🤖 Configure AI in X-AutoMate (Azure OpenAI)

To enable AI capabilities in X-AutoMate — the assistant, SQL generation, the query translator, and MCP — you need an LLM backend. This guide walks through deploying an Azure OpenAI model, retrieving the credentials, and wiring it into X-AutoMate.

💡 You only need to do this once per environment. After it's configured, all AI features become available across the app.

✅ Prerequisites

An active Azure subscription
Access to Azure OpenAI Service (request access at https://aka.ms/oai/access if your subscription doesn't have it yet)
Owner or Contributor role on the target resource group
(Optional) Azure CLI installed locally — brew install azure-cli

🏗 Step 1: Create an Azure OpenAI Resource

🅰 Option A — Azure Portal

Sign in to https://portal.azure.com.
Search for Azure OpenAI in the top search bar and select it.
Click Create.
Fill in the Basics tab:
- Subscription — your subscription
- Resource group — create a new one or pick existing (e.g. rg-openai-prod)
- Region — choose a region that supports your model (e.g. East US, Sweden Central, West Europe). Model availability varies by region — see model availability docs
- Name — globally-unique name (e.g. xam-openai-prod)
- Pricing tier — Standard S0
On Network, choose access type (public endpoint for quickstart; private endpoint for production).
Add Tags if needed, then Review + create → Create.
Wait for deployment to finish, then click Go to resource.

🅱 Option B — Azure CLI

# Variables
RG="rg-openai-prod"
LOCATION="eastus"
NAME="xam-openai-prod"

# Create resource group (skip if it already exists)
az group create --name "$RG" --location "$LOCATION"

# Create the Azure OpenAI account
az cognitiveservices account create \
  --name "$NAME" \
  --resource-group "$RG" \
  --location "$LOCATION" \
  --kind OpenAI \
  --sku S0 \
  --yes

🧬 Step 2: Deploy a Model

A deployment is a named instance of a base model (e.g. gpt-5.4-nano) that your application targets.

🅰 Option A — Azure AI Foundry (Portal)

From the resource page, click Go to Azure AI Foundry portal (formerly Azure OpenAI Studio).
In the left nav, select Deployments → + Deploy model → Deploy base model.
Choose a model from the GPT-5 family, for example:
- ⭐ gpt-5.4-nano — preferred for X-AutoMate: fastest and most cost-efficient, ideal for the assistant, SQL generation, and the query translator
- gpt-5.4-mini — balanced speed and quality
- gpt-5.4 — highest-quality reasoning, best for complex SQL generation tasks
- text-embedding-3-large — embeddings
Click Confirm, then configure:
- Deployment name — friendly identifier you'll use in code (e.g. gpt-5.4-nano-chat)
- Model version — pin a specific version or use Auto-update to default
- Deployment type:
  - Standard — pay-per-token, regional
  - Global Standard — pay-per-token, global capacity (highest throughput)
  - Provisioned (PTU) — reserved throughput for predictable latency
- Tokens per Minute Rate Limit (TPM) — quota for this deployment
- Content filter — default or a custom policy
Click Deploy.

🅱 Option B — Azure CLI

DEPLOYMENT_NAME="gpt-5.4-nano-chat"
MODEL_NAME="gpt-5.4-nano"
MODEL_VERSION="2026-04-15"

az cognitiveservices account deployment create \
  --name "$NAME" \
  --resource-group "$RG" \
  --deployment-name "$DEPLOYMENT_NAME" \
  --model-name "$MODEL_NAME" \
  --model-version "$MODEL_VERSION" \
  --model-format OpenAI \
  --sku-capacity 50 \
  --sku-name "Standard"

🔑 Step 3: Retrieve Endpoint and Keys

From the Portal

Open the Azure OpenAI resource.
In the left nav, select Keys and Endpoint.
Copy:
- Endpoint — e.g. https://xam-openai-prod.openai.azure.com/
- KEY 1 (or KEY 2)

From the CLI

# Endpoint
az cognitiveservices account show \
  --name "$NAME" --resource-group "$RG" \
  --query "properties.endpoint" -o tsv

# Keys
az cognitiveservices account keys list \
  --name "$NAME" --resource-group "$RG"

🧪 Step 4: Test the Deployment

Before wiring into X-AutoMate, give it a quick smoke test.

REST (curl)

ENDPOINT="https://xam-openai-prod.openai.azure.com"
DEPLOYMENT="gpt-5.4-nano-chat"
API_VERSION="2024-10-21"
API_KEY="<your-key>"

curl -X POST "$ENDPOINT/openai/deployments/$DEPLOYMENT/chat/completions?api-version=$API_VERSION" \
  -H "Content-Type: application/json" \
  -H "api-key: $API_KEY" \
  -d '{
        "messages": [
          {"role": "system", "content": "You are a helpful assistant."},
          {"role": "user",   "content": "Say hello from Azure OpenAI."}
        ],
        "max_tokens": 100
      }'

Python (openai SDK)

pip install openai

import os
from openai import AzureOpenAI

client = AzureOpenAI(
    api_key=os.environ["AZURE_OPENAI_API_KEY"],
    azure_endpoint=os.environ["AZURE_OPENAI_ENDPOINT"],
    api_version="2024-10-21",
)

resp = client.chat.completions.create(
    model="gpt-5.4-nano-chat",  # <-- the DEPLOYMENT name, not the model name
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user",   "content": "Say hello from Azure OpenAI."},
    ],
)
print(resp.choices[0].message.content)

🔗 Step 5: Wire It into X-AutoMate

Once your Azure OpenAI deployment is ready, add the following environment variables to your X-AutoMate configuration.

✅ Required Variables

XAM_LLM_ENABLED=True
XAM_LLM_PROVIDER=azure
AZURE_INFERENCE_ENDPOINT="https://xam-openai-prod.openai.azure.com"
AZURE_INFERENCE_CREDENTIAL="<your-azure-openai-key>"
AZURE_API_VERSION="2024-10-21"
AZURE_CHAT_MODEL="gpt-5.4-nano-chat"  # the DEPLOYMENT name, not the base model

⚙️ Optional Variables

MAX_TOKENS=4096
STREAMING=True   # True | False

💡 Tip

AZURE_CHAT_MODEL should be your deployment name from Step 2 (e.g. gpt-5.4-nano-chat) — not the base model name (e.g. gpt-5.4-nano). This is the most common configuration mistake.

🔐 Step 6 (Recommended): Use Managed Identity Instead of Keys

API keys are convenient but harder to rotate and audit. For services running in Azure (App Service, AKS, Container Apps, VM), prefer Entra ID auth:

Enable a system-assigned managed identity on the compute resource.
On the OpenAI resource → Access control (IAM) → Add role assignment → grant the identity the role Cognitive Services OpenAI User.
In code, drop the API key and use DefaultAzureCredential:

from azure.identity import DefaultAzureCredential, get_bearer_token_provider
from openai import AzureOpenAI

token_provider = get_bearer_token_provider(
    DefaultAzureCredential(),
    "https://cognitiveservices.azure.com/.default",
)

client = AzureOpenAI(
    azure_endpoint=os.environ["AZURE_OPENAI_ENDPOINT"],
    azure_ad_token_provider=token_provider,
    api_version="2024-10-21",
)

📊 Step 7: Monitoring and Quota

Metrics — Resource → Monitoring → Metrics. Track Processed Inference Tokens, Generated Completion Tokens, throttled requests.
Logs — enable Diagnostic settings to send logs to Log Analytics.
Quota — AI Foundry → Quotas. Increase TPM per deployment or request a subscription-level quota increase via support ticket.
Cost — Cost Management + Billing → filter by the resource group. Pricing is per 1K input/output tokens and varies by model.

🚑 Troubleshooting

Symptom	Likely cause	Fix
`404 DeploymentNotFound`	`AZURE_CHAT_MODEL` is the base model name instead of the deployment name	Use your deployment name (e.g. `gpt-5.4-nano-chat`)
`401 Unauthorized`	Wrong key or wrong endpoint region	Re-check Keys and Endpoint blade
`429 Too Many Requests`	TPM / RPM exceeded	Raise deployment capacity or switch to Global Standard
`Region does not support model`	Selected region lacks that model SKU	Pick a supported region from the model availability matrix
Slow first response	Cold start on Standard tier	Use PTU / Provisioned deployment for steady low latency
AI features missing in X-AutoMate UI	`XAM_LLM_ENABLED` is `False` or unset	Set `XAM_LLM_ENABLED=True` and restart the app

📚 Useful References

📌 Summary

Step	What you do
1	Create an Azure OpenAI resource
2	Deploy a model from the GPT-5 family (preferred: `gpt-5.4-nano`)
3	Grab the endpoint and key
4	Smoke-test the deployment
5	Add `XAM_LLM_` and `AZURE_` env vars to X-AutoMate
6	(Recommended) switch to Managed Identity for production
7	Monitor TPM, quota, and cost

🚀 Next step: Once configured, head to the AI Assistant, MCP Integration, AI SQL Generation, or AI Query Translator pages to start using AI in X-AutoMate.

✅ Prerequisites​

🏗 Step 1: Create an Azure OpenAI Resource​

🅰 Option A — Azure Portal​

🅱 Option B — Azure CLI​

🧬 Step 2: Deploy a Model​

🅰 Option A — Azure AI Foundry (Portal)​

🅱 Option B — Azure CLI​

🔑 Step 3: Retrieve Endpoint and Keys​

From the Portal​

From the CLI​

🧪 Step 4: Test the Deployment​

REST (curl)​

Python (openai SDK)​

🔗 Step 5: Wire It into X-AutoMate​

✅ Required Variables​

⚙️ Optional Variables​

🔐 Step 6 (Recommended): Use Managed Identity Instead of Keys​

📊 Step 7: Monitoring and Quota​

🚑 Troubleshooting​

📚 Useful References​

📌 Summary​

✅ Prerequisites

🏗 Step 1: Create an Azure OpenAI Resource

🅰 Option A — Azure Portal

🅱 Option B — Azure CLI

🧬 Step 2: Deploy a Model

🅰 Option A — Azure AI Foundry (Portal)

🅱 Option B — Azure CLI

🔑 Step 3: Retrieve Endpoint and Keys

From the Portal

From the CLI

🧪 Step 4: Test the Deployment

REST (curl)

Python (openai SDK)

🔗 Step 5: Wire It into X-AutoMate

✅ Required Variables

⚙️ Optional Variables

🔐 Step 6 (Recommended): Use Managed Identity Instead of Keys

📊 Step 7: Monitoring and Quota

🚑 Troubleshooting

📚 Useful References

📌 Summary