OpenAI

OpenAI provides state-of-the-art AI models through its API, allowing developers to integrate advanced AI capabilities into their applications easily:

• Access pre-trained models for natural language processing, reasoning, embeddings, and more.

• Fine-tune the included pre-trained models with your organizational IP for a specialized model.

For more information on OpenAI, please see the official site.

The OpenAI Inferencing Agent makes a direct, stateless call to a model of your choice and includes the response in the Data Stream.

The OpenAI Assistant Agent makes a direct call to a pre-built Assistant you already defined in OpenAI (Assistants API), referenced by its Assistant ID. The model, instructions, and tools live in that Assistant, not on the agent; the agent passes your prompt to it and returns the Assistant's reply into the Data Stream.

Pre-requisites

The following is required to use the OpenAI Inferencing Agent:

• OpenAI API Key

• Stream Host (SH) v4.4.22 or greater

• Network access to the OpenAI services

The following is required to use the OpenAI Assistant Agent.

• OpenAI API Key

• OpenAI Assistant ID.

• Stream Host v4.6.0 or greater

• Python is installed on the system where the Stream Host is installed - along with the OpenAI Python Library (refer to the Python Agent for further details on the install options)

• Python version 3.9 - 3.12

• Network access to the OpenAI services

Current Version

• Download the OpenAI Inferencing Agent v1.03 (Last Update: 14 Jul 2026)

• Download the OpenAI Assistant v2.0 (Last Update: 15 Jun 2026)

Please contact XMPro if you're looking for an older version of this Agent.

OpenAI Inferencing Release Notes

Version	Date	Description
1.03	14 Jul 2026	Initial Release.

OpenAI Assistant Release Notes

Version	Date	Description
2.0	15 Jun 2026	Support Event-Level Streaming (when using XMPro Platform v4.6+) for reduced memory usage and earlier downstream processing. Fix FileLoadException: Could not load file or assembly 'ManagedCuda.dll' on GPU-less Stream Hosts, by validating for GPU availability using CUDA detection on data stream publish. Fix Stream Host container crash with no error message by validating Python paths on data stream publish.
1.10	08 Jan 2026	Repackaged to add missing dropdown translations.
1.02	23 Dec 2024	Repackaged to translate the Agent's properties and change the category to Generative AI.
1.01	10 May 2024	Initial Release.