Connect Open WebUI without guessing the Base URL or model ID.

Use these connection values

What is verified—and what is not

The official Open WebUI compatible-provider guide says basic chat requires POST /v1/chat/completions and recommends GET /v1/models for model discovery. It also says connection verification calls /models with a Bearer token.

Run two preflight checks outside Open WebUI

First confirm that the key and Base URL return a model list. Do this in a trusted server-side shell and never print the key itself:

test -n "$PRIMORDIAL_API_KEY" || {
  echo "PRIMORDIAL_API_KEY is not set" >&2
  exit 1
}

curl --fail-with-body --silent --show-error \
  https://www.primoraihub.com/v1/models \
  -H "Authorization: Bearer $PRIMORDIAL_API_KEY" \
  | jq -r '.data[]?.id'

Copy one exact ID from that output into the next test. Do not invent a model name or use an old ID from a blog post:

MODEL_ID="MODEL_ID_FROM_V1_MODELS"

curl --fail-with-body --silent --show-error \
  https://www.primoraihub.com/v1/chat/completions \
  -H "Authorization: Bearer $PRIMORDIAL_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\": \"$MODEL_ID\",
    \"messages\": [{\"role\": \"user\", \"content\": \"Reply with OK\"}],
    \"stream\": false
  }"

The second request can consume account quota. A successful result verifies one model, one endpoint, and one non-streaming request shape only.

Configure the provider in Open WebUI

1. Sign in to Open WebUI with an administrator account.
2. Open Admin Settings → Connections → OpenAI and select Add Connection.
3. Keep the provider type as OpenAI and enter https://www.primoraihub.com/v1 as the URL.
4. Paste the API key into the API Key field. Do not include Bearer , quotes, or a trailing newline.
5. Leave Model IDs empty for the first verification. Save the connection.
6. If discovery succeeds, add only the models intended for this Open WebUI instance to Model IDs (Filter).

Open WebUI's menu labels can change between releases. If the current interface differs, follow its official connection guide and keep the same provider URL and credential boundary.

Test the smallest useful path

1. Choose one model that appears under the new provider connection.
2. Start a new chat with tools, web search, knowledge, images, and other optional features disabled.
3. Send a short prompt and confirm that a complete non-streaming or basic streaming response appears.
4. Review Primordial AI usage and Open WebUI server logs for the same attempt without copying credentials into logs.
5. Only after plain chat works, enable one optional feature at a time and retest.

If the model picker is empty, compare the Open WebUI server's model request with the direct model discovery test. A browser loading the Open WebUI page does not prove that its backend container can reach the provider.

Verify optional features independently

Open WebUI's official guide lists separate optional routes: /v1/embeddings for RAG, /v1/images/generations for images, and audio routes for speech and transcription. Tool calling and streaming also depend on the selected model and response schema.

• Do not enable RAG until an embeddings request succeeds with an intended embedding model.
• Do not infer image support from a text chat result; use the documented image endpoint and payload.
• Do not infer tool support from a plain response; test one minimal tool schema and inspect the returned structure.
• Keep a fallback connection or disable a feature when its exact route is unavailable.

Troubleshoot by symptom

Protect a shared Open WebUI connection

A provider key configured by an Open WebUI administrator can spend the associated provider account's quota for allowed users. Protect the admin panel, require HTTPS, constrain the key by workload where possible, set quota or expiry controls, restrict the visible model allowlist, and monitor usage.

The official Open WebUI guide recommends avoiding management or master keys for multi-user connections and preferring least-privilege credentials. This is Open WebUI security guidance, not an endorsement of Primordial AI.

Frequently asked questions