🧠 Onyx AI Platform Integration

Deploy Onyx (open-source AI knowledge assistant) on DCDN Cloud with one click, powered by our OpenAI-compatible inference backend.

What is Onyx?

Onyx is an open-source AI platform that connects to your company's documents (Slack, Google Drive, Confluence, etc.) and provides an intelligent search and chat interface. Think of it as a private ChatGPT that knows your organization's data.

• 🔍 Semantic search across all connected sources
• 💬 AI chat with document-grounded answers
• 🔗 50+ connectors (Slack, GitHub, Notion, Google Drive, etc.)
• 🔒 Self-hosted — your data stays on your infrastructure

Three Integration Layers

1. OpenAI-Compatible LLM Backend

DCDN Cloud provides a fully OpenAI-compatible API that Onyx can use as its LLM provider. No OpenAI API key needed.

Setting	Value
API Base URL	https://dcdncloud.com/v1
Model (Smart)	dcdn-smart
Model (Fast)	dcdn-fast
Model (Reasoning)	dcdn-reasoning
API Key	Your DCDN Cloud API token

In Onyx's environment config:

GEN_AI_API_BASE=https://dcdncloud.com/v1
GEN_AI_MODEL_VERSION=dcdn-smart
GEN_AI_API_KEY=your_dcdn_token
FAST_GEN_AI_MODEL_VERSION=dcdn-fast

2. One-Click Onyx Deploy Template

Deploy a full Onyx stack on a DCDN VPS from the Compute dashboard:

1. Go to Dashboard → Compute
2. Click + New Instance
3. Select the Onyx AI Platform template
4. Choose a plan (recommended: s-large or higher — 4 GB RAM minimum)
5. Click Create

The template deploys via Docker Compose:

• Onyx Web — Frontend + API server (port 3000)
• Onyx Background — Document indexing workers
• PostgreSQL — Metadata storage
• Vespa — Search index engine
• Redis — Caching layer

3. Browser API + MCP Connector

DCDN Cloud's Browser API can be used as an MCP (Model Context Protocol) tool server, giving Onyx agents the ability to browse the web.

Feature	Details
MCP Server	dcdncloud.com:7379
Protocol	MCP over HTTP (JSON-RPC)
Capabilities	Navigate, screenshot, click, type, extract text
Nodes	7 nodes with headless Chrome
Regions	EU, US, SA — multi-region browsing

Available MCP tools:

• browser_navigate — Open a URL
• browser_screenshot — Capture page screenshot
• browser_click — Click an element
• browser_type — Type text into an input
• browser_extract — Extract page text/data
• browser_evaluate — Run JavaScript on the page

Recommended Plans

Use Case	Plan	Price
Small team (1-10 users)	s-large (2 vCPU, 4 GB)	$20/mo
Medium team (10-50 users)	s-xl (4 vCPU, 8 GB)	$40/mo
Large deployment	d-medium (4 vCPU, 8 GB dedicated)	$48/mo

Supported Models

All 21 DCDN Cloud models work with Onyx. Recommended:

• dcdn-smart — Best quality for document Q&A (Claude Sonnet 4.6)
• dcdn-fast — Lower latency for real-time chat
• dcdn-reasoning — Complex analysis and multi-step reasoning
• Any OpenRouter model ID also works as passthrough

Self-Hosting vs. DCDN Template

	Self-Host	DCDN Template
Setup time	1-2 hours	~2 minutes
LLM provider	You configure	Pre-configured (DCDN)
Infrastructure	Your servers	DCDN VPS (managed)
Updates	Manual	Docker pull + restart
Browser tools	Not included	MCP connector ready

Related Docs

• GPU Marketplace (Models & Inference)
• Cloud VPS Guide
• GPU API Reference
• Agent Platform