XylemNode

Integration guides

Dify

An open-source platform for building LLM apps with visual orchestration of agents, workflows and RAG. Through its OpenAI-API-compatible provider it connects any OpenAI-compatible service — enter the address and key to use XylemNode as a model source.

Get API Key
Site / download
  1. 01

    Prepare Dify

    Sign up for Dify Cloud (cloud.dify.ai) to start right away; or self-host — clone the official repo, bring up the services with Docker Compose and open the local address in your browser to log in.

  2. 02

    Install the OpenAI-API-compatible provider

    Go to Settings → Model Providers in the top-right and find and install OpenAI-API-compatible from the list or plugin marketplace. It's made for connecting all kinds of OpenAI-compatible endpoints.

  3. 03

    Add the XylemNode model

    On that provider's card, click 'Add model' and fill in the table below. The API endpoint URL needs /v1, and the model name must match the console.

    Model type
    LLM
    Model name
    gpt-5.6-sol
    API Key
    sk-your-key
    API endpoint URL
    https://api.xylemnode.com/v1
  4. 04

    Save and use

    After saving, you can set it as the default reasoning model under 'System Model Settings' in the top-right; then you can select the XylemNode model when orchestrating apps.

Two notes

1) For a knowledge base / RAG, add an Embedding model under the same provider (model type Text Embedding) and set it as the default in the knowledge base. 2) Some newer Dify versions can also use the built-in OpenAI provider directly by changing the base URL to XylemNode's /v1 address — same result; pick whichever fits your version.

Common questions

"OpenAI-API-compatible" is missing from Dify's model provider list

From Dify 1.x model providers are plugins, so the matching plugin has to be installed from the Marketplace before it shows up in the list — an empty list before that step is expected, not a sign you installed the wrong build. Once it is there, configure it as a customizable model with your model name and base URL. One known Dify-side incompatibility to be aware of: for newer models that require max_completion_tokens, Dify's OpenAI-compatible implementation still sends max_tokens, so upstream can reject the request over that parameter. That is Dify's own open issue — when you hit a parameter error of that kind, start by not setting a max token limit in Dify.