Language models are excellent at processing and creating text. However, they struggle to deliver correct answers when interacting with relational databases. This is because database structures are typically described in cryptic terms which have no meaning to LLMs.
eRAG, leveraging GigaSpacesโ extensive expertise in operational data, interprets the context of structured data – filling the gap for LLMs – and ensuring accurate trustworthy responses delivered in an end-to-end SaaS application.ย
Deep semantic understanding of operational data
eRAGโs powerful reasoning engine automatically extracts relevant metadata and enriches it with the correct organizational context for LLMs. It augments data with meaningful table/column names and descriptions, dictionary and more, and stores it in an internal knowledge graph for split-second retrieval. Additionally,ย eRAG offers โhuman in the loopโ calibration for optimal accuracy and verification.
Provides precise queries & superb accuracy
Delivers precise SQL queries and highly accurate responses from multiple data sources:
- Automatically checks the cache for similar results
- Uses schema filtering and example selection to generate precise SQL statements
- Validates results with syntax checks, sanity tests and retries if needed
- Executes the statement in natural languageย
Highly accurate answers from diverse data sets
Quickly connect to multiple data sources and unlock the hidden value of structured data across your business.ย
- Easily add new tables or pipelines from cloud and on-premises sources
- Ensure accurate responses with advanced data virtualization that harmonizes SQL dialects
- Define the data scope for a specific business uses case or function, ensuring data relevancy and privacy
In your words: Asked and Answered
- Enables fast responses to ad-hoc questionsย
- Ensures that only those with required permissions can access specific dataย
- Refines and enriches previously provided answers when negative feedback is received
- eRAG advises the user about missing data and invites user inputs and validation
- Non-English queries and responses are normalized in English to limit inaccuracies due to language barriers
Ensure safe GenAI interactions with enterprise databasesย
eRAG offers real-time risk monitoring and mitigation, focusing on high-risk scenarios that can arise when customers interact with operational enterprise data (in the works).
It evaluates inputs for specific AI governance risks, generates an AI risk score and gives real-time alerts on regulatory or company policy violations.
