Skip to content
Start freeGPT-5 · Claude · DeepSeek V4 · Qwen3 — in ₹One OpenAI-compatible endpointBilled in rupeesGST invoiceNo international cardGet started →Start freeGPT-5 · Claude · DeepSeek V4 · Qwen3 — in ₹One OpenAI-compatible endpointBilled in rupeesGST invoiceNo international cardGet started →
Use Cases

WhatsApp Chatbot API in India | unoblox

Build a WhatsApp chatbot backed by unoblox's rupee-billed AI API — architecture, model picks by budget, and honest data-handling notes.

A WhatsApp chatbot is really two pieces bolted together: the WhatsApp Business messaging layer that sends and receives messages, and a language model that decides what to reply. unoblox is the second half — one OpenAI-compatible endpoint, billed in rupees, that plugs behind your WhatsApp webhook to handle order status questions, lead qualification, FAQs, or support triage without you managing a separate AI vendor relationship per model.

Why teams pair WhatsApp with an AI API

WhatsApp is where a large share of Indian customers already expect to reach a business, but hand-written reply trees run out of coverage fast — anything outside the scripted paths either goes unanswered or gets escalated to a human. Routing free-text messages through a model lets the bot understand intent more flexibly, while still handing off to a person when it's unsure.

A simple architecture

  • Inbound: WhatsApp Business API (or Cloud API) delivers the incoming message to your webhook.
  • Decision layer: your webhook calls https://api.unoblox.ai/v1/chat/completions with the conversation so far and, optionally, a tools definition for order lookups or CRM actions.
  • Outbound: your webhook sends the model's reply back through the WhatsApp API.
  • Escalation: a simple confidence or keyword check routes anything the model can't resolve to a human agent.
curl https://api.unoblox.ai/v1/chat/completions \
  -H "Authorization: Bearer ub-gw-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{"model":"deepseek-ai/deepseek-v4-flash","messages":[{"role":"system","content":"You are a helpful support assistant for an online store."},{"role":"user","content":"Where is my order 4021?"}]}'

Choosing a model by conversation volume

WhatsApp bots tend to run high message volumes with short exchanges, which makes per-token cost matter more than raw model size:

ModelInput (₹/1M)Output (₹/1M)Fit
Qwen3 1.7BFree (₹0)Free (₹0)simple FAQ bots, high volume
DeepSeek V4 Flash₹9.07₹18.14everyday support conversations
Llama 4 Scout₹10.08₹30.24broader intent coverage
GPT-5 mini₹25.2₹201.6more nuanced, brand-sensitive replies

Anything else in the catalog, including GPT-4o mini and Claude Haiku, works the same way through the same endpoint — check /models for current ₹ pricing.

Tone, language, and escalation

Keep the system prompt tight about what the bot should and shouldn't promise — refunds, delivery dates, and pricing are common places a bot can overreach. Indian WhatsApp support often mixes Hindi, English, and regional languages in the same thread; most catalog models handle code-mixed text reasonably, but test with your own real conversation samples rather than assuming coverage, since we don't publish language-specific accuracy figures.

Frequently asked questions

Does unoblox connect directly to WhatsApp? No — unoblox is the AI layer behind your bot. You still integrate with WhatsApp Business API or Cloud API separately and call unoblox from your own webhook.

Is customer chat data stored in India? For unoblox's own small hosted models, like the free Qwen3 1.7B, processing happens on Indian infrastructure. For third-party models such as GPT or Claude, the India benefit is rupee billing, one GST invoice, and no international card — not Indian data residency, so avoid routing anything you can't send abroad to those models.

Can the bot hand off to a human agent? Yes, but that logic lives in your own application — have it watch for low-confidence replies or specific keywords and route the conversation to a live agent queue.

What does a WhatsApp bot cost to run on unoblox? It depends on message volume and model choice; short support replies on a model like DeepSeek V4 Flash cost a few rupees per thousand exchanges at published per-token rates — check /models for exact figures at your volume.

Can I use the same setup for outbound broadcast messages? The AI layer can draft or personalise message text, but WhatsApp's own rules for template messages and opt-in apply regardless of which model generates the content.

Get started in rupees → https://unoblox.ai/sign-in

whatsapp chatbot api indiause-casescustomer support ai
ShareLinkedInWhatsAppTelegram

Start building in rupees

Call every major model through one OpenAI-compatible endpoint, billed in ₹ on a GST invoice.