Cherry Studio, Obsidian Copilot, Immersive Translate and n8n

These tools are set up in their own settings screens, with nothing to paste into a file. Each takes a base URL, a key and model ids.

Cherry Studio

Settings → Model Services → + Add → add a custom provider of type OpenAI.

• API Host: https://api.ai.ml. Cherry Studio adds /v1 itself.
• API Key: your ai.ml key.
• Models: Manage reads GET /v1/models; add the ones you want with +.
• Turn the provider on with the switch at the top right. A new provider starts switched off.

Obsidian Copilot

Settings → Copilot → BYOK → Add a provider → Add a custom provider.

• Base URL: https://api.ai.ml/v1
• API key: your ai.ml key
• Models: found from GET /v1/models, or typed as ids.

If Test passes but chat fails, turn on Enable CORS. Replies then arrive whole rather than streamed.

Immersive Translate

Translation services → Add custom AI translation service compatible with the OpenAI interface.

• API endpoint: the full chat URL, https://api.ai.ml/v1/chat/completions
• API key: your ai.ml key
• Model: any id, such as qwen/qwen-mt-flash, the translation model, where your catalog serves it.

Translating a page sends many requests at once. If you get 429s, lower the requests-per-second setting.

n8n

Create an OpenAI credential with your ai.ml key and Base URL https://api.ai.ml/v1, then pick it in the OpenAI Chat Model node. Models come from GET /v1/models, or can be entered by ID.

• The node uses the Responses API by default (Use Responses API, node version 1.3 and later). ai.ml serves it, but keeps no conversation state. Leave Conversation ID empty, because ai.ml refuses it, and turn off Use Responses API for models that are not OpenAI's.
• n8n's built-in Web Search, File Search and Code Interpreter tools run only where the provider runs them.

Each section was checked against that tool's documentation on 2026-10-02; none was run here.