It is intended for customers who already have a working chatbot on v1.2.0 and need to recreate it on v1.3.0 before the Assistants API is shut down.
Why you need to migrate
OpenAI has deprecated the Assistants API and scheduled its shutdown for August 26, 2026. After that date, any chatbot still based on v1.2.0 may stop working. To avoid service interruptions, migrate to a v1.3.0 chatbot based on the Responses API before that date.
| The migration is not automatic | There is no button that upgrades v1.2.0 to v1.3.0. You create a new v1.3.0 extension and rebuild the configuration. |
| A new extension is created | Your v1.2.0 extension is left untouched. You install a separate v1.3.0 extension alongside it. |
| Your v1.2.0 chatbot keeps running | Keep the old chatbot live until the new one is fully validated, so your customers are never left without a reply. |
| End users are not affected | Customers keep writing from the same WhatsApp, Instagram, Facebook Messenger, or Webchat channel. Only the internal connection to OpenAI changes. |
Recommended approach: migrate with testing.Build the new v1.3.0 chatbot, validate it in a test environment with a small group of testers, and only switch it to production once you are confident. This way your customers keep talking to the proven v1.2.0 chatbot the whole time, with no interruption.
Make sure you have:
The fastest way to migrate is to open your v1.2.0 configuration in one tab and the new v1.3.0 extension in another, then copy each setting across.
Do not disconnect or delete your current chatbot yet. It stays live and keeps replying to customers for the entire time you are building and testing the new one. You will only switch it off at the very end, during cutover (Step 6).
You will create the new chatbot from scratch as a separate extension.
my-gpt-assistant-v13.
For the complete, screenshot-by-screenshot setup, follow the full guide: How to Create an AI Chatbot with OpenAI in WOZTELL. This migration guide focuses on the steps that are specific to moving from v1.2.0.
Important: You cannot configure the chatbot or connect it to a channel until the API Key is saved.In v1.2.0, all of your chatbot's behavior lived in the prompt (Instructions). Migrating it is mostly a matter of copying that prompt across and selecting the right model.
In the new v1.3.0 extension, open Configuration → AI Chatbot and set the instructions (prompt), model, transcription model and advanced settings.
Important: Save each tab before switching to another one — unsaved changes may be lost.
This is the concept that makes a safe migration possible, so it is worth understanding before you connect anything.
During migration, environments let you run the new v1.3.0 chatbot in a test environment while your v1.2.0 chatbot stays on the live (Default/production) environment. Your customers keep talking to the old, proven chatbot until you are ready to switch.
Keep the two chatbots on separate environments. Two chatbots ending up active within the same environment will cause unpredictable replies (users reaching one bot or the other with no control). Using a separate test environment is exactly how you prevent this.
Now connect the new chatbot — but to a test environment, not to production.
AI-CHATBOT-WHATSAPP-TEST.When the button changes to Disconnect, the new chatbot is live in the test environment only.
Now only your whitelisted testers are routed to the test environment and the new v1.3.0 chatbot. Production — and your real customers — stay on v1.2.0, untouched.
Have your testers put the new chatbot through its paces:
Once you are confident in the new chatbot, switch your live channel from v1.2.0 to v1.3.0.
Important: Disconnect v1.2.0 first, then connect v1.3.0, so the two are never active in the same environment at once.
Once you have confirmed v1.3.0 is working in production, you can safely remove the old v1.2.0 extension.
| Issue | Cause | What to do |
|---|---|---|
| Unpredictable replies / users reaching the wrong bot | Two chatbots ended up active within the same environment. | Keep v1.2.0 and v1.3.0 on separate environments. During cutover, disconnect one before connecting the other in the same environment. |
| Testers can't reach the new chatbot | No Priority Group assigned, or testers not added to it. | Confirm the Priority Group is assigned to the test environment and that each tester is whitelisted (WhatsApp testers must message the WABA first). |
| Some prompt-based logic behaves differently | The Responses API handles context differently from the Assistants API. | Review and adjust the prompt during validation; consider moving reference content into File Search. |
| Assistant not responding in the test environment | API Key not saved, chatbot not connected, or wrong environment selected. | Confirm the API Key is saved, the chatbot is connected, and you are messaging the correct environment. |