Skip to Content
Althea Studio

Models

The Models layer is the foundational component of our software stack, providing the core capabilities that enable our AI agents to perform their tasks. These models are responsible for processing and understanding various types of input, generating responses, and performing specialized functions.

Models

  1. Voice Intake:

    • Description: This model processes and understands voice inputs from users, converting spoken language into a format that can be used by other components of the system. It includes several sub-components:
      • STT (Speech-to-Text): Converts spoken language into written text.
      • Voice Tagging: Labels and categorizes different parts of the conversation for better analysis and understanding.
      • Denoising: Reduces background noise to improve the clarity of voice inputs.
      • Turn Prediction: Anticipates user responses to improve the efficiency of interactions.

  2. NLU (Natural Language Understanding):

    • Description: This model enables the AI agent to comprehend and interpret user inputs, understanding the intent and context behind the words. It is based on Large Language Models (LLMs).

  3. NLG (Natural Language Generation):

    • Description: This model generates human-like responses, enabling the AI agent to communicate effectively with users. It is based on Large Language Models (LLMs).

  4. RAG (Retrieval-Augmented Generation):

    • Description: This model combines retrieval-based methods with generation capabilities to provide accurate and contextually relevant responses. It is based on Large Language Models (LLMs).

  5. LMMs (Large Multimodal Models):

    • Description: These models integrate multiple data types (e.g., text, images, audio) to provide a comprehensive understanding of user inputs and generate appropriate responses. We offer the utilization of off-the-shelf vision-based LMMs from providers such as OpenAI, Anthropic, and Google, while our own voice-based LMMs are currently available for voice-text to text and will soon support voice-text to voice-text.

  6. LLMs (Large Language Models):

    • Description: These models provide advanced language processing capabilities, enabling the AI agent to handle complex language tasks and generate sophisticated responses.

  7. SLMs (Specialized Language Models):

    • Description: These models are tailored for specific healthcare-related language processing tasks, ensuring accurate and relevant responses in specialized contexts.

Voice Engine

The Voice Engine manages the interaction between the user and the AI seamlessly, particularly when customers choose to use our proprietary voice models. It includes:

  • Voice Intake: Manages the STT, voice tagging, denoising, and turn prediction to ensure clear and efficient voice interactions.
  • TTS (Text-to-Speech): Converts text responses generated by the AI into natural-sounding speech.

Controls

For models, there are also parameters that can be configured to optimize their performance and behavior. These controls include:

  • Model Selection: Users can choose between using Althea's proprietary AI models or well-known supported engines for STT and TTS.
    • STT Options: Althea, Azure, Google, AWS, Deepgram
    • TTS Options: Azure, Google, AWS, Deepgram, 11 Labs, OpenAI
  • Output Format: Define the format in which the model's output should be presented, ensuring compatibility with other system components.
  • Voice-Related Parameters: Configure parameters such as standard delay, turn prediction, and voice tagging to enhance the interaction experience.
  • LLM Selection: Users can choose from supported Large Language Models (LLMs) to power their widgets and workflows.
    • LLM Options: Althea, OpenAI, Anthropic, Google

By leveraging these detailed models and controls, our AI agents can effectively process and understand various types of input, generate accurate and relevant responses, and perform specialized functions, ensuring a seamless and efficient experience for both patients and healthcare providers.

previousWidgets introduction