Build a Hindi voice agent on Bolna

​

Choose a Hindi transcriber

​

Choose a Hindi voice

​

Write the prompt for Hindi and Hinglish

• Write in Devanagari, not romanized Hindi. नमस्ते! आप कैसे हैं? rather than Namaste! Aap kaise ho?. Native script gives more accurate pronunciation and stops the text being treated as English.
• Keep common English loanwords in Latin script. Callers say “order”, “OTP” and “support ticket” in English, so write आपका order status check करने के लिए मुझे थोड़ा समय दीजिए। instead of forcing a pure Hindi translation.
• Set gender and politeness explicitly. For a female persona, ask for consistent feminine verb forms in Hindi.
• Read alphanumerics character by character, in English. For OTPs, registration numbers and order IDs, this avoids misheard and mispronounced digits.

​

Create a Hindi agent over the API

{
  "agent_config": {
    "agent_name": "Hindi Support Agent",
    "agent_welcome_message": "नमस्ते! मैं आपकी क्या मदद कर सकती हूँ?",
    "tasks": [{
      "task_type": "conversation",
      "toolchain": { "execution": "sequential", "pipelines": [["transcriber", "llm", "synthesizer"]] },
      "tools_config": {
        "llm_agent": {
          "agent_type": "simple_llm_agent",
          "agent_flow_type": "streaming",
          "llm_config": { "provider": "openai", "model": "gpt-4.1-mini", "max_tokens": 150, "temperature": 0.2 }
        },
        "transcriber": {
          "provider": "sarvam",
          "model": "saaras:v4",
          "language": "hi",
          "stream": true,
          "endpointing": 500,
          "keywords": "Bolna, PhonePe, Flipkart, New Delhi",
          "context": "Inbound support line for an electronics retailer in India. Callers ask about order status and returns, and often read out order numbers."
        },
        "synthesizer": {
          "provider": "sarvam",
          "provider_config": {
            "voice": "Aparna (Customer)",
            "voice_id": "aparna_hi_customer",
            "model": "bulbul:v4-flash",
            "language": "hi",
            "speed": 1.0
          },
          "stream": true,
          "buffer_size": 150
        },
        "input": { "provider": "plivo", "format": "wav" },
        "output": { "provider": "plivo", "format": "wav" }
      },
      "task_config": { "call_terminate": 90, "hangup_after_silence": 10 }
    }]
  },
  "agent_prompts": {
    "task_1": {
      "system_prompt": "आप एक सहायक सपोर्ट एजेंट हैं। संक्षेप में उत्तर दें।"
    }
  }
}

curl https://api.bolna.ai/v2/agent \
  -H "Authorization: Bearer $BOLNA_API_KEY" \
  -H "Content-Type: application/json" \
  -d @agent.json

import os, urllib.request, json

with open("agent.json") as f:
    body = json.load(f)

req = urllib.request.Request(
    "https://api.bolna.ai/v2/agent",
    data=json.dumps(body).encode(),
    headers={"Authorization": f"Bearer {os.environ['BOLNA_API_KEY']}",
             "Content-Type": "application/json"},
)
print(json.load(urllib.request.urlopen(req)))

​

Improve recognition with keywords and context

• keywords: a comma-separated list of terms callers say. On Sarvam this needs saaras:v4, and takes up to 50 terms of up to 64 characters each.
• context: a sentence or two describing the call. Sarvam reads it on every Saaras model.

​

Fix pronunciation of numbers and names

• Write how a word should sound in the script the voice speaks. For a Hindi voice, write नो योर कस्टमर rather than know your customer.
• Each language has its own list, so 56 can be “five six” in English and पाँच छह in Hindi.
• A rule only matches text the agent actually says. If the LLM writes a number out in words, a rule for the digits won’t apply, so ask the prompt to keep numbers exactly as written.

​

Serve Hindi and English on one agent

• Code-mixing within a sentence. Sarvam, Soniox, AssemblyAI and Gladia transcribe code-mixed speech, and Kalpa Labs speaks Hinglish from one model. A sentence with a few English words is still treated as Hindi.
• A full language switch. Add both English and Hindi to one agent with multilingual_config, each with its own transcriber, voice and prompt. Hindi is one of the languages Bolna detects automatically mid-call. The multilingual quickstart builds an English and Hindi agent over the API.

• Start a call in Hindi with "agent_data": { "language": "hi" } on POST /call, or an agent_language variable in a batch CSV.
• Let system messages such as the hangup and “are you still there?” checks follow the active language.
• Read How language switching works for what does and doesn’t trigger a switch.

​

Phone numbers for Hindi agents in India

• Buy an Indian number in the dashboard through Plivo or Vobiz.
• For commercial or telemarketing calls in India, read How to get 140 & 160-series phone numbers in India for the TRAI and DLT requirements.
• To use numbers from your own carrier, connect it with a SIP trunk.

​

Next steps