Bring your own LLM for Context Grounding

Context Grounding supports Bring Your Own Model (BYOM) and Bring Your Own Subscription (BYOS) through the UiPath AI Trust Layer. Administrators can replace UiPath-managed embedding and inference models with their own subscriptions, or connect models not included in the default set, while retaining the governance and audit capabilities of the AI Trust Layer.

Supported features and models

The following Context Grounding features support custom LLM configuration through any provider whose API follows the OpenAI V1 standard, connected via the OpenAI V1 Compliant LLM connector. Each feature has specific model requirements.

Feature	Supported models	Recommended model	Requirements	Recommended specs
Advanced ingestion	Any OpenAI V1-compatible LLM	gemini-2.5-flash	Forced tool calling; multimodal image support	16k input tokens; 32k output tokens
Batch Transform	Any OpenAI V1-compatible LLM	gemini-2.5-flash (smart), gemini-2.5-flash-lite (fast)	Forced tool calling	128k input tokens
Batch Transform with Web Search	Any Gemini web search model	N/A	Gemini web search tool	N/A
DeepRAG	Any OpenAI V1-compatible LLM	gemini-2.5-flash	Forced tool calling; count > 1 support	1M input tokens; 64k output tokens
Embeddings	Any OpenAI V1-compatible Embedding model	gemini-embedding-001	< 4,096 dimensions; ≥ 8k input tokens	N/A

For the full model scope across all UiPath products, refer to Configuring LLMs.

Requirements — without this capability, the feature does not function.
Recommended — without this capability, the feature is untested and may perform worse or fail more often.

Configuration and routing

Custom LLM configuration for Context Grounding is managed through the LLM configurations tab in Admin > AI Trust Layer. For setup steps, select Context Grounding as the product when following the steps in Configuring LLMs.

For details on how AI features route requests to models, including fallback behavior and model governance, refer to AI features and model routing.