Discussions

Issue Type: Technical Support Request
📋 Form Fields:

1. Product/Service:
   HeyGen Streaming Avatar SDK
2. SDK Version:
   @heygen/streaming-avatar v4.0.8
3. API Endpoint:
   /v1/streaming.new
4. Avatar ID:
   Katya_Chair_Sitting_public
5. Task Type:
   TALK
6. Issue Description:
   The knowledgeBase parameter in the HeyGen Streaming Avatar SDK is not being applied to the avatar's behavior. The avatar continues to speak in demo mode instead of using the custom knowledge base provided.
7. Configuration Used:

{
"quality": "high",
"avatarName": "Katya_Chair_Sitting_public",
"knowledgeBase": "You are Dr. AI Elena, a professional sexual therapy counselor. You provide expert guidance on intimate relationships, sexual health, and emotional well-being. Always maintain professional boundaries while being empathetic and supportive.",
"systemPrompt": "You are Dr. AI Elena, a professional sexual therapy counselor...",
"customInstructions": "You are Dr. AI Elena, a professional sexual therapy counselor...",
"knowledgeId": "Katya_Chair_Sitting_public",
"task_type": "TALK",
"voice": {
"voiceId": "default",
"rate": 1.0,
"emotion": "friendly"
}
}

8. Current Behavior:
   The avatar continues to speak in demo mode, ignoring the custom knowledge base provided.
9. Expected Behavior:
   The avatar should adopt the personality and expertise defined in the knowledgeBase parameter.
10. Steps to Reproduce:
    Initialize HeyGen Streaming Avatar SDK
    Set task_type to TALK
    Provide custom knowledgeBase parameter
    Start avatar session
    Avatar speaks in demo mode instead of using custom knowledge base
11. Additional Information:
    Multiple system prompt variations have been attempted
    knowledgeId parameter has been included
    TALK task type is being used as required
    All parameters are being sent correctly to the API
12. Questions:
    Is there an additional configuration required for the knowledgeBase parameter to work?
    Are there any specific requirements or limitations for the knowledge base content?
    Should we use a different parameter or approach for customizing avatar behavior?
13. Priority:
    High - This is blocking our production implementation