Skip to main content

Documentation Index

Fetch the complete documentation index at: https://gomodel-docs-providers-restructure.mintlify.app/llms.txt

Use this file to discover all available pages before exploring further.

GoModel routes Azure OpenAI requests through the deployment URL you provision in the Azure portal. Azure speaks the OpenAI API shape with two differences GoModel handles transparently: an api-key header instead of Authorization: Bearer, and a required api-version query parameter on every request. Flow: Client -> GoModel -> Azure OpenAI

​
Before you start

  • An Azure OpenAI resource in your subscription.
  • At least one deployment created for a model (e.g. gpt-5, gpt-4o, text-embedding-3-small).
  • The deployment’s base URL and resource API key from the Azure portal.

​
1. Set the deployment URL and API key

export AZURE_API_KEY="..."
export AZURE_BASE_URL="https://your-resource.openai.azure.com/openai/deployments/your-deployment"
export AZURE_API_VERSION="2024-10-21"
AZURE_API_VERSION is optional. GoModel defaults to 2024-10-21 when unset. Set it explicitly when you need a different stable or preview version (for example, a newer Responses API surface). Or in config.yaml: 
providers:
  azure:
    type: azure
    base_url: "${AZURE_BASE_URL}"
    api_key: "${AZURE_API_KEY}"
    api_version: "2024-10-21"

​
2. (Optional) Multiple deployments

Each Azure deployment is its own model endpoint. Register them as separate providers using suffixed env vars: 
export AZURE_GPT5_API_KEY="..."
export AZURE_GPT5_BASE_URL="https://your-resource.openai.azure.com/openai/deployments/gpt-5"

export AZURE_EMBED_API_KEY="..."
export AZURE_EMBED_BASE_URL="https://your-resource.openai.azure.com/openai/deployments/text-embedding-3-small"
These register providers azure-gpt5 and azure-embed. When you want all deployments under one Azure resource to share an API key and expose every deployment as a model, point AZURE_BASE_URL at the resource root (without the /deployments/... segment) and use aliases or AZURE_MODELS to advertise specific deployment names.

​
3. Start GoModel

go run ./cmd/gomodel

​
4. Verify the model registry

curl -s http://localhost:8080/v1/models
Expected result:
  • a 200 OK
  • the deployments the Azure resource exposes via /openai/models, with owned_by: "azure".

​
5. Verify Chat Completions

curl -s http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5",
    "messages": [{"role": "user", "content": "Reply with the single word ok."}],
    "max_tokens": 80
  }'
Expected result:
  • a 200 OK
  • assistant content containing ok
Streaming, the Responses API, embeddings, and batch operations are wired up; batch results route through the /openai/batches resource path under your Azure resource.

​
Notes on URLs

GoModel accepts the deployment-scoped URL most users copy from the portal: 
https://<resource>.openai.azure.com/openai/deployments/<deployment>
For account-wide operations (model listing, batches) GoModel automatically strips the /openai/deployments/<deployment> suffix and talks to the resource root, so a single AZURE_BASE_URL works for both chat and resource-level calls.

​
Troubleshooting

  • 401 Unauthorized / 403 Check AZURE_API_KEY matches the resource that hosts the deployment, and that the deployment name in the URL is correct.
  • Resource not found or DeploymentNotFound Azure returns 404 when the deployment name in the URL does not match. The model name in the request body must be the deployment name, not the underlying base model.
  • unsupported api version or invalid api-version Bump AZURE_API_VERSION to a version the feature you’re calling supports. Newer Responses API features require recent preview API versions.

​
References