.unlink(tmp_file.name)
+    
+    with tab2:
+        st.write("Or upload a pre-recorded file:")
+        uploaded_file = st.file_uploader("Upload an audio file (WAV format)", type=['wav'])
+        
+        if uploaded_file:
+            # Display the uploaded audio
+            st.audio(uploaded_file, format="audio/wav")
+            
+            # Save audio temporarily
+            with tempfile.NamedTemporaryFile(delete=False, suffix='.wav') as tmp_file:
+                tmp_file.write(uploaded_file.getvalue())
+                process_audio(tmp_file.name)
+                # Clean up
+                os.unlink(tmp_file.name)
+
+with col2:
+    # Display results
+    if "transcription" in st.session_state and st.session_state.transcription:
+        st.subheader("📝 Transcription")
+        st.write(st.session_state.transcription)
+        
+    if "clinical_notes" in st.session_state and st.session_state.clinical_notes:
+        st.subheader("🏥 Clinical Notes")
+        st.write(st.session_state.clinical_notes)
+
+# Instructions
+st.markdown("---")
+st.markdown("""
+### How to use:
+1. Choose your preferred input method:
+   - **Record**: Click the microphone button to start/stop recording directly in your browser
+   - **Upload**: Upload a pre-recorded WAV file
+2. Wait for the transcription and clinical notes to be generated
+3. Review the generated notes
+
+**Note**: For best results, ensure:
+- Clear audio quality
+- Minimal background noise
+- Proper microphone placement
+- Allow microphone access in your browser when prompted
+""")
+
+# Footer
+st.markdown("---")
+st.markdown("*Note: This is an AI-assisted tool. Please review and verify all generated notes.*")
+st.markdown("*For issues or feedback, please visit the [GitHub repository](https://huggingface.co/spaces/fernandoly/AI-Clinical-Notes/discussions)*")