deepseek-vn

DeepSeek AI community node for n8n with Thinking Mode and JSON Output support

Package Information

Released: 3 mo ago
Downloads: 0 weekly / 0 monthly
Latest Version: 1.0.1

Available Nodes

Documentation generating…

DeepSeek VN icon
DeepSeek VN
Consume DeepSeek AI

Documentation

n8n-nodes-deepseek-vn

npm version
License: MIT

A community node for n8n that integrates DeepSeek AI into your workflows.

This project is a maintained fork of rubickecho/n8n-deepseek, extended with Thinking Mode and JSON Output controls. It is published under the MIT license.

Features

  • Chat Completion — Create model responses for chat conversations via POST /chat/completions
  • FIM Completion (Beta) — Fill-In-the-Middle completions via POST /beta/completions
  • Thinking Mode — Toggle chain-of-thought reasoning before the final answer (docs)
  • JSON Output — Force structured JSON responses with a simple boolean toggle (docs)
  • Simplify Output — Return a simplified response instead of raw API data

Installation

Via n8n UI (recommended)

  1. Open your n8n instance

  2. Go to Settings → Community Nodes

  3. Click Install

  4. Enter the npm package name:

    n8n-nodes-deepseek-vn

  5. Accept the risk acknowledgment and click Install

  6. Restart n8n if prompted

Via npm (self-hosted)

Follow the n8n community nodes installation guide.

Credentials

  1. Obtain an API key from the DeepSeek Platform
  2. In n8n, create a new DeepSeek VN API credential
  3. Paste your API key and save

The node connects to https://api.deepseek.com.

Usage

Add a DeepSeek VN node to your workflow, select the resource and operation, then configure the parameters.

Note: This package uses the internal node name deepSeekVn so it can be installed alongside the original n8n-nodes-deepseek package without database conflicts. The node appears as DeepSeek VN in the editor.

Chat Completion

Parameter Description
Model DeepSeek model to use (loaded dynamically from the API)
Thinking Mode When enabled, sends thinking: { type: "enabled" } to activate chain-of-thought reasoning
Prompt System, user, and assistant messages
Simplify Return simplified output instead of raw API response
Options Temperature, max tokens, penalties, JSON Output, and more

JSON Output

When enabling JSON Output in Options:

  1. The API receives response_format: { type: "json_object" }
  2. Your system or user prompt must include the word JSON
  3. Provide an example of the expected JSON format in the prompt
  4. Set a reasonable Maximum Number of Tokens to avoid truncated JSON

Example system prompt:

Parse the input and return JSON with "question" and "answer" fields.

EXAMPLE JSON OUTPUT:
{
  "question": "What is the tallest mountain?",
  "answer": "Mount Everest"
}

Thinking Mode

When Thinking Mode is enabled:

  • The model outputs reasoning_content (chain-of-thought) before the final content
  • Sampling parameters (temperature, top_p, penalties) have no effect in thinking mode
  • See the Thinking Mode guide for multi-turn and tool-call behavior

Local Development

git clone https://github.com/bbrandao/n8n-nodes-deepseek-vn.git
cd n8n-nodes-deepseek-vn
npm ci
npm run build
npm link

Then, in your n8n installation directory:

npm link n8n-nodes-deepseek-vn

Restart n8n to load the node.

Validation

npm run build   # TypeScript compile + copy icons
npm run lint    # ESLint checks

Contributing

See CONTRIBUTING.md for guidelines on reporting issues and submitting pull requests.

Security

To report a vulnerability, see SECURITY.md. Please do not open a public issue for security reports.

Changelog

See CHANGELOG.md for release history.

License

MIT — see LICENSE.md.

This project is derived from rubickecho/n8n-deepseek. The original copyright notice is preserved in the license file.

Resources

Discussion