n8n-nodes-dzrouter

Use 200+ AI models (GPT, Claude, Gemini, DeepSeek, Llama, Qwen, Mistral and more) in your n8n workflows through a single DzRouter API key, paid with Pulses, DzRouter's AI credit (1 Pulse = 1 DA), bought with Edahabia/CIB, BaridiMob or a card.

DzRouter is an OpenAI-compatible AI router built for Algeria and North Africa. This package gives you:

Every call reports its cost in Pulses (1 Pulse = 1 DA), so you always know what a workflow run costs.

Installation · Credentials · DzRouter node · DzRouter Chat Model · Examples · Compatibility

Installation

Follow the community nodes installation guide in the n8n documentation.

In n8n: Settings → Community Nodes → Install, then enter:

n8n-nodes-dzrouter

Credentials

1. Create an account at dzrouter.com and top up your balance with Pulses (from 500 DA, 1 Pulse = 1 DA).
2. In the DzRouter console, open API keys and create a key. It starts with sk-dz-.
3. In n8n, create a DzRouter API credential and paste the key. Leave Base URL as https://dzrouter.com/v1.

n8n tests the credential with GET /v1/key. The test is free and also works on a new account with a zero balance.

DzRouter node

Chat → Message a Model

Send a conversation to any model.

• Model: search the live catalog by name or provider. Each entry shows its price per 1M tokens, context size and capabilities (vision, tools, reasoning). You can also type a model ID.
• Messages: any number of System, User and Assistant messages, which lets you give the model conversation history.
• Simplify Output (on by default) returns:

{
  "text": "Marhba! How can I help you today?",
  "model": "openai/gpt-4o-mini",
  "finishReason": "stop",
  "usage": { "promptTokens": 21, "completionTokens": 9, "totalTokens": 30 },
  "cost": { "pulses": 0.0024 }
}

• Options: Sampling Temperature, Maximum Number of Tokens, Top P, Frequency and Presence Penalty, Reasoning Effort, Response Format (Text or JSON Object, which is parsed into a json field), Seed, Stop Sequences, Timeout.

Turn Simplify Output off to get the full OpenAI-compatible response plus cost.

Text: ready-made AI tasks

You don't have to write a prompt. Each task runs at temperature 0, so the same input gives consistent results. It tolerates replies that wrap JSON in code fences or add extra text, and adds a dzrouter field with the model, usage and cost.

Extracted values are converted to the declared types. For example, "1 200,50" becomes 1200.5 and "oui" / "نعم" become true. Classification labels are matched to your list regardless of case.

Model

• Get Many: the catalog with prices per 1M tokens in Pulses. Filter by provider, required capabilities (vision, tool calling, reasoning), maximum input/output price, minimum context, or search words. Sort by recommended order, price, context or name.
• Find Cheapest: the cheapest model that matches your filters, compared on input, output or blended price (3 input : 1 output). Pass its id to a later node to always use the cheapest suitable model.
• Get: one model's details and prices.

Account → Get Balance

Returns the balance and total spend in Pulses, rate limit and last use of the key. It's free and never billed. Put it before an expensive step, together with an If node, to stop a workflow when the balance is low.

DzRouter Chat Model

Connect it to the Chat Model input of an AI Agent, a Basic LLM Chain, a Question and Answer Chain, a Summarization Chain, an Information Extractor, a Text Classifier, and so on.

• Pick any catalog model. For AI Agents, choose a model whose description in the list includes tools.
• Options: Sampling Temperature, Maximum Number of Tokens, Top P, Frequency and Presence Penalty, Reasoning Effort, Response Format, Timeout, Max Retries.

Examples

Customer messages in Darja → ticket
Webhook → DzRouter Translate (to French) → DzRouter Classify (order, complaint, question, spam) → Switch on category → create a ticket or reply.

Invoices by email → spreadsheet
Gmail Trigger → DzRouter Extract Data (supplier Text, total Number, dueDate Date, paid Boolean) → Google Sheets Append.

Support agent
Chat Trigger → AI Agent (DzRouter Chat Model + your tools) → reply. Add a DzRouter Get Balance + If step to get an alert before credit runs out.

Cheapest model that fits
DzRouter Find Cheapest (capabilities: Tool Calling, min context 128000) → DzRouter Message a Model with the model set by the expression {{ $json.id }}.

Errors

DzRouter errors are shown with a hint:

With On Error → Continue, a failed item is returned as { "error": "…" } and the other items keep running.

Compatibility

The Chat Model sub-node is built on n8n's AI node SDK, which ships with n8n 2.0 and later. On n8n 1.x the package still loads and the DzRouter node works fully; the DzRouter Chat Model shows a message explaining the alternative instead of running.

AI Agents on n8n 1.x: DzRouter is OpenAI-compatible, so n8n's own OpenAI Chat Model works with it:

1. Create an OpenAI credential: API Key = your DzRouter key (sk-dz-…), Base URL = https://dzrouter.com/v1.
2. Add the OpenAI Chat Model node to your agent or chain, select that credential, and type any DzRouter model id (for example openai/gpt-4o-mini or claude-opus-5-5) in the Model field using By ID.

Other requirements: Node.js 20.15 or newer. No runtime dependencies (@n8n/ai-node-sdk is an optional peer, provided by n8n 2.x).

Development

npm install
npm run build
npm run lint
npm run dev        # starts n8n on http://localhost:5678 with these nodes loaded
npm test           # end-to-end test against a local DzRouter server + mock model provider

npm test needs the DzRouter server repository built next to this one (or DZROUTER_DIR=/path/to/server). It runs both nodes against a real DzRouter instance whose upstream is a local mock, so nothing is billed.

Resources

• DzRouter: models, prices and documentation
• DzRouter API documentation
• n8n community nodes documentation

License

MIT