- Getting started
- Data security and compliance
- Organizations
- Authentication and security
- Licensing
- About licensing
- Unified Pricing: Licensing plan framework
- Flex: Licensing plan framework
- Activating your Enterprise license
- Upgrading and downgrading licenses
- License migration
- Requesting a service trial
- Assigning licenses to tenants
- Assigning user licenses
- Deallocating user licenses
- Monitoring license allocation
- License overallocation
- Licensing notifications
- User license management
- Tenants and services
- Accounts and roles
- AI Trust Layer
- External applications
- Notifications
- Logging
- Testing in your organization
- Troubleshooting
- Migrating to Automation Cloud
Automation Cloud admin guide
About Context Grounding
Context Grounding is a component of the UiPath AI Trust Layer that makes your business data LLM-ready — without additional subscriptions to embedding models, vector databases, or LLMs. It creates indexes and embeddings of your data that UiPath GenAI features can reference at runtime via retrieval augmented generation (RAG).
As a tenant-scoped RAG-as-a-service, Context Grounding grounds your prompts with relevant information before LLM execution. It supports the following UiPath generative AI experiences:
- GenAI Activities
- Autopilot for Everyone
- UiPath Agents
Providing RAG to these experiences helps to:
- Overcome context window limitations and improve model accuracy, reliability, and efficiency.
- Reduce hallucination risk through references to ground truth data.
- Give generative apps access to specialized, proprietary, and up-to-date knowledge.
- Enable feedback loops between data stores and user queries.
Core components
The terminology and core components of Context Grounding include:
Figure 1. Context Grounding component architecture
Ingestion and indexing
- Ingestion: Converts business data into representative embeddings using UiPath-managed embedding models.
- Embedding: A representation of business data that an LLM can understand and search.
- Index: A folder in a vector database that organizes embeddings.
- Vector DB: A UiPath-managed vector database that stores embeddings organized in indexes.
Retrieval
Context Grounding searches through LLM-ready business data to find the most relevant information, using extraction, chunking, retrieval, and re-ranking techniques optimized for different data formats and queries. GenAI Activities, Autopilot for Everyone, and Agents interpret prompts as queries and return the most relevant results via cosine similarity search — an intermediate step before RAG.
Retrieval Augmented Generation (RAG)
Grounds prompts with the most relevant information from semantic similarity search results, then executes generation via an LLM hosted through the AI Trust Layer's LLM Gateway.
DeepRAG (Deep Research-Augmented Generation)
An advanced RAG capability, generally available for Agents, that enables multi-document synthesis and retrieval for complex agent use cases. Agents can synthesize information across multiple documents and deliver comprehensive, citation-backed answers. Currently supported for PDF files only.
Key features
- Multi-document support: CSV, DOCX, JPG, JSON, PDF, PNG, TXT, and XLSX.
- Multi-modal ingestion: Processes documents containing both images and text, including non-native (scanned) PDFs.
- Structured query support: Advanced CSV querying, available when adding an index to an agent.
- Multilingual support: Ingests and queries documents in all UTF-8 encoded languages.
- DeepRAG: Multi-document synthesis for complex agent queries, with citation-backed answers. Available for PDF files only.
- Proof of knowledge: Cites the reference source and text from semantic similarity search results.
- Custom LLM support: Bring your own model and Bring your own subscription configurations through the AI Trust Layer, allowing administrators to use their own embedding and inference models. For details, refer to Bring your own LLM for Context Grounding.
- Multiple data sources:
- UiPath Orchestrator bucket entities: Ingest, index, and query data stored in shared Orchestrator folders.
- Document storage systems: Access data from Dropbox, Google Drive, and Microsoft OneDrive & SharePoint via Integration Service connectors.
Licensing
Refer to Context Grounding licensing.
Limitations and considerations
- Supported file types: CSV, DOCX, JPG, JSON, PDF, PNG, TXT, and XLSX.
- Index limit: Ten indexes per tenant, expandable on request. We recommend a 1:1 relationship between indexes and folder paths in your data source.
- Folder permissions: Indexes inherit folder permissions. Users without access to a shared folder cannot view, update, delete, or use its associated indexes.
- Studio version requirement: To use Context Grounding through GenAI Activities, you must use Studio Web or Studio Desktop version 2024.4 or later. For more information, refer to Working with Context Grounding.