Skip to main content

ChindaLLM 4B with n8n

At the end you will have an n8n workflow that sends a Thai prompt to ChindaLLM 4B and passes the answer to the next node.

You need Node.js or Docker for n8n, and either Ollama to run the model yourself or a ChindaX account for the hosted API.

1. Start n8n​

npx n8n

Open http://localhost:5678, create a new workflow and add a Manual Trigger node.

2. Get the model​

curl -fsSL https://ollama.com/install.sh | sh
ollama pull iapp/chinda-qwen3-4b
ollama serve

The API listens on http://localhost:11434.

3. Add an HTTP Request node​

Add an HTTP Request node after the trigger, set Method to POST, turn on Send Body and set Specify Body to Using JSON, then fill in the fields for your model.

URL: http://localhost:11434/api/generate

Body:

{
  "model": "iapp/chinda-qwen3-4b",
  "prompt": "สวัสดีครับ ช่วยอธิบายเกี่ยวกับปัญญาประดิษฐ์ให้ฟังหน่อย",
  "stream": false
}

An n8n HTTP Request node calling ChindaX, with the answer in the Output panel

Click Execute step; the answer appears in the Output panel.

4. Pass the answer on​

Map {{ $json.response }} (Ollama) or {{ $json.choices[0].message.content }} (ChindaX) into the next node, such as Set or Gmail.

A workflow of Manual Trigger, HTTP Request to ChindaX, then Gmail

Troubleshooting
  • Error: connect ECONNREFUSED 127.0.0.1:11434: Ollama is not running; start it with ollama serve. If n8n runs in Docker, localhost is the container rather than your machine, so start n8n with npx or npm, or use ChindaX.
  • The workflow times out: in the HTTP Request node, open Options, add Timeout and set it to 30000 (30 seconds) or more.
  • The answer starts with <think>: that is the model's reasoning. Remove everything up to and including </think> in a Set or Code node before you send it on.

The model's benchmarks and its Apache 2.0 licence are on the ChindaLLM 4B page; n8n is documented at docs.n8n.io.