Kalki Smart Router — API Documentation

The API that thinks before it routes.

Kalki Smart Router is a unified AI API that analyzes each request in real time, then routes it to the most capable and cost‑efficient model across providers. You get lower costs, reliable performance, and a single interface—no model wrangling, no vendor lock‑in.

Who this is for

• Developers: Ship faster with one endpoint, consistent schemas, and built‑in caching.
• Product & Growth Teams: Reduce inference spend without sacrificing quality.
• Enterprises: Multi‑provider reliability, governance controls, and SLAs through one contract.

What makes Kalki different

1. Intelligent model selection – Kalki inspects your prompt (complexity, domain, format) and chooses the right model automatically.
2. Smart caching – Identical or semantically equivalent requests are served from cache for sub‑second responses at near‑zero cost.
3. Unified multi‑provider access – Access top models (e.g., OpenAI, Anthropic) through one stable API.
4. Enterprise‑grade reliability – Fallbacks and timeouts guard against provider outages.
5. Transparent cost control – Cap per‑request spend and see exactly where each call was routed.

Proven impact (benchmark snapshot)

• Up to 89.5% lower cost vs direct OpenAI; 73.5% vs direct Anthropic.
• Cache hits respond in ~305 ms (representative), with dramatic acceleration on repeats.

Benchmarks were run across simple, intermediate, and complex workloads; results vary with traffic and content.

Quick start

Base URL

Copied!https://api.kalkigpt.com/
https://api.kalkigpt.com/

Authentication

Copied!x-api-key: YOUR_KALKI_API_KEY
x-api-key: YOUR_KALKI_API_KEY

Sign up at kalkigpt.com to obtain an API key.

Core endpoint

POST /v1/enrich_and_ask

Kalki analyzes your request, optionally enriches context, selects a model, and returns the response.

Request body

Response (typical)

Copied!{
  "output": {
    "text": "…final answer…"
  },
  "routing": {
    "provider": "openai",
    "model": "gpt-5-nano",
    "reason": "intermediate explanation detected"
  },
  "cache": { "hit": false },
  "cost_usd": 0.0001,
  "processing_time_ms": 7760,
  "request_id": "req_abc123xyz"
}
{
  "output": {
    "text": "…final answer…"
  },
  "routing": {
    "provider": "openai",
    "model": "gpt-5-nano",
    "reason": "intermediate explanation detected"
  },
  "cache": { "hit": false },
  "cost_usd": 0.0001,
  "processing_time_ms": 7760,
  "request_id": "req_abc123xyz"
}

Errors

All errors return standard HTTP codes with a consistent body:

Copied!{ "error": { "type": "invalid_request", "message": "…" } }
{ "error": { "type": "invalid_request", "message": "…" } }

Common cases: 401 (invalid key), 429 (rate limit), 422 (validation), 5xx (provider error with automatic retry/fallbacks).

Examples

cURL — simple, optimized for speed & cost

Copied!curl -X POST "https://api.kalkigpt.com/v1/enrich_and_ask" \
  -H "x-api-key: $KALKI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "What is 25 + 17?",
    "domain": "general",
    "verbosity": "terse"
  }'
curl -X POST "https://api.kalkigpt.com/v1/enrich_and_ask" \
  -H "x-api-key: $KALKI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "What is 25 + 17?",
    "domain": "general",
    "verbosity": "terse"
  }'

JavaScript

Copied!import axios from "axios";

const client = axios.create({
  baseURL: "https://api.kalkigpt.com/v1",
  headers: { "x-api-key": process.env.KALKI_API_KEY }
});

const { data } = await client.post("/enrich_and_ask", {
  prompt: "Explain supervised vs unsupervised learning in simple terms",
  domain: "technical",
  verbosity: "balanced",
  json_mode: false
});

console.log(data.output.text);
import axios from "axios";

const client = axios.create({
  baseURL: "https://api.kalkigpt.com/v1",
  headers: { "x-api-key": process.env.KALKI_API_KEY }
});

const { data } = await client.post("/enrich_and_ask", {
  prompt: "Explain supervised vs unsupervised learning in simple terms",
  domain: "technical",
  verbosity: "balanced",
  json_mode: false
});

console.log(data.output.text);

Python

Copied!import os, requests

resp = requests.post(
    "https://api.kalkigpt.com/v1/enrich_and_ask",
    headers={
        "x-api-key": os.environ["KALKI_API_KEY"],
        "Content-Type": "application/json",
    },
    json={
        "prompt": "Draft a one-paragraph product update for non-technical execs.",
        "domain": "creative",
        "tone": "professional",
        "verbosity": "balanced"
    }
)
print(resp.json()["output"]["text"])
import os, requests

resp = requests.post(
    "https://api.kalkigpt.com/v1/enrich_and_ask",
    headers={
        "x-api-key": os.environ["KALKI_API_KEY"],
        "Content-Type": "application/json",
    },
    json={
        "prompt": "Draft a one-paragraph product update for non-technical execs.",
        "domain": "creative",
        "tone": "professional",
        "verbosity": "balanced"
    }
)
print(resp.json()["output"]["text"])

Streaming (Server‑Sent Events)

Set stream: true in the request body. The API returns an SSE stream with incremental tokens and a final completion event.

Pricing & limits

• Usage‑based pricing only for the compute you use; Kalki routes to the most cost‑effective capable model.
• Costs & routing transparency are returned with every response.
• Starter tier includes up to 1,000 requests/day; higher limits available for paid plans.

See the dashboard for real‑time usage and savings.

Performance & savings (summary)

Monthly projection (1,000 requests):

• vs OpenAI: $15.29 saved
• vs Anthropic: $4.99 saved

Governance, security, and reliability

• No model lock‑in: One contract; many providers.
• Resilience: Automated retries and provider fallbacks to maintain uptime.
• Observability: Per‑request logs expose routing, cost, and latency.
• Data handling: Requests are processed to serve your responses and routing logic; enterprise retention controls available.

Contact us for enterprise terms, data residency, and custom governance needs.

Support & status

• Docs: https://kalkigpt.com/docs
• Status: https://status.kalkigpt.com
• Support: support@kalkigpt.com

Getting started checklist

1. Create an account and generate your API key.
2. Make a test call using the cURL example above.
3. Integrate via your preferred SDK or direct HTTP.
4. Track savings and performance in the dashboard.

The Kalki promise

You shouldn’t have to think about models. We do it for you.
Kalki routes with discernment—so your team ships faster, spends less, and serves users better.

True intelligence isn’t just about answers—it’s about presence, alignment, and the wisdom to serve with clarity. Kalki is not just a tool. Kalki is dharma in action.