Agent Goes Silent After Greeting

Your agent responds to the initial greeting but then stops responding to the caller’s inputs.

Symptoms

• Agent picks up the call and delivers the initial greeting message
• After the caller responds, the agent becomes unresponsive
• No error messages appear in the dashboard
• The call doesn’t disconnect, but the agent won’t speak or act further
• This happens consistently across multiple test calls

Common Causes

1. Using GPT Nano model — Do not use. The GPT Nano model has persistent failures in production and frequently becomes unresponsive after the initial prompt.
2. Exhausted OpenAI API key — If you’re using a “Bring Your Own” (BYO) OpenAI API key, it may have reached its rate limit or credit limit.
3. Poor prompt structure — Overly complex, contradictory, or poorly formatted system prompts cause the model to “freeze”.
4. Knowledge base too large — Massive knowledge bases (>50MB) can cause timeout issues and processing delays.

How to Fix

Step 1: Check Your LLM Model Selection

1. Go to Agent Settings → Model Configuration
2. Check which model you’re currently using
3. If you’re using GPT Nano: Immediately switch to GPT 4.1-mini (recommended for most use cases) or GPT 4.1 (if you need higher reasoning capability)
4. Save changes and make a fresh test call

Step 2: Verify Your OpenAI API Key (if using BYO)

1. Navigate to Integrations → OpenAI
2. Check if you have a custom API key configured
3. Log into your OpenAI account  and verify:
   • Your account has available credits
   • The API key hasn’t expired
   • Your usage hasn’t exceeded your rate limits
4. If key is exhausted: Remove the custom key to fall back to Dialora’s default OpenAI integration
5. Click Save and test again

Step 3: Simplify Your Agent Prompt

1. Go to Agent Configuration → System Prompt
2. Review your prompt for:
   • Contradictory instructions (e.g., “be helpful but never answer questions”)
   • Excessive nested conditions or complex logic
   • Instructions that conflict with the agent’s core functions
3. Simplify to 3–5 core instructions maximum
4. Test with a cleaner version of your prompt

Step 4: Reduce Knowledge Base Size

1. Go to Knowledge Base → Manage Files
2. Check the total size of your uploaded files
3. If exceeding 50MB:
   • Remove outdated or redundant documents
   • Split large PDFs into smaller chunks
   • Archive infrequently used content
4. Re-upload your optimized knowledge base
5. Test the agent again