N-ATLAS
Developer Gateway
Build Nigerian AI with N-ATLAS.
Developer infrastructure for integrating, testing and building Nigerian-language AI applications with N-ATLAS — the Llama-3 8B model released by Awarri Technologies and the Federal Ministry of Communications, Innovation and Digital Economy.
- Model
- NCAIR1/N-ATLaS
- Weights
- bf16 · 8B
- Serving
- vLLM · GPU
- API
- REST + SSE
curl -X POST https://natlas.aiwaziri.com/api/v1/chat \
-H "Content-Type: application/json" \
-d '{
"model": "N-ATLaS",
"language": "hausa",
"messages": [
{ "role": "user", "content": "Menene basirar wucin gadi?" }
]
}'pip install "git+https://github.com/AIWAZIRI-LIMITED/aiwaziri-natlas#subdirectory=packages/python-sdk"
from aiwaziri_natlas import NAtlas
client = NAtlas(base_url="https://natlas.aiwaziri.com")
response = client.chat(message="Menene basirar wucin gadi?", language="hausa")
print(response.text)
print(response.latency_ms, response.request_id)SUPPORTED BY N-ATLAS
Four languages, one model
Languages listed on the official N-ATLaS model card. Quality varies by language; see the model card's own human-evaluation results and our limitations.
en
English
ha
Hausa
yo
Yorùbá
ig
Igbo
WHAT YOU GET
From model weights to a developer platform
N-ATLaS is published as model weights. Running an 8B model needs GPUs, serving infrastructure and integration work. The gateway provides that layer so developers can call N-ATLaS over HTTP or from Python in minutes.
Playground
Run real N-ATLaS inference in the browser with streaming, latency and request IDs.
Open →
Language Lab
Send one prompt in English, Hausa, Yorùbá and Igbo and compare outputs side by side.
Open →
REST API
Versioned /api/v1 endpoints for chat, generation, models and health. Streaming over SSE.
Open →
Python SDK
aiwaziri-natlas: typed client with streaming and structured errors.
Open →
Documentation
Quick start, integration guide, error reference, deployment and limitations.
Open →
Live status
Health endpoint probes the GPU worker and reports whether N-ATLaS is actually being served.
Open →
ARCHITECTURE
How a request reaches N-ATLaS
01
Developer
Browser playground, curl, or the Python SDK
02
natlas.aiwaziri.com
Cloudflare Pages: web app + same-origin /api proxy
03
AIWAZIRI Gateway
FastAPI: validation, language prompts, rate limits, errors, SSE
04
N-ATLaS on GPU
vLLM serving NCAIR1/N-ATLaS (bf16) on a Modal GPU
Inference credentials live only in server-side secrets. The browser talks to the gateway; the gateway talks to the GPU worker.
PROJECT STATUS
Public beta, built by AIWAZIRI LIMITED for the National AI Innovation Challenge (Innovation & Enterprise Track, PS1 — Developer Infrastructure). Not a commercial service. N-ATLaS is used under its Terms of Use v1.0, which include a 1,000 active-user cap. The GPU scales to zero when idle, so the first request after a quiet period can take a few minutes.