Skip to content

[Suggestion] Voice Agent with proactive speech: context-aware prompting during user silence (Python) #571

Description

@deepgram-robot

What to build

A Python recipe demonstrating a Voice Agent that intelligently fills conversational silence — when the user pauses for an extended period, the agent proactively offers helpful prompts, clarifications, or suggestions based on the conversation context, rather than waiting indefinitely.

Why this matters

Most voice agent implementations sit silently when the user stops talking, creating an awkward experience that mimics a phone call where nobody speaks. In real human conversations, the other party would naturally fill silence with helpful cues ("Would you like me to repeat that?", "I can also help with X if you're interested", "Take your time — I'm here when you're ready"). This pattern is especially important for accessibility (users who need more processing time), complex workflows (users reviewing information), and customer support (users looking up account details). This recipe showcases Deepgram's Voice Agent capabilities in a way that produces noticeably more natural conversations.

Suggested scope

  • Language: Python
  • Deepgram APIs: Voice Agent API
  • Features: Configurable silence thresholds (short pause vs long silence vs very long silence), context-aware prompt generation (different prompts based on conversation state), escalating helpfulness (first prompt is gentle, subsequent prompts offer more guidance), silence categorization (thinking pause vs confused silence vs distracted)
  • Conversation states: Waiting for initial response, waiting for decision, waiting for information lookup, idle
  • Complexity: Moderate — silence detection + context tracking + dynamic prompt generation

Acceptance criteria

  • Runnable with minimal setup (clone, add API key, run)
  • README explains the pattern clearly
  • Uses current SDK version
  • Demonstrates at least 3 different silence response strategies
  • Configurable silence thresholds and prompt templates
  • Shows how conversation context influences the proactive prompt content
  • Does not interrupt the user when they begin speaking (respects endpointing)

Raised by the DX intelligence system.


Queued by PM — Engineer will pick this up as a priority:user recipe.

Metadata

Metadata

Assignees

No one assigned

    Labels

    action:generateQueue action: generate code examplespriority:userUser-submitted request — prioritizetype:queueIssue is in the automation queue

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions