Private AI,
running on our own GPU.
Weststar AI is a suite of production APIs — language, vision, speech, agents, security scanning and live video analytics — served from a single on-prem NVIDIA GX10. One token, one base URL, and no third-party AI clouds in the loop.
Every service, one platform.
Each API is documented with a live “try it” console. Pick one to see its endpoints, parameters and examples.
Auth & Tokens
One bearer token per domain that works on every Weststar API — with per-domain usage and quota reporting.
Explore docs →LLM
Direct, OpenAI-compatible access to the language models on the GX10 — chat, completions and embeddings.
Explore docs →Chatbot
A managed conversational layer on the LLM — persistent conversations, memory and system personas.
Explore docs →Agents
Configurable RAG chat agents with their own knowledge base and permissioned tools — grounded, cited answers over REST.
Explore docs →Visual
On-device image generation, reference-guided generation, instruction edits, enhancement and cartoon characters.
Explore docs →Sound
Transcribe speech to text, synthesize natural speech, and translate spoken audio.
Explore docs →Architect
Turn a one-line requirement into a dark-themed Mermaid architecture diagram as self-contained HTML.
Explore docs →Org Chart
Describe an organisation, committee or team in one line and get back a hierarchy chart — no JavaScript, iframe-safe.
Explore docs →Timeline
Describe a delivery plan or roadmap and get a Gantt chart with status colours, milestones and a “today” marker.
Explore docs →Face
Registration, recognition, 1:1 comparison, emotion prediction and 1:N identification with stranger enrolment.
Explore docs →SenseTime
SenseStudio HTTP-Push middleware — every event is ingested, reviewed by the GX10 LLM and turned into an analytics API.
Explore docs →Inference
CP Monitoring bridge — pushes fire, PPE, intrusion and face detections to CP’s ingest webhook as Incidents.
Explore docs →Live Video
go2rtc on the GX10 turns camera RTSP into browser-playable WebRTC / MSE over HTTPS — embed with one iframe.
Explore docs →AIVA Inference
YOLO object detection on live camera streams — each detection becomes JSON + an annotated snapshot pushed to your webhook.
Explore docs →Crawler
Crawl websites and scrape pages into clean, LLM-ready text or Markdown.
Explore docs →News
Give it a topic — get a JSON array of current stories with headline, source, date, image and full text.
Explore docs →Site Health
Is the site reachable, browsable — or defaced and hijacked? Fast checks, signatures and a local-LLM verdict.
Explore docs →Security Assessment
Authorized, non-destructive web security assessment, written up by a local LLM into a PDF report emailed to you.
Explore docs →SentinelScan
LLM-powered vulnerability assessment — Nuclei, WSTG checks and CVE correlation, with a comprehensive PDF. Nothing stored.
Explore docs →LogicDB
Schemaless NoSQL with a logic engine, versioned time-travel and plain-language questions over your data.
Explore docs →Marine
Tides, weather, waves and an AI go/no-go advisory in Bahasa Melayu for the Terengganu coast — in one call.
Explore docs →Your data stays on our hardware.
Every model — language, vision, speech and detection — runs on a single NVIDIA GX10 in our own facility, behind one HTTPS gateway.
Get one token
Your domain is issued a single bearer token that is valid on every API, with usage and quota reported per service.
Call any API
Everything is served from
https://api.weststar-ai.com— the LLM endpoints are OpenAI-compatible, so existing SDKs work unchanged.Inference runs on-prem
Requests are processed on the GX10’s GB10 GPU. Selected services automatically fall back to CPU when the GPU is busy, so calls still complete.
Results come back — nothing leaves
Responses, reports and generated files are returned to you. No prompts, images or scans are forwarded to third-party AI clouds.
curl https://api.weststar-ai.com/v1/chat/completions \ -H "Authorization: Bearer $WESTSTAR_TOKEN" \ -H "Content-Type: application/json" \ -d '{ "model": "<model>", "messages": [{"role":"user","content":"Hello"}] }' # → served by the GX10, on-prem
Private by design
Data is processed on-prem, not sent to external AI providers.
Standard interfaces
Plain REST + JSON, with an OpenAI-compatible LLM surface.
Metered per domain
Usage and quotas tracked for each caller and each service.
Composable
Crawl → summarize → chart → alert: chain services into workflows.
Build on Weststar AI today.
Open the dashboard, talk to the assistant, or dive straight into the API reference.