Developer Hub

Find everything you need to start building.

APIs, SDKs, cookbooks, and reference material for shipping sovereign AI into production — on your infrastructure, under your control.

Get API Key Read the quickstart

Build with Kervalt

Tools for every stage of integration

API Quickstarts

Make your first call to Tafari Core in under five minutes. Step-by-step guides for authentication, chat completions, embeddings, and reranking.

Start building →

Python SDK

A typed, idiomatic client for every Kervalt endpoint. Streaming, retries, and batch workloads built in. pip install kervalt

View on PyPI →

TypeScript SDK

First-class support for Node.js and edge runtimes, with full type inference across request and response payloads. npm install @kervalt/sdk

View on npm →

API Reference

Complete endpoint documentation with rate limits, error semantics, tokenizers, and per-model parameters for Tafari Core, Zuberi Vector, and Sankofa Rank.

Browse the reference →

Cookbooks

Pre-built samples that integrate Zuberi Vector and Sankofa Rank into real data pipelines — ingestion, embedding, hybrid retrieval, and reranking at production scale.

Open the cookbooks →

LLM University

Our educational portal for AI engineers: vector databases, dense-sparse retrieval, agentic tool calls, and prompt optimization — from fundamentals to advanced patterns.

Start learning →

Quickstart

Your first call to Tafari Core

One endpoint, one API key, full sovereignty. Point the client at api.kervalt.be — or at your own Kanda Vault deployment — and the same code runs unchanged.

01 Provision an API key from your Kervalt console.
02 Send a chat completion to Tafari Core over TLS.
03 Swap the base URL to your VPC endpoint when you go to production.

Get API Key

quickstart.py

# curl
curl https://api.kervalt.be/v1/chat/completions \
  -H "Authorization: Bearer $KERVALT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tafari-core",
    "messages": [
      {"role": "user", "content": "Summarize this contract."}
    ]
  }'

# Python SDK
from kervalt import Kervalt

client = Kervalt()  # reads KERVALT_API_KEY
response = client.chat.completions.create(
    model="tafari-core",
    messages=[{"role": "user", "content": "Summarize this contract."}],
)
print(response.choices[0].message.content)

Why build on Kervalt

Your models. Your cloud. Your control.

Safe

Your data stays under your control with multi-layered protection and industry-certified security standards.

Flexible

Secure within your virtual private cloud (VPC), on-premises, or dedicated, Kervalt-managed Kanda Vault.

Independent

Train on your proprietary data and build unique AI solutions made for your use cases, needs, and infrastructure.