Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.voxworks.ai/llms.txt

Use this file to discover all available pages before exploring further.

​
Why Dynamics Matter

Natural conversation requires careful timing. Respond too quickly and you interrupt users mid-thought. Wait too long and the conversation feels awkward. Voxworks provides fine-grained control over these dynamics to create natural, human-like interactions.

​
Key Settings

SettingWhat It ControlsOptions
Response EagernessHow quickly the assistant responds after user stops speakingkeen, normal, patient
Thinking EffortHow deeply the LLM reasons about responsesfast, normal, deep
Silence ToleranceHow long to wait before prompting idle usershurried, patient, extremely patient, infinite
KeywordsTerms to improve transcription accuracyCustom word list

​
How Dynamics Work Together

During a conversation turn:
  1. User speaks — Voice activity detection (VAD) detects speech
  2. User pauses — System starts the eagerness timer
  3. Eagerness applied — Timer duration based on eagerness setting
  4. Turn completion check — AI assesses whether the user has finished their thought
  5. Response generated — LLM generates response with configured thinking effort
  6. Assistant speaks — Once the eagerness timer has expired, the Text-to-speech delivers the assistant response
  7. Idle monitoring — If user doesn’t respond, silence tolerance kicks in

​
Finding the Right Balance

The goal is natural conversation flow:
Too FastJust RightToo Slow
Interrupts usersNatural back-and-forthAwkward pauses
Feels rushedComfortable paceSeems unresponsive
Misses contextComplete understandingOver-thinking
Incorrect responsesCorrect responsesSeems unresponsive
Start with default settings and adjust based on testing with real conversations. See the individual setting pages for detailed guidance:
  • Response Eagerness — Adjust how quickly the assistant responds after the user stops speaking
  • Thinking Effort — Control the depth of reasoning for more accurate or faster responses
  • Silence Tolerance — Configure how long the assistant waits before prompting idle users

​
Per-Step Configuration

Each step in your flow can have its own dynamics settings: 
Step: Complex questions that require thoughtful consideration from both user and assistant
  - Eagerness: patient (wait for complete thoughts)
  - Thinking Effort: deep (deeper reasoning)
  - Silence Tolerance: patient

Step: Simple confirmation
  - Eagerness: keen (quick yes/no expected)
  - Thinking Effort: fast (straightforward response)
  - Silence Tolerance: hurried

​
Next Steps