Tokkio Migration Guide (4.1 → 5.0 Beta)#

This guide outlines the process for migrating from Tokkio 4.1 to Tokkio 5.0 Beta. It provides detailed instructions for updating each component of your existing Tokkio implementation.

1. Connecting to your Knowledge Source:#

  1. Connecting your NVIDIA RAG compatible RAG: Tokkio 5.0 Beta supports connecting to NVIDIA RAG Blueprint and is compatible with any RAG built using the Generative Examples repo. To update the RAG endpoint, follow the steps in Digital Human Agent Customization.

  2. Connecting your custom RAG service: Tokkio 5.0 Beta supports connecting to your custom RAG by writing a custom RAG frame processor. Refer to the NVDIARAGService library to define your RAG schema and then refer to adding a frame processor customization for integrating it into the Tokkio 5.0 Beta workflow.

2. Applying your Bot Behavior:#

  1. You can customize bot behaviors including the avatar name, greetings, proactivity messages, acknowledgement words, and bot role by configuring the appropriate frame processors. Follow the steps in Digital Human Agent Customization for more information.

  2. Conversation history: You can enable custom bot history in Tokkio 5.0 Beta by using either the OpenAILLM_Service Context Aggregator Frame Processor or the NVIDIA Context Aggregator with speculative speech processing enabled. Follow the Advanced Customizations to implement this feature.

3. Migrating Speech AI Components:#

  1. Modifying TTS behavior: Changing voice, pronunciation (IPA) and other TTS parameters: editing voice ID, pronunciation hints, language and other TTS behaviors can be configured following Advanced Customizations.

  2. Swapping to ElevenLabs or other TTS providers: Swapping to another TTS provider is enabled in Tokkio 5.0 Beta to any of these providers by importing the target service and then following Advanced Customizations. See the ElevenLabs docs for more information.

  3. Enable speculative speech processing: When using RIVASRService in the pipeline, you can enable speculative speech processing by following this guide and then implementing the code-based customizations described in Advanced Customizations.

4. Migrating Dialog Management:#

  1. Fixed intents: Fixed dialog intents are handled through the custom RAG service or a custom knowledge source service in Tokkio 5.0 Beta.

  2. Adding guard railing: Basic guardrails can be added to Tokkio 5.0 Beta by adding the guardrails frame processor to the pipeline by following Advanced Customizations. View Code

5. Migrating the Avatar and Animation:#

  1. Migrating avatars from the Omniverse Renderer to the Unreal Engine Renderer microservice:

  1. Avatar Configurator avatar: Tokkio 5.0 Beta currently does not support the stylized avatars from the avatar configurator.

  2. Custom avatars / scene: Custom avatars, depending on the creation or DCC tool used, would require the character to be exported to Unreal Engine in FBX format with the ARKit blendshapes, once the materials and shaders are tweaked to the desired output, follow Avatar and Scene Customization to build & integrate the compatible avatar.

  1. Audio2Face-3D configuration changes: Tokkio 5.0 Beta supports Audio2Face-3D configurations through a tuning workflow to support any stylization and customizations dynamically using Audio2Face-3D microservice customization. See Audio2Face-3D Microservice Customization - Experimental for more information.

6. Migrating the UI:#

  1. Tokkio Iframe: Tokkio 5.0 Beta combines the Iframe and UI into a unified web application that runs either as a standalone application or as an Iframe within an existing website. The UI provides toggleable features including a start page, transcripts, and full-screen toggle. You can customize these features by following Customization Options.

7. No Vision AI#

  1. There is no Vision AI pipeline support in this Tokkio 5.0 Beta release. The pipeline will function similar to the no vision deployment from the previous 4.x releases.

On this page