🎤 CLARISSA Voice Input Showcase¶

Talk to Your Reservoir Simulation

This notebook demonstrates CLARISSA's voice interface for controlling reservoir simulations through natural language commands.

---

🔄 Workflow¶

Step	Where	What
1️⃣	Web Demo	Record your voice command
2️⃣	This Notebook	Upload + Transcribe + Process

---

🚀 Step 1: Record Your Voice¶

👉 Open Voice Demo (opens in new tab)

1. Click "Start Recording"
2. Speak your command (e.g., "show permeability distribution")
3. Click "Stop"
4. Click "Download WAV"
5. Come back here and upload the file

---

In [ ]:

# Setup (run once)
!pip install -q openai

import os
from IPython.display import display, HTML, Audio
from google.colab import files

print("✅ Ready! Upload your audio file below.")

# Setup (run once) !pip install -q openai import os from IPython.display import display, HTML, Audio from google.colab import files print("✅ Ready! Upload your audio file below.")

---

📤 Step 2: Upload Your Recording¶

In [ ]:

# Upload your recorded audio file
print("📂 Select your audio file (WAV, WebM, MP3, etc.)")
print()

uploaded = files.upload()

if uploaded:
    filename = list(uploaded.keys())[0]
    audio_bytes = uploaded[filename]
    
    print()
    print(f"✅ Uploaded: {filename} ({len(audio_bytes):,} bytes)")
    print()
    print("🔊 Playback:")
    display(Audio(audio_bytes, autoplay=False))
else:
    print("❌ No file uploaded")
    audio_bytes = None

# Upload your recorded audio file print("📂 Select your audio file (WAV, WebM, MP3, etc.)") print() uploaded = files.upload() if uploaded: filename = list(uploaded.keys())[0] audio_bytes = uploaded[filename] print() print(f"✅ Uploaded: {filename} ({len(audio_bytes):,} bytes)") print() print("🔊 Playback:") display(Audio(audio_bytes, autoplay=False)) else: print("❌ No file uploaded") audio_bytes = None

---

📝 Step 3: Transcribe with Whisper¶

Enter your OpenAI API key to transcribe the audio:

In [ ]:

# Transcribe the uploaded audio
import tempfile
from getpass import getpass

# Get API key
if not os.environ.get('OPENAI_API_KEY'):
    os.environ['OPENAI_API_KEY'] = getpass("OpenAI API Key: ")

if audio_bytes:
    from openai import OpenAI
    client = OpenAI()
    
    # Save to temp file
    ext = filename.split('.')[-1] if '.' in filename else 'wav'
    with tempfile.NamedTemporaryFile(suffix=f'.{ext}', delete=False) as f:
        f.write(audio_bytes)
        temp_path = f.name
    
    print("🔄 Transcribing...")
    
    with open(temp_path, 'rb') as audio_file:
        transcript = client.audio.transcriptions.create(
            model="whisper-1",
            file=audio_file,
            language="en",
            prompt="Reservoir simulation commands: permeability, porosity, pressure, saturation, water cut, oil rate, injection, production, layers, wells, grid"
        )
    
    os.unlink(temp_path)
    
    print()
    print("═" * 50)
    print("📝 TRANSCRIPT")
    print("═" * 50)
    print()
    print(f'   "{transcript.text}"')
    print()
    
    # Store for next cell
    command_text = transcript.text
else:
    print("❌ No audio to transcribe. Run the upload cell first.")
    command_text = None

# Transcribe the uploaded audio import tempfile from getpass import getpass # Get API key if not os.environ.get('OPENAI_API_KEY'): os.environ['OPENAI_API_KEY'] = getpass("OpenAI API Key: ") if audio_bytes: from openai import OpenAI client = OpenAI() # Save to temp file ext = filename.split('.')[-1] if '.' in filename else 'wav' with tempfile.NamedTemporaryFile(suffix=f'.{ext}', delete=False) as f: f.write(audio_bytes) temp_path = f.name print("🔄 Transcribing...") with open(temp_path, 'rb') as audio_file: transcript = client.audio.transcriptions.create( model="whisper-1", file=audio_file, language="en", prompt="Reservoir simulation commands: permeability, porosity, pressure, saturation, water cut, oil rate, injection, production, layers, wells, grid" ) os.unlink(temp_path) print() print("═" * 50) print("📝 TRANSCRIPT") print("═" * 50) print() print(f' "{transcript.text}"') print() # Store for next cell command_text = transcript.text else: print("❌ No audio to transcribe. Run the upload cell first.") command_text = None

---

🧠 Step 4: Parse Intent¶

Analyze the voice command to extract the user's intent:

In [ ]:

# Parse the intent from the transcript
if command_text:
    # Simple keyword-based intent detection
    text_lower = command_text.lower()
    
    # Define intents
    intents = []
    
    # Property queries
    if any(w in text_lower for w in ['permeability', 'perm']):
        intents.append(('SHOW_PROPERTY', 'permeability'))
    if any(w in text_lower for w in ['porosity', 'poro']):
        intents.append(('SHOW_PROPERTY', 'porosity'))
    if 'pressure' in text_lower:
        intents.append(('SHOW_PROPERTY', 'pressure'))
    if any(w in text_lower for w in ['saturation', 'sat']):
        intents.append(('SHOW_PROPERTY', 'saturation'))
    if any(w in text_lower for w in ['water cut', 'watercut']):
        intents.append(('SHOW_METRIC', 'water_cut'))
    if any(w in text_lower for w in ['oil rate', 'production rate']):
        intents.append(('SHOW_METRIC', 'oil_rate'))
    
    # Actions
    if any(w in text_lower for w in ['run', 'simulate', 'execute']):
        intents.append(('RUN_SIMULATION', None))
    if any(w in text_lower for w in ['stop', 'cancel', 'abort']):
        intents.append(('STOP_SIMULATION', None))
    
    # Questions
    if any(w in text_lower for w in ['what is', 'what are', 'how much', 'how many']):
        intents.append(('QUERY', 'general'))
    
    print("═" * 50)
    print("🧠 INTENT ANALYSIS")
    print("═" * 50)
    print()
    print(f"Input: \"{command_text}\"")
    print()
    
    if intents:
        print("Detected intents:")
        for intent_type, intent_value in intents:
            if intent_value:
                print(f"  • {intent_type}: {intent_value}")
            else:
                print(f"  • {intent_type}")
    else:
        print("  ⚠️ No specific intent detected")
        print("  → Would pass to LLM for interpretation")
    
    print()
    print("─" * 50)
    print("💡 In production, this would trigger CLARISSA actions")
else:
    print("❌ No transcript available. Run the transcription cell first.")

# Parse the intent from the transcript if command_text: # Simple keyword-based intent detection text_lower = command_text.lower() # Define intents intents = [] # Property queries if any(w in text_lower for w in ['permeability', 'perm']): intents.append(('SHOW_PROPERTY', 'permeability')) if any(w in text_lower for w in ['porosity', 'poro']): intents.append(('SHOW_PROPERTY', 'porosity')) if 'pressure' in text_lower: intents.append(('SHOW_PROPERTY', 'pressure')) if any(w in text_lower for w in ['saturation', 'sat']): intents.append(('SHOW_PROPERTY', 'saturation')) if any(w in text_lower for w in ['water cut', 'watercut']): intents.append(('SHOW_METRIC', 'water_cut')) if any(w in text_lower for w in ['oil rate', 'production rate']): intents.append(('SHOW_METRIC', 'oil_rate')) # Actions if any(w in text_lower for w in ['run', 'simulate', 'execute']): intents.append(('RUN_SIMULATION', None)) if any(w in text_lower for w in ['stop', 'cancel', 'abort']): intents.append(('STOP_SIMULATION', None)) # Questions if any(w in text_lower for w in ['what is', 'what are', 'how much', 'how many']): intents.append(('QUERY', 'general')) print("═" * 50) print("🧠 INTENT ANALYSIS") print("═" * 50) print() print(f"Input: \"{command_text}\"") print() if intents: print("Detected intents:") for intent_type, intent_value in intents: if intent_value: print(f" • {intent_type}: {intent_value}") else: print(f" • {intent_type}") else: print(" ⚠️ No specific intent detected") print(" → Would pass to LLM for interpretation") print() print("─" * 50) print("💡 In production, this would trigger CLARISSA actions") else: print("❌ No transcript available. Run the transcription cell first.")

---

🔗 Resources¶

Resource	Link
🎤 Voice Demo	irena-40cc50.gitlab.io/demos/voice-demo.html
🎬 Recording Suite	irena-40cc50.gitlab.io/demos/demo-recording-suite.html
📚 CLARISSA Docs	irena-40cc50.gitlab.io
💻 Source Code	GitLab Repository

---

📋 Example Commands to Try¶

• "Show me the permeability distribution"
• "What is the current water cut?"
• "Display pressure in layer 3"
• "Run the simulation"
• "What are the oil production rates?"