Open WebUI

Open WebUI can front any number of OpenAI-compatible endpoints at once. Add Grafilab as one connection and every model in the catalog appears in the chat model picker, alongside whatever else you already run.

• An Open WebUI instance you can reach, and an account on it with administrator rights. Only administrators see the connection settings.

• A Grafilab API key. In the console open API (https://app.grafilab.ai/api) and click Generate New Key — see API Keys.

Name the key open-webui. Every chat in the instance authenticates with this one key, so a dedicated key keeps that shared traffic separate from your other tools in the usage view.

1. Select your avatar, open the Admin Panel, then go to Settings → Connections. The page lists your model providers under Manage OpenAI API Connections.

2. Select the + button beside Manage OpenAI API Connections to add a connection.

3. Set the URL to:

   https://llm.grafilab.ai/v1

   The field suggests well-known endpoints as you type — ignore the suggestions and paste the value above. Open WebUI appends /chat/completions and /models itself.

4. Paste your Grafilab key into the API Key field.

5. Select Verify Connection. Open WebUI calls GET /v1/models; a success message means the URL and key are both good.

6. Select Save. The connection is live for every user of the instance.

7. Start a new chat and open the model picker. The Grafilab model ids are in the list.

Send a message on a Grafilab model and watch it stream. Then open API in the console, select View Usage on the open-webui key, and confirm a request landed today. Rollups can lag by up to about 30 minutes.

The catalog is long, and a picker with everything in it is a picker nobody reads.

• Show only the models you want. The connection has a Model IDs (Filter) allowlist. Type a Grafilab model id, select + to add it, and repeat. With at least one entry, the picker shows only those models. Leave it empty to show everything the connection returns.
• Rename and describe models. In Workspace → Models, give a model a friendly display name, a description, a system prompt, and default parameters. This changes only how the model appears in your instance; the request still carries the Grafilab id.
• Generate images. Open WebUI’s image settings can point at an OpenAI-compatible image endpoint, which for Grafilab is https://llm.grafilab.ai/v1 with /v1/images/generations behind it.

Symptom	Cause	Fix
Verify Connection fails	The URL is missing /v1, or has a trailing slash or path suffix	Set it to exactly https://llm.grafilab.ai/v1
Connection verifies but no models appear	A Model IDs (Filter) entry does not match any catalog id	Clear the filter, or re-list with GET /v1/models and retype the id
401 on every chat	The key has stray whitespace, or it was rotated in the console	Paste the key again from API
429 with a quota message	Your Grafilab credit balance is empty — this is not a throughput limit	Top up in the console
Replies arrive all at once instead of streaming	A reverse proxy in front of Open WebUI is buffering server-sent events	Disable response buffering for the Open WebUI route