Add Azure OpenAI

Azure OpenAI is not a normal OpenAI key. The gateway needs the resource URL and the key. Deployment names are not stored. Call each one as the request model.

Save the endpoint#

In the console keys view, pick Azure OpenAI and paste:

• Resource URL: https://{resource}.openai.azure.com/openai/v1
• The Azure API key for that resource

The store is PUT /v1/keys/azure with {"api_key":"...","base_url":"https://{resource}.openai.azure.com/openai/v1"}. No deployments field (retired). A 422 names azure if the row fails; a 404 means no row. The header the gateway sends upstream is api-key.

Same values from the CLI, stored on the gateway (callers keep using CONIFER_API_KEY):

printf %s "$AZURE_KEY" | conifer keys add azure --url https://{resource}.openai.azure.com/openai/v1

To register Azure as a local CLI endpoint instead, so conifer run can hit the resource directly:

printf %s "$AZURE_KEY" | conifer endpoints add byok azure --url https://{resource}.openai.azure.com/openai/v1

The key is verified live against the resource and stored sealed. You can see it (provider, last four, resource URL) and remove it on the same console page. Mint a Conifer API key first at the console if callers will hit api.conifer.build.

Call a deployment#

curl https://api.conifer.build/v1/chat/completions \
  -H "Authorization: Bearer $CONIFER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "YOUR_DEPLOYMENT_NAME",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "three names for a build cache"}]
  }'

Set model to the Azure deployment name, or azure/<deployment>. The gateway maps nothing. Your deployments are not listed in GET /v1/models. A name the resource does not serve returns Azure’s own error.

One collision rule: a deployment named exactly like a catalog model (for example gpt-4o) routes to the catalog, not your resource. Give deployments distinct names.