GPT 5.5 Support & Voice AI Scenario Tool

Release 1.153.0 | 30th July 2026

You can now use newer OpenAI models in Talkative Chatbot. Switch your config to the GPT 5.5 model to enable them, and it will continue to point to the latest OpenAI version so you stay on the newest release automatically.

The new Scenario tool for Voice AI lets you surface parts of your prompt only when they are needed. For example, adding guidance that only applies when a customer asks for a refund. This keeps your main prompt focused while still handling specific situations.

LLM Chatbot System Tools - LLM Chatbot now supports system tools. This is a set of built-in tools that provide baseline features such as the "current time" tool.
Interaction IDs in Tool Calls - the LLM Chatbot is now aware of the Interaction UUID and ID, so both can be used within tool calls.
Voice AI Model Upgrade - Voice AI GPT 5 bots have been upgraded to GPT 5.4, as GPT 5 is now deprecated.
SIP Caller Number - SIP calls into Voice AI now include the extracted phone number.
Transfer Validation - Voice AI now prevents you from configuring transfers that cannot succeed.
Duplicate KB URL Prevention - duplicate AI Knowledge Base URLs can no longer be added.
Clearer AI Disclosure - newly created Talkative Chatbots now clearly flag to customers that they are AI.
French Language Preset - added a French preset for Language Strings in Chat widgets.
Trigger Voice Assist Recordings Without the Console - you can now trigger Voice Assist recordings without opening the console.
Localised AI - you can now force localised AI. Reach out to your Customer Success Representative to turn this on.

Long Tool Responses - extremely long tool responses are now recorded correctly within LLM Chatbot.
Transcript Email Names - transcript emails now use the correct translation for the customer's name.
WhatsApp Usernames - improved behaviour around WhatsApp usernames.
Withheld SIP Numbers - withheld numbers over SIP now transfer correctly when using ORIGINATOR.
Empty Tool Responses - fixed an issue where empty tool responses could break Voice AI.

You can now use newer OpenAI models in Talkative Chatbot. Switch your config to the GPT 5.5 model to enable them, and it will continue to point to the latest OpenAI version so you stay on the newest release automatically.

The new Scenario tool for Voice AI lets you surface parts of your prompt only when they are needed. For example, adding guidance that only applies when a customer asks for a refund. This keeps your main prompt focused while still handling specific situations.

LLM Chatbot System Tools - LLM Chatbot now supports system tools. This is a set of built-in tools that provide baseline features such as the "current time" tool.
Interaction IDs in Tool Calls - the LLM Chatbot is now aware of the Interaction UUID and ID, so both can be used within tool calls.
Voice AI Model Upgrade - Voice AI GPT 5 bots have been upgraded to GPT 5.4, as GPT 5 is now deprecated.
SIP Caller Number - SIP calls into Voice AI now include the extracted phone number.
Transfer Validation - Voice AI now prevents you from configuring transfers that cannot succeed.
Duplicate KB URL Prevention - duplicate AI Knowledge Base URLs can no longer be added.
Clearer AI Disclosure - newly created Talkative Chatbots now clearly flag to customers that they are AI.
French Language Preset - added a French preset for Language Strings in Chat widgets.
Trigger Voice Assist Recordings Without the Console - you can now trigger Voice Assist recordings without opening the console.
Localised AI - you can now force localised AI. Reach out to your Customer Success Representative to turn this on.

Long Tool Responses - extremely long tool responses are now recorded correctly within LLM Chatbot.
Transcript Email Names - transcript emails now use the correct translation for the customer's name.
WhatsApp Usernames - improved behaviour around WhatsApp usernames.
Withheld SIP Numbers - withheld numbers over SIP now transfer correctly when using ORIGINATOR.
Empty Tool Responses - fixed an issue where empty tool responses could break Voice AI.