Assist (formerly known as aiR Assist) is a conversational search tool integrated within Relativity aiR, designed to let legal teams interact with case data using natural language rather than traditional keyword search.
Key features and functions include:
Natural language querying with citations
Users ask questions in plain language against a selected index of documents. The system identifies the most relevant content, generates an answer using a large language model, and supports the response with citations linked back to source documents.
Retrieval-augmented generation (RAG) architecture
Combines a document retrieval step with LLM-based answer generation: the system plans and runs searches against the index, then passes the most relevant passages to the LLM to produce a grounded, citation-supported response.
Index and document limits
Supports up to five indexes per workspace, each containing up to 300,000 documents; individual documents must be under 5 MB of extracted text and must have text stored in Data Grid to be indexed.
Permission-aware responses
Every response is generated based only on the documents the requesting user is authorized to access at the time of the query, so different users may receive different answers to the same question.
Multi-language support
The underlying LLM has been evaluated across 83 languages; while testing has focused primarily on English-language documents, the tool is designed to support non-English datasets, with user-directed prompting available to specify a query or answer language.
Use-case flexibility
Supports common workflows including early case insight (surfacing key documents and themes), case strategy development (timelines, communications mapping), and deposition/trial preparation (exhibit suggestions, confirming communications).
Activity auditing
Logs user activity — including questions, answers, conversations, and index changes — in the Audit application for governance and review purposes.
Loading...