Voice Mental Performance Coach

A voice-first AI-powered mental performance coach for athletes, built with Streamlit, CrewAI, Ollama, and voice technologies. The app allows athletes to create profiles, engage in interactive voice-based consultations, receive personalized coaching recommendations, and get daily audio reminders. It focuses on mental aspects like confidence, stress management, focus, and motivation.

Features

• Interactive Voice Consultation: Natural conversation with Coach Alex using speech recognition and TTS.
• Personalized Coaching: Generates tailored mental performance strategies based on athlete profile and responses.
• Data Collection: Collects athlete details (age, sport, goals) through conversational flow.
• Assessment Questions: Asks and processes responses to personalized questions.
• Coaching Analysis: Provides comprehensive analysis with key insights and action plans.
• Audio Output: Uses ElevenLabs TTS for high-quality voice (fallback to gTTS or pyttsx3).
• Speech Recognition: Supports voice input using microphone.
• Database Persistence: Saves athlete profiles and consultation logs in SQLite.
• Daily Reminders: Generates and plays audio reminders with coaching tips.
• Streamlit UI: User-friendly interface for profile setup, assessment, and coaching.
• Pygame Audio Playback: Handles audio playback for a seamless experience.

Requirements

• Hardware:
  • Microphone for voice input.
  • Speakers/headphones for audio output.
  • Quiet environment for speech recognition.
• Software:
  • Python 3.10 or higher (tested on 3.11).
  • Ollama with a model (e.g., llama3.2:latest or phi-2 for lower memory usage).
  • Internet connection for ElevenLabs TTS (optional; fallback to offline TTS).
• API Keys:
  • ElevenLabs API key (for high-quality TTS; get from ElevenLabs).
• System Memory:
  • At least 4 GB RAM (for llama3.2:latest) or 2 GB (for phi-2).

Installation

1. Clone the Repository (if applicable):

   git clone <repository_url>
   cd voice.py
   

2. Set Up Virtual Environment:

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   

3. Install Dependencies:

   pip install streamlit crewai langchain-ollama pyttsx3 speechrecognition gtts pygame requests
   

4. Install Ollama:

   • Download and install Ollama from https://ollama.com/download.
   • Pull and run the model:

     ollama pull llama3.2:latest  # Or phi-2 for lower memory
     ollama serve
     

   • Ensure it's running at http://localhost:11434.

5. Set ElevenLabs API Key:

   • Add to your environment:

     export ELEVENLABS_API_KEY="your_elevenlabs_api_key_here"
     

   • Or update the script's ELEVENLABS_API_KEY variable.

6. Set Up Audio Dependencies:

   • On Windows, install Microsoft Visual C++ Build Tools for pyaudio (part of speechrecognition): Download here.

Usage

1. Run the App:

   streamlit run voice.py
   

   • Open your browser to http://localhost:8501.

2. Interact with the App:

   • Profile Setup: Enter your name, age, weight, height, sport type, experience level, training frequency, and goals.
   • Welcome Phase: Listen to Coach Alex's welcome audio and respond.
   • Assessment Phase: Answer 6 personalized questions via voice or text.
   • Coaching Phase: Receive personalized coaching analysis, recommendations with confidence scores (e.g., "90%", "85%"), and action plans.
   • Audio and JSON: Play/download audio, view JSON output with recommendations and confidence scores.
   • Daily Reminders: Generate and download daily audio reminders.
   • Save Session: Save your profile and session data to SQLite.

3. Example JSON Output (includes confidence scores):

   {
     "session_id": "<uuid>",
     "athlete_name": "<name>",
     "sport_type": "<sport>",
     "athlete_level": "<level>",
     "training_frequency": <number>,
     "goals": ["<goal1>", "<goal2>"],
     "responses": {
       "question_1": "<response>",
       ...
     },
     "recommendations": [
       {
         "recommendation": "<recommendation>",
         "justification": "<justification>",
         "confidence_score": "90%"
       },
       ...
     ]
   }

4. Audio Playback:

   • Uses pygame for automatic playback.
   • Audio includes recommendations with confidence scores (e.g., "Practice deep breathing... with 90% confidence").

Configuration

• Ollama Model: Change Config.MODEL_NAME to a different model (e.g., llama3.2:latest for higher accuracy, but requires more RAM).
• API Base: Update Config.API_BASE if Ollama runs on a different port.
• ElevenLabs Voice ID: Change ELEVENLABS_VOICE_ID for a different voice (e.g., "pNInz6obpgDQGcFmaJgB" for Adam).
• Database Path: Change DATABASE_PATH if you want a different SQLite file location.
• Max Response Length: Adjust MAX_RESPONSE_LENGTH for TTS response length.

Troubleshooting

• **StreamlitAPIException (Duplicate Forms