AI-grounded WhatsApp operations platform — built for teams that want a real assistant on WhatsApp, not just an API gateway.

noxusai.online · Why · AI Stack · API Slash · Quick Start · Docs

Most WhatsApp gateways stop at "send a message." NOXUS AI Open WhatsApp goes further: it ships a grounded AI agent that answers customers from your own knowledge — and can call your live APIs by slash command before replying.

You bring the data and the APIs. NOXUS does the WhatsApp, the AI, and the operations console.

Hosted by us at noxusai.online · self-host with the steps below.

NOXUS ships with a real AI layer for WhatsApp customer support, not just an HTTP wrapper. It is grounded — the assistant is locked to your company-provided knowledge and live API responses.

┌─────────────────┐    1. message      ┌──────────────────┐
│  WhatsApp user  │ ─────────────────► │   NOXUS engine   │
└─────────────────┘                    └────────┬─────────┘
                                                │ 2. extract /slugs from Reply Data
                                                ▼
                              ┌─────────────────────────────────────┐
                              │  Reply Data (your knowledge.md)     │
                              │  "Live menu and prices: /menu"      │
                              │  "Stock check: /stock"              │
                              └────────────┬────────────────────────┘
                                           │ 3. fetch live APIs in parallel
                                           ▼
                              ┌─────────────────────────────────────┐
                              │  /menu  → https://api.you/menu      │
                              │  /stock → https://api.you/stock?... │
                              └────────────┬────────────────────────┘
                                           │ 4. inject into prompt context
                                           ▼
                              ┌─────────────────────────────────────┐
                              │  Gemini (grounded) ─ company data + │
                              │  live API JSON + safety rules       │
                              └────────────┬────────────────────────┘
                                           │ 5. concise WhatsApp reply
                                           ▼
                              ┌─────────────────────────────────────┐
                              │  Sent back to the customer          │
                              └─────────────────────────────────────┘

Static knowledge files lie the moment your prices, stock, or schedule change. Slash commands fix that.

1. Open the dashboard → AI Manager → APIs (slash commands) → Add API

2. Give it a slug (e.g. menu), a name, a description ("Live restaurant menu and prices"), the URL, optional headers, and method.

3. Save and click Test — you'll see HTTP 200, response time, and a 600-char sample.

4. Open Reply Data and write your knowledge naturally, referencing the API with /:

   Our hours are 10 AM – 11 PM, daily.
   
   The live menu and current prices are provided via /menu.
   For real-time stock on a specific item, use /stock.

5. Type / anywhere in Reply Data to open a picker that lists every registered API with its description — pick one to insert it.

At reply time, NOXUS scans your Reply Data for /slug tokens, fetches each enabled API in parallel, and slots the response into the prompt as [API /menu] {…} with the description, so Gemini knows what the data is and grounds its answer in it.

If an API is down, the model is told it failed and falls back to the "the team will follow up" guardrail rather than making something up.

Endpoints: GET /api/ai/apis, POST /api/ai/apis, PUT /api/ai/apis/:id, DELETE /api/ai/apis/:id, POST /api/ai/apis/:id/test

A modern React 19 + Vite + TanStack Query dashboard with full English/Arabic + RTL support. Every module is a real first-class page, not a settings panel.

_{Dashboard — live overview of sessions, messages, webhooks, and AI status}

_{AI Manager — Gemini settings + Reply Data with the / picker for live API slash commands}

_{AI Interactions Log — every question, decision and reason. The assistant mirrors the customer's language automatically (Arabic / RTL shown).}

git clone https://github.com/rmyndharis/OpenWA.git noxus
cd noxus
docker compose -f docker-compose.dev.yml up -d

• Dashboard: http://localhost:2886
• API: http://localhost:2785/api
• Swagger: http://localhost:2785/api/docs

git clone https://github.com/rmyndharis/OpenWA.git noxus
cd noxus
npm install            # also installs the dashboard
npm run dev            # starts API + dashboard, config auto-generated

Open http://localhost:2886 → log in with the API key printed in the console → go to AI Manager to set up Gemini.

1. AI Manager → Settings → paste your Gemini key, enable auto-reply, save.
2. AI Manager → APIs → click Add API, register your first endpoint:
   • Slug: prices
   • Name: Live prices
   • Description: Returns the current price list as JSON.
   • URL: https://your.api/prices
   • Click Test — confirm HTTP 200.
3. AI Manager → Reply Data → write your knowledge in natural language and reference the API:

   We are open daily, 10 AM – 11 PM.
   Our up-to-date prices are provided via /prices.

4. Sessions → start a session, scan the QR with WhatsApp.
5. Message that number from another phone — NOXUS will reply using your knowledge + live /prices response, in the customer's language.

docker compose up -d                       # SQLite, local storage
docker compose --profile postgres up -d    # + PostgreSQL
docker compose --profile full up -d        # PostgreSQL + Redis + Dashboard + Traefik

curl -X POST http://localhost:2785/api/sessions/{sessionId}/messages/send-text \
  -H "Content-Type: application/json" \
  -H "X-API-Key: YOUR_API_KEY" \
  -d '{ "chatId": "628123456789@c.us", "text": "Hello from NOXUS!" }'

curl -X POST http://localhost:2785/api/ai/apis \
  -H "Content-Type: application/json" \
  -H "X-API-Key: YOUR_API_KEY" \
  -d '{
    "slug": "menu",
    "name": "Live menu",
    "description": "Returns the live restaurant menu and prices as JSON.",
    "url": "https://api.yourbiz.com/menu",
    "method": "GET",
    "headers": { "Authorization": "Bearer xyz" },
    "enabled": true
  }'

curl -X PUT http://localhost:2785/api/ai/knowledge \
  -H "Content-Type: application/json" \
  -H "X-API-Key: YOUR_API_KEY" \
  -d '{
    "content": "We open daily 10 AM – 11 PM.\nThe live menu is provided via /menu."
  }'

curl -X POST http://localhost:2785/api/ai/apis/{id}/test \
  -H "X-API-Key: YOUR_API_KEY"

curl -X POST http://localhost:2785/api/sessions/{sessionId}/webhooks \
  -H "Content-Type: application/json" \
  -H "X-API-Key: YOUR_API_KEY" \
  -d '{
    "url": "https://your-server.com/webhook",
    "events": ["message.received", "session.status"],
    "secret": "your-hmac-secret"
  }'

noxus-ai-open-whatsapp/
├── src/
│   ├── main.ts                  # Bootstrap
│   ├── app.module.ts            # Root module
│   ├── config/                  # Configuration loaders
│   ├── common/                  # Cache, storage, shared utilities
│   ├── core/                    # Hooks + plugin system
│   ├── engine/                  # WhatsApp engine abstraction
│   └── modules/
│       ├── ai/                  # Gemini reply + API-link slash commands
│       │   ├── ai-reply.service.ts
│       │   ├── ai-api-links.service.ts
│       │   ├── ai-api-links.controller.ts
│       │   └── ai.controller.ts
│       ├── session/             # Session management
│       ├── message/             # Message handling
│       ├── webhook/             # Webhooks
│       ├── group/               # Groups API
│       ├── contact/             # Contacts
│       ├── auth/                # API keys + roles
│       ├── infra/               # Infrastructure config
│       └── health/              # Health probes
├── dashboard/                   # React dashboard
│   └── src/pages/AiManager.tsx  # AI + APIs slash UI
├── data/
│   ├── ai/
│   │   ├── settings.json        # Runtime AI settings
│   │   ├── api-links.json       # Registered API slash commands
│   │   └── interactions.jsonl   # AI decision audit log
│   └── knowledge/
│       └── noxus.md             # Reply Data (default path)
└── docs/                        # In-repo documentation

1. Fork the repo
2. Branch: git checkout -b feature/your-feature
3. Commit and push
4. Open a PR

Please read the Development Guidelines.

MIT — free for personal and commercial use. See LICENSE.

NOXUS AI Open WhatsApp — AI-grounded WhatsApp, open source.

noxusai.online · Docs · API Docs · Issues