​
Platform Concepts

An overview of the key components that make up CallHQ Voice AI agents and the tasks these agents are capable of handling.

​
Voice AI Agent Architecture

CallHQ allows you to create intelligent voice agents that can interact in real time and perform useful post-call actions. A CallHQ agent follows a modular pipeline:

​
1. Input Medium

The input method determines how the user interacts with the agent:
  • Voice: Via phone call or embedded web widget (microphone input)
  • Text: (Coming soon) Typed input via web chat
  • Visual: (Coming soon) Upload images or documents as part of the interaction

​
2. ASR (Automatic Speech Recognition)

Converts spoken input into text format. This transcription is sent to the LLM for further processing.

​
3. LLM (Large Language Model)

The LLM processes the transcribed text and generates a relevant response based on the agent’s instructions, tone, and logic. This response is routed to the output processor.

​
4. Output Generator

Depending on the type of conversation:
  • For voice: TTS (Text-to-Speech) converts LLM output into natural audio
  • For visual or text output: responses may be rendered visually or as plain text (Coming soon)

​
5. Output Medium

Delivers the generated output to the user:
  • Voice: Played back on a call or in-browser via the web widget
  • Text/Visual: (Coming soon) Rendered as chat or rich content

​
Tasks After Conversation

CallHQ agents can also be configured to perform additional tasks once the primary conversation ends:

​
Post-Call Actions

  • Call Summarization
    Automatically summarize the entire conversation for review or CRM logging.
  • Information Extraction
    Extract structured fields such as customer name, reason for call, appointment time, etc.
  • Function Tools (APIs & Triggers)
    Use function tools to invoke APIs or external actions—e.g., booking a slot, sending a webhook, or updating a database.
These components work together seamlessly, but you can swap and configure models (e.g., use Deepgram for ASR or ElevenLabs for TTS) via your CallHQ dashboard or API.