Tool configuration
Configure SeaLink as an OpenAI-API-compatible chat model provider in Dify for assistants and workflows. Configure embedding and rerank models separately under Dify's matching model types.
Open your Dify console (cloud.dify.ai or your self-hosted URL) and log in with your account.
Click your avatar (top-right) → 'Settings' → 'Model Provider' in the left sidebar. You can also click 'Settings' directly from the Dify home page.
Find 'OpenAI-API-compatible' in the provider list and click 'Add Provider'. If you don't see it, look for 'Custom' or 'Add Custom Provider' — the exact label varies by Dify version, which is normal.
Provider Name: "SeaLink" (or anything you'd like) API Base URL: https://api.sealink.io/v1 API Key: Paste the SeaLink API Key (starts with sk-sealink-) You can leave the other fields empty. Click Save.
After saving, Dify may automatically fetch the model list; some versions require entering model IDs manually. We recommend testing with claude-opus-5-5 first. If no models appear, double-check that your Base URL ends with /v1 — this is the most common oversight.
Create a new app or edit an existing one → choose your SeaLink connection in 'Model Settings' → select or enter a specific chat model ID. One thing to keep in mind: if you're using RAG in your workflow, embedding and rerank models need their own SeaLink provider entries under Dify's Embedding and Rerank model types — they can't share the chat model connection directly.
No models appear after adding the provider
First, check that the Base URL ends with /v1 — this is the most common issue. If you're using self-hosted Dify, verify the server can reach sealink.io by running: curl https://sealink.io/health. If the network is fine but models still don't appear, confirm the API Key is active and valid.
Dify shows 'Connection failed'
This is usually caused by network access, Base URL, or API Key issues — any one of these can trigger the error. We recommend checking in this order: first, confirm the machine running Dify can reach sealink.io; second, verify the Base URL is complete (ending with /v1); third, check the API Key is still active and hasn't been revoked. Most cases are resolved at step two or three.
Model list seems incomplete
Dify may filter models by type. Chat models belong under the OpenAI-API-compatible chat provider; embedding and rerank models need their own SeaLink provider entries under Dify's Embedding and Rerank model types and won't show up in the chat model list.
Model call fails in workflow but works in chatbot
Dify version differences can cause incompatibility with certain advanced parameters, leading to failures in workflows even when the same model works in chat. We suggest testing workflow connectivity with a general-purpose chat model like claude-opus-5-5 first. Once the basic call path is confirmed, you can adjust workflow parameters or switch models as needed.
