Skip to main content
Open WebUI’s built-in OpenAI-compatible Chat Completions connection can use Ace Data Cloud without a custom Pipe. This quickstart covers an existing instance; use the full setup guide for deployment, persistence, and troubleshooting. Its connection steps were checked in v0.11.4; v0.12.0 model discovery and ordinary-user Read permissions were also checked separately.

Before you start

  • Get an application API key from the Ace Data Cloud console. Check the selected model’s access, current price, and balance. A restricted key needs both /v1/models and /v1/chat/completions.
  • Use an Open WebUI administrator account to add the server-side connection and grant model access. Authorized users share this connection’s key and can consume its Credits. Do not paste a key into chat messages or screenshots.

Add the connection

  1. Open Open WebUI. In the user menu, choose Admin Panel → Settings → Connections.
  2. Under OpenAI API → Manage OpenAI API Connections, click Add Connection (+).
  3. Leave Connection Type at External. Set URL to https://api.acedata.cloud/v1 (no trailing slash), Auth to Bearer, and API Key to your application key. Leave API Type at Chat Completions.
  4. Click Verify Connection. It checks the /v1/models endpoint; the button is optional when you save. Click Save in the connection dialog, then Save in the main Connections settings if prompted.
  5. In Admin Panel → Settings → Models, grant the intended users or group Read access to each model. Then, as an authorized user, start a new chat and select a current Chat Completions model. Send Reply with the single word OK. Check the completed reply and the corresponding usage record.
Model IDs and availability change; use the returned list and the current chat model catalog. To show a smaller list, add exact IDs under Advanced → Model IDs using Add (+) and save. This allowlist controls discovery; it does not replace each model’s Read permission. Seeing an ID does not prove support for images, tools, or other protocols.

Check the request and solve problems

The URL already contains /v1; Open WebUI appends /models or /chat/completions. Do not enter the full chat endpoint or /openai/v1 as the connection URL. Verify Connection checks discovery; only a completed chat proves the chat route. If an ordinary user sees no models, check Read grants separately from application access, key validity, and balance. Disable any unused default connection to avoid selecting or probing it. The Ace key stays in the server-side connection settings; restrict server and model access. A personal-key model Pipe or service Tool has its own user settings and permissions; the shared connection does not populate those keys. Each model request can consume Credits. Title, tag, and follow-up generation can send additional requests; disable these features for a controlled first check. Match charges in usage history by key and time window. An accepted media task is not a finished result; this text-chat quickstart does not validate media, files, Community tool publication, or every model. For connection rules, see Open WebUI’s official compatibility guide. Continue with the full setup guide or Chat Completions schema.

Install a separate service Tool

Suno, Seedance, Happy Horse, GPT Image and Nano Banana each have an independently downloadable Tool in the Ace Data Cloud v0.1.1 release, with English/Chinese walkthroughs. This is an Ace GitHub distribution; verified Open WebUI Community post/Get URLs are still pending. Do not treat a repository download as an official Community listing or default-installed provider. Choose only the needed service’s JSON or ZIP, review the Python and checksum, then ask an administrator to import it into Workspace → Tools and grant Read. Each user fills that Tool’s own User Valves → Api Key and keeps Allow Billable Calls off except for an intended paid submit. Model access and tool access are separate. Async generation returns a task ID first; retrieve the same ID until terminal success and open the returned media, instead of generating again to poll. See the service walkthrough for exact Key permissions, current-price checks, actions and output limits.