PromethistAI — Release Notes

Track your agent’s configuration history with automatic versioning. Every time you save changes, a new revision is created. You can preview, publish, or revert to any previous revision.

New controls include taking photos with the camera or choosing images from the gallery on OpenAI agents. This enables more interactive and contextual discussions.

Visual quality settings have been simplified. Scene resolution is now automatic, and the new "Auto" quality mode adapts to your device’s performance, no manual adjustments needed.

Kiosk stations now support using your mobile device as a microphone. Scan the QR code on the kiosk with your phone to use it as the audio input for the session.

We made improvements to extractors helping your agents track and optimize specific metrics during conversations.

Agents can now transfer conversations to other relational agents using the new Handover multimodal interaction. Configure automatic or specific agent routing to escalate conversations or direct users to specialists.

You can now embed Relational Agents directly into your website using the PromethistAI Widget. Add a simple script tag to display a floating assistant that users can interact with.

The Web Client now supports Android devices and Linux desktop browsers, making it accessible on even more platforms without installing a native application.

Visual quality now automatically adjusts to your device for smoother operation (Android and iOS only).

Performance has been improved in multiple areas for a faster, more stable experience.

You can now control the platform through a conversational interface. Create agents, navigate between sections, and make configuration changes using natural language instead of clicking through menus.

Access your Relational Agents directly in your browser without installing any applications. The Web Client is now available for Windows (Chrome, Edge) and iOS 16+ (Safari) with WebGPU support.

Connect your agents to external services and data sources using webhooks and MCP servers.

We improved avatar visual fidelity with a new realistic hair shader and better lighting, including fixes for hair blurring.

We refined overall movement and scene behavior with reduced stance-change swinging and a unified, simplified loading sequence across Avatar and Environment.

We improved compatibility and reliability, including fixes to static camera behavior.

We added Turn Taking option in Settings (iOS and Android), when enabled, prevents interruptions.

Kiosks can now display a QR code that mobile users can scan to start a personalized conversation

Project assets (Knowledge Bases, Integrations and Multimodal Interactions) can now be marked as Global to make them accessible to all agents within the same project

A primary featured agent now displays prominently at the top of the home screen for signed-in users

Push-to-talk mode is now available also on iPhone

We improved audio streaming with reduced buffering time for faster, more responsive voice interactions

We improved scene and avatar loading reliability

Let agents optimize conversations for custom, measurable goals (e.g., brand sentiment, product knowledge, issue detection).

Choose whether agents recall past conversations for each user

Compare performance across versions to see what changes worked

The agent detects when you stop speaking and adjusts continuing or offering to end the session

The QR button has moved from the Identity sidebar to the main Talk screen (lower-right corner) for quicker sharing and device pairing

We improved conversational responsiveness with faster audio processing

We made avatar animations smoother across different devices

Cleaner, smoother background blur with less pixelation

We made major performance improvements with an updated rendering system

We refined avatar posture and movement to look more natural

We improved audio reliability for smoother conversations

We’re putting the finishing touches on a new planning component that will let you opt in per agent for quicker reasoning and more reliable function calls

If enabled, kiosk agents can now use the device camera to perceive their surroundings (e.g., detect presence, read signage, or react to context)

Addressed cases where images weren’t shown to users; display is now consistent

The agent automatically shows a close-session prompt when it detects that you’re done with the conversation

In the PromethistAI iOS and Android apps, you can now switch between US and EU clusters directly in Settings → Region. The default is autodetected based on your location for optimal performance

Now you can scan a QR code for an agent hosted in a different region (e.g., EU QR on a US-configured device). A warning will appear, but the session will launch without interruption

You can now choose where your agent runs—US or EU—and which backend cluster the Administration Console connects to

Keep your workspace tidy and searchable. You can now rename any Knowledge Base from its menu or within the Knowledge Base settings page

Limit which languages an agent will speak using the new Accepted Languages control

Want a little extra privacy? Start a session without linking it to your identity—toggle Anonymously under the agent’s Talk button (per session)

Your agent just got smoother: new "thinking" animations, better lip-sync, and steadier resolution scaling for a more natural feel