How to Combine Azure OpenAI with Elastic Observability AI Assistant in 10 Minutes

Prerequisites

To use Elastic Observability’s AI Assistant you need Elastic Stack version 8.9 or higher, an Elastic Enterprise subscription, an Azure OpenAI account with a GPT‑4 model (e.g., gpt‑4‑32k 0613), and, for private‑knowledge‑base use, a machine‑learning node with at least 4 GB of RAM.

Creating the Azure OpenAI Service

In the Azure portal create an OpenAI resource (e.g., ai4elasticstack) in a region that supports GPT‑4, such as Central Switzerland. Deploy the desired model (gpt‑4‑32k) and set the token limit to the maximum to ensure the AI Assistant can operate without throttling.

Configuring the Elastic AI Assistant Connector

In Elastic Cloud, add a new connector of type Azure OpenAI and fill in the service details. The essential parameter is the endpoint URL, which must follow this pattern:

https://{your-resource-name}.openai.azure.com/openai/deployments/{deployment-id}/completions?api-version={api-version}

Replace {your-resource-name} with your Azure OpenAI resource name, {deployment-id} with the deployment identifier (e.g., gpt4-32k0613), and {api-version} with a supported version such as 2023-07-01-preview. An example URL looks like:

https://ai4elasticstack.openai.azure.com/openai/deployments/gpt4-32k0613/chat/completions?api-version=2023-07-01-preview

After entering the URL and API key, click the “Test” button; a green success message confirms a proper connection.

Using Contextual Insights

Within the Observability app you can click the AI Assistant icon in various modules (Logs, APM, Infrastructure, Alerts, etc.) to obtain contextual explanations, suggested KQL queries, and remediation advice. For example, selecting a failing Docker‑related log entry triggers the assistant to explain the error and propose possible causes such as a stopped Docker daemon or mis‑configured Metricbeat.

Chat‑Based Interaction

The chat mode lets you ask open‑ended questions based on observability data, such as “Will my server run out of disk space?” The assistant responds by analyzing relevant metrics, generating visual summaries, and offering step‑by‑step reasoning. It can also invoke built‑in functions like summarize, recall, lens, elasticsearch, kibana, and various APM‑related functions to fetch data or create visualizations.

Benefits

Accelerates root‑cause analysis by providing AI‑generated explanations of logs and alerts.

Reduces manual KQL writing through automatically suggested queries.

Enables multilingual queries; the assistant understands both English and Chinese.

Integrates seamlessly across multiple Observability modules, offering a unified AI‑driven troubleshooting experience.

Summary

By following the steps above you can configure Azure OpenAI as a backend for Elastic’s AI Assistant, enabling both contextual insights and conversational troubleshooting across the Elastic Observability suite. The entire setup can be completed in roughly ten minutes, providing immediate AI‑assisted operational intelligence.