On-Prem LLM Integration with UiPath Cloud?

Hi everyone,

We are planning an architecture where UiPath Orchestrator is in Automation Cloud, but the LLM we want to use will be deployed in our on-premises environment.

I would like to confirm if this setup is supported and understand the best practice.

Our Scenario

  • UiPath Orchestrator / Automation Cloud
  • Custom LLM hosted on-prem (inside our private network)
  • User want use their own-llms trained so that data is not sent anywhere

Questions

  1. Connectivity
    Can Automation Cloud (via the AI Trust Layer / GenAI Activities) securely call an LLM endpoint that is hosted on-prem, provided we expose it through a secure HTTPS endpoint?

  2. Network Requirements
    Are there any official guidelines on:

    • IP allowlisting for UiPath outbound traffic
    • Required security setup (API gateway, reverse proxy, etc.)
  3. Product Compatibility
    If this connection is established, can the on-prem LLM be used with:

    • GenAI Activities in Studio
    • Document Understanding or IXP (where applicable)

Our goal is to keep the model and data processing inside our environment while still using UiPath Cloud for orchestration.

Thanks in advance for your guidance!

@Lahiru.Fernando @loginerror @Palaniyappan

Hi @Mahendra_T,

Welcome to the UiPath Community.

If you want to use your on-prem hosted LLM model in UiPath, then there are 2 options-

  1. Use the OpenAI V1 Compliant LLM activities. This currently supports only one activity - Generate Chat Completion

2- You can use the Bring Your Own Model(BYOM) service of UiPath. Please refer to this blog for more info - https://www.uipath.com/blog/product-and-updates/introducing-bring-your-own-model-to-uipath

And for configuring the external LLMs, please refer to this - https://www.uipath.com/blog/product-and-updates/introducing-bring-your-own-model-to-uipath

Talking about the Product compatibility, currently you cannot use custom LLM models for IXP and Gen AI activities but it may be possible in the future. But you can use custom models to power your Agents in UiPath.

Thanks

1 Like

Hi @Karan_Rautela

Thanks for the response

I am exploring options to use our own on-prem LLM model integrated with GenAI Activities to extract data from documents. If UiPath does not currently support this, we may need to consider the IXP approach instead.

Hi @Mahendra_T

Yes, it’s supported if your on-prem LLM is exposed as a secure HTTPS API and integrated via the AI Trust Layer (custom connector). Automation Cloud cannot directly access private networks, so the endpoint must be reachable over the internet. GenAI Activities can use it; DU/IXP support depends on the specific component.

1 Like

Has anyone used or does anyone have knowledge of how to use a custom LLM for IXP or DU, and how we can do the retraining when using a custom LLM?

Yes, custom LLMs can be used with IXP/DU only where GenAI endpoints are supported, via the AI Trust Layer. UiPath does not handle retraining you retrain and redeploy the model externally, then point UiPath to the updated endpoint.