Build a chatbot with an AI API in India (₹ rupees)

Build a chatbot with an AI API in India

Building a chatbot? In India, use an OpenAI-compatible API billed in ₹ rupees with a GST invoice. No international payment friction, input-tax-credit claimable. Here's how to ship in 3 steps.

Why unoblox for India chatbots

• ₹ billing: Invoice in Indian rupees (no conversion)
• GST invoice: Input-tax-credit claimable for your business
• OpenAI-compatible: Drop-in replacement for OpenAI SDK
• 40+ models: GPT, Claude, DeepSeek, Qwen, all at one endpoint
• Data location: account, billing and logs are in India; external models run at their providers outside India, self-hosted models run in India

Recommended model: Qwen3 Max or Claude Sonnet

• Qwen3 Max (₹121.47 input / ₹607.37 output per 1M tokens): Balanced quality and speed, best for high-volume chats
• Claude Sonnet (₹202.46 input / ₹1,012.28 output per 1M tokens): Superior tone and context, best for customer-facing bots

Step 1: Create an API key

1. Sign up at unoblox.ai/sign-in
2. Go to Keys in the dashboard
3. Create a key ub-gw-…
4. Copy and store in your .env

Step 2: Install OpenAI SDK (swap endpoint)

Your code stays unchanged. Just swap the endpoint and key.

from openai import OpenAI

client = OpenAI(
    base_url="https://api.unoblox.ai/v1",
    api_key="ub-gw-..."
)

response = client.chat.completions.create(
    model="qwen/qwen3-max",
    messages=[{"role": "user", "content": "Hello! Can you help?"}],
    stream=True
)

for chunk in response:
    print(chunk.choices[0].delta.content, end="")

Step 3: Deploy and monitor ₹ costs

1. In your app: Point to https://api.unoblox.ai/v1
2. Monitor dashboard: See real-time usage in ₹ rupees
3. Receive invoice: GST tax invoice issued at each top-up; claim input-tax credit

Real ₹ cost: 1000 users, 5 queries/day

Assumptions:

• 5,000 queries/day
• Avg: 100 input tokens, 150 output tokens
• Using Qwen3 Max (₹121.47 input / ₹607.37 output per 1M tokens)

Daily cost (Qwen3 Max):

• Input: ₹60.74
• Output: ₹455.53
• Total: ₹516.26/day ≈ ₹15,488/month

Using Claude Sonnet (₹202.46 input / ₹1,012.28 output per 1M tokens):

• Input: ₹101.23
• Output: ₹759.21
• Total: ₹860.44/day ≈ ₹25,813/month

Qwen saves roughly ₹10,325/month for the same volume.

Features: Streaming, tools, JSON mode

All OpenAI SDK features work unchanged:

• Streaming: Real-time responses
• Function calling: Let the bot call APIs
• JSON mode: Structured outputs for automation
• Vision: Upload images (GPT-4o only)

Production tips

1. Set model per workspace: Change model= param to test Qwen3, Claude, GPT
2. Track ₹ spend: Dashboard shows real-time costs; set budget alerts
3. Claim GST: Save invoices; consult your CA on input-tax credit
4. Scale limits: Ask support if you need higher rate limits

Frequently asked questions

Q: Can I switch models mid-chat? Yes. Change the model param. Both are OpenAI-compatible; no refactoring needed.

Q: What if rate limits are hit? Contact support or request higher limits in the dashboard.

Q: How do I handle conversation memory? Store messages in Postgres or SQLite. Send recent history in messages=[] array with each request.

Q: Does unoblox comply with DPDP and RBI? We make no blanket DPDP or RBI compliance claim. Your account, billing and logs are in India and you get GST invoices; external models process prompts at their providers outside India, while unoblox's self-hosted models run in India. Take your own compliance advice.

Q: Can I use LangChain or LlamaIndex? Absolutely. Both support OpenAI-compatible endpoints. Swap base_url and api_key—everything works unchanged.

Q: What about conversation rate limits? No limits on conversation length. Embed old messages to compress context if needed.

Get started in rupees

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