Release Notes

ACE Agent Release 4.0.0

This is the NVIDIA ACE Agent 4.0.0 Release Notes.

Key Features and Enhancements

  • ACE Agent supports streaming for both text and speech bot responses, as well as word level streaming for text and sentence level streaming for speech response.

  • Added prebuilt support for NVIDIA GenerativeAIExamples RAG workflows in ACE Agent bots.

  • Easy integration of LangChain and LlamaIndex based Agents with ACE microservices using the ACE Agent Plugin server.

  • Added support for Colang 2.0 and Latest NVIDIA NeMo Guardrails (0.8.1). Colang 1.0 is still supported and will be deprecated in future releases.

  • Improved ACE Agent Event Interface to support complex multimodal interactions between a bot and a user.

  • Added support for NVIDIA API Catalog models and on-prem LLM deployment using NVIDIA NIM.

  • Added the capability to override a prebuilt generate_user_intent and generate_bot_message system actions to support custom logic.

  • Included a set of sample bots showcasing various ACE Agent capabilities.

Compatibility

Breaking Changes

  • In the last release, NVIDIA Bot Maker was renamed to NVIDIA ACE Agent. In this release, the containers, microservices, and resources are updated with the latest naming. Bot Maker Bot Controller has been renamed to ACE Agent Chat Controller, Bot Maker Dialog Manager has been renamed to ACE Agent Chat Engine, the nvbotmaker tool has been renamed to the aceagent tool, and Bot Maker Fulfillment Server has been renamed to ACE Agent Plugin Server.

  • As part of this release we have deprecated embedding search and information retrieval support from the ACE Agent NLP server microservice. Use the Retrieval Augmented Generation (RAG) workflow as an alternative.

  • ACE Agent Microservices API schemas have been simplified and updated for supporting streaming responses.

Limitations

  • ACE Agent does not support features like secure communication and secret encryption which are critical for a production deployment. Users are expected to carefully evaluate the nature of their deployment and usage, and implement additional features and safeguard which may include identify and authentication, secure API communication, input and output filtering, and safeguards to ensure safe and trustworthy AI.

  • The sample bots and sample clients illustrated in the documentation are meant for testing and demonstrating different capabilities of the NVIDIA ACE Agent. They are not optimized for best accuracy or performance.

Known Issues

  • We use yfinance to extract market capital and yahoo-fin to extract real-time stock prices. If you’re seeing issues with packages, you can build the stock plugin using Alpha Vantage or similar API to get this information and integrate it with the existing bot.

  • In Docker deployment, you might observe permissions issues for a few specific Host UID and GID. Update docker-compose.yml with different users or maybe root user for avoiding the issue.

ACE Agent Release 3.0.1 EA

This is the NVIDIA ACE Agent 3.0.1 Early Access (EA) Release Notes. This EA release is for early testing and feedback.

Fixed Bugs

  • Removed the usage of the deprecated OpenAI model text-davinci-003 from the documentation and sample bots.

  • Fixed errors observed in the ACE Agent Chat Engine for the OpenAI model gpt-3.5-turbo-instruct.

  • Deprecated the NeMo Inform and Financial sample bots available as part of 3.0.0-ea release.

Known Issues

  • Sample bots provided as part of this release are for experimentation purposes, are not exhaustive, and shouldn’t be used in production without due diligence.

ACE Agent Release 3.0.0 EA

This is the NVIDIA ACE Agent 3.0.0 Early Access (EA) Release Notes. This EA release is for early testing and feedback.

Key Features and Enhancements

  • NVIDIA NeMo Guardrail and Colang-based dialog management and guardrailing support

  • Support for plugging in custom Pythonic dialog management code

  • NvBotMaker CLI tool for building your bot in a native Python environment

  • Docker compose-based deployment support

  • Deployment support on Kubernetes environment

  • Integration with NVIDIA NeMo LLM and NeMo Inform EA services

  • Integration with NVIDIA Riva translation

  • Standalone NLP server with dedicated endpoint for common NLP tasks

  • Connectors for custom NLP model and retrieval pipelines

  • Standalone Fulfillment server with ability to support custom endpoints

  • Conversion script and guidance for porting previous release-based bots to Colang

  • Concurrent deployment of multiple bots with different configs

  • Support for plugging in 3rd party TTS pipeline

  • Set of sample bots showcasing various ACE Agent capabilities

Announcements

  • NVIDIA Bot Maker has been renamed to NVIDIA ACE Agent starting with this release. Bot Maker Bot Controller has been renamed to ACE Agent Chat Controller and Bot Maker Dialog Manager has been renamed to ACE Agent Chat Engine. In this release, the Containers, Microservices, and resources are using old names, which will be updated in the next release

  • This NVIDIA ACE Agent release supports Riva Skills 2.13.0.

  • This NVIDIA ACE Agent release is based on NeMo Guardrails v0.5.0.

Limitations

  • This is an Early Access (EA) release, therefore, not all functionality is fully supported and working. This EA release is meant for testing, feedback, and research.

  • NVIDIA ACE Agent is provided as an SDK for testing and development and it is not a production ready solution. Users are expected to carefully evaluate the nature of their deployment and usage, and implement additional features and safeguard which may include identify and authentication, secure API communication, input and output filtering, and safeguards to ensure safe and trustworthy AI.

Known Issues

  • It is recommended to return string responses instead of numeric values in the Fulfillment server when there is a possibility of 0 as the response. This is because the ACE Agent interprets 0 and empty strings as None when getting values from the Fulfillment server, which can lead to unexpected behavior.

  • We use yfinance to extract market capital and yahoo-fin to extract real-time stock prices. If you’re seeing issues with packages, you can build the stock fulfillment using Alpha Vantage or similar API to get this information and integrate it with the existing bot.