Azure OpenAI
- Artificial Intelligence
- Microsoft Azure, WOZTELL
Train It Your Way
Web Search + Live Actions
Enable Web Search or connect to external systems via Function Calling — your assistant can look things up and take action without leaving the conversation.
Voice Message Ready
Audio messages are automatically transcribed before reaching the assistant. Your AI handles voice and text the same way.
What is this extension for?
Azure GPT Assistant v1.1.0 connects WOZTELL to Azure OpenAI — Microsoft’s enterprise-grade deployment of OpenAI models, with the data residency, compliance, and access controls that come with running on Azure.
This extension gives you access to the OpenAI models you deploy in your own Azure environment, plus three integrated tools — Web Search, File Search (Knowledge Base), and Function Calling — that go beyond answering questions. Your assistant can search the web, retrieve answers from your own uploaded documents, and trigger real actions in external systems, all within a customer conversation.
Key Features of the Extension
- Custom Prompt Training: Start by defining who your assistant is. Write instructions that set its role, tone, personality, and boundaries — just like briefing a new team member.
- Knowledge Base via Vector Stores: Upload your own files — product catalogues, FAQs, return policies, pricing sheets — and your assistant will search them to answer customer questions accurately. Use this in combination with the File Search tool.
- Web Search: Give your assistant access to information from the web.
- Function Calling: Connect your assistant to external systems — checking order status, retrieving customer data, submitting a form, updating a CRM record. Any action your systems support can be triggered directly from a customer conversation. This requires external technical integration, but unlocks fully automated end-to-end workflows.
- Voice Message Support: Audio messages are automatically transcribed before being passed to the assistant. Note: Azure requires a separate resource for transcription — different API Key, Endpoint, and deployment from your main model. You’ll configure both sets of credentials in the extension.
- AI Inbox Replies: An optional module that generates reply suggestions. Useful for guiding customers into the right flow, or helping your team respond faster with AI-generated drafts.
- Smart Conversation Memory:Two settings control how your assistant manages context across sessions:
- Number of chat history messages how many previous messages are included as context when a new thread starts.
- Max Chat Session Duration (days) how long a thread stays active before a new one is created.
Setup: What's different about Azure
Azure OpenAI works a bit differently from the OpenAI API, so the configuration flow has a few extra steps. You’ll need to:
- Create your Azure resources and obtain an API Key and Endpoint
- Deploy a model inside Azure and give it a deployment name during setup — it doesn’t need to match the original model name. Then, select that same deployment name in this extension.
- Optional: Create a separate resource for transcription if you want voice message support — this needs its own API Key, Endpoint, and deployment.
- Configure the extension with your endpoint, API key, deployment name, tools, and Vector Stores.
- Connect to your channel and test.
Most Azure OpenAI users will already be familiar with this setup. A step-by-step tutorial will be available at launch.
