UI Enhancements, API Improvements, Cost Optimisations & Stability Fixes

Release Notes - 25.2.2026

This update brings a major upgrade to the Chat Widget experience, better styling control, stronger API flexibility, smarter conversation handling, and important reliability improvements across the platform.


✨ What’s New

A faster, more flexible Chat Widget (now rebuilt in React)

We’ve fully rebuilt the Chat Widget in ReactJS to improve performance, flexibility, and compatibility with modern websites and frontend setups.

What this means for you:

  • Smoother widget experience
  • Better long-term performance and stability
  • Easier integration into modern web apps

A completely redesigned Appearance tab

The Appearance section has been redesigned to make widget customisation much easier — whether you prefer visual controls or full design control.

Two ways to customise your widget

  • UI Settings Mode – adjust the design visually using structured settings
  • Custom CSS Mode (New) – apply your own CSS for advanced styling

Expanded design controls

You now have more control across areas like:

  • Content & messages
  • Layout & display
  • Features
  • Automation & timing
  • Typography
  • Colors & styling
  • Images & assets

Live Preview (real-time)

See your widget updates instantly before saving.

Why it matters:
You can now create a much more brand-aligned widget experience — from simple tweaks to fully custom styling.


Human / Agent mode now available via API

The Human / Agent toggle from the Conversations UI is now available through a public API endpoint.

Previously, this could only be changed manually in the dashboard. Now it can be triggered programmatically.

You can now switch a conversation to:

  • human → hand the conversation over to a human
  • agent → return control to the AI Chat Agent

Endpoint supports:

  • conversation_id
  • mode

It returns the updated conversation state and mirrors the exact behavior of the UI toggle.

Why it matters:
This makes it easier to build automated handoff workflows, escalation logic, and integrations with external tools.


Smarter conversation title generation (with lower API usage)

Conversation titles are now generated within the same OpenAI request as the assistant response.

Previously, title generation required a separate request. This update removes that extra step.

What’s improved

  • Fewer API calls
  • Lower token usage
  • Faster overall conversation flow

New setting added

A new LLM-Driven Conversation Titles toggle is now available in Chat Agent Settings:

  • Enabled → titles are generated automatically
  • Disabled → no LLM-generated titles

Why it matters:
Better efficiency, lower cost, and more control over how titles are handled.


Lead Collection Form improvements

We’ve improved the Lead Collection Form to make lead capture more structured and more useful during live conversations.

What’s new

  • Dropdown field support (with configurable options)
  • Chat Widget form submissions are now visible directly to agents

Why it matters:
Better lead capture quality, more context for agents, and smoother sales/support workflows.


Webhook history reliability improvements

We fixed an issue that could cause incomplete records to appear in webhook history when pagination was involved.

Also added

  • A new Resend action in the webhook history table

Why it matters:
Improved visibility, easier recovery, and more reliable webhook management.


Better JSON formatting in Custom Function history

JSON payloads in the Custom Function history table are now formatted more clearly.

Why it matters:
Cleaner output and easier debugging when reviewing function responses.


Model upgrade: Grok migration

We’ve upgraded from deprecated grok-2-1212 to grok-3.

Why it matters:
Improved compatibility, ongoing support, and better performance going forward.


✅ Summary

This update is focused on improving the parts that matter most in day-to-day usage:

  • Better widget performance
  • Stronger design flexibility
  • More automation control via API
  • Lower model/API overhead
  • More reliable operational workflows

More upgrades are on the way.