Agents

​

Overview

​

Agent structure

​

Accessing agent data

curl https://api.iqra.ai/api/v1/business/12345 \
  -H "Authorization: Token YOUR_API_KEY"

​

Example agent response

{
  "Id": "agent_abc123",
  "General": {
    "Name": "Sales Agent",
    "Description": "Handles customer sales inquiries",
    "Enabled": true
  },
  "Context": {
    "Variables": {
      "CompanyName": "Acme Corp",
      "ProductLine": "Enterprise Software"
    }
  },
  "Personality": {
    "Tone": "professional and friendly",
    "Style": "consultative sales approach",
    "Constraints": [
      "Never discuss pricing without qualification",
      "Always offer to schedule a demo"
    ]
  },
  "Utterances": {
    "Greeting": "Hello! Thanks for calling Acme Corp. How can I help you today?",
    "Fallback": "I'm not sure I understood that. Could you rephrase?",
    "Goodbye": "Thank you for your time. Have a great day!"
  },
  "Interruptions": {
    "Enabled": true,
    "TurnEndViaAI": {
      "Enabled": true
    }
  },
  "KnowledgeBase": {
    "Enabled": true,
    "References": ["kb_xyz789"]
  },
  "Integrations": {
    "STT": {
      "en": ["deepgram_integration_id"]
    },
    "LLM": {
      "en": ["openai_integration_id"]
    },
    "TTS": {
      "en": ["elevenlabs_integration_id"]
    }
  },
  "Settings": {
    "BackgroundAudioUrl": "https://s3.amazonaws.com/audio.mp3",
    "MaxDuration": 1800,
    "Language": "en"
  }
}

​

Agent configuration best practices

​

Context and personality

• Keep context variables specific and relevant to the agent’s purpose
• Define clear personality traits that align with your brand
• Set appropriate constraints to keep conversations on track

​

Utterances

• Write natural, conversational greetings and responses
• Test utterances with different customer types
• Localize utterances for multi-language support

​

Interruptions

• Enable interruptions for natural conversation flow
• Use AI-based turn detection for better accuracy
• Configure verification to reduce false positives

​

Knowledge base

• Link relevant knowledge bases to provide accurate information
• Use query refinement to improve search accuracy
• Monitor knowledge base usage and update content regularly

​

Integrations

• Configure STT, LLM, and TTS per language for best results
• Test different AI models to find the best fit for your use case
• Monitor integration performance and costs

​

Notes

• Background audio URLs are pre-signed and expire after approximately 8 hours
• Agents can be enabled/disabled without deleting configuration
• Multi-language agents require separate STT/LLM/TTS integrations per language
• Agent settings are inherited from business-level defaults unless overridden