Data retrieval using embeddings for data in backup systems
AssigneeCohesity, Inc.
InventorsGregory Statton, Sanjay Poonen, Mohit Aron, Apurv Gupta
Life of this patent
Estimating the term…
Abstract
In general, techniques for efficient data retrieval from a backup system are described. An example computing system includes one or more storage devices and processing circuitry having access to the one or more storage devices and configured to: process an input to generate a filter, wherein the input indicates a context for one or more queries; apply the filter to backup data to obtain filtered data from the backup data; generate an index of embeddings from the filtered data; process, based on the index of embeddings, a query to generate a response for the query; and output the response.
Claims20
A computing system comprising:
- one or more processors;
- and
- one or more storage devices comprising instructions that are to configure the one or more processors to:
- receive an input indicative of a context for one or more queries;
- generate, based on the input, a filter that matches data relevant to the context for the one or more queries;
- apply, to backup data stored at a storage system, the filter to obtain a subset of the backup data that is relevant to the context for the one or more queries;
- generate an index of embeddings from the subset of the backup data;
- obtain, from the backup data stored at the storage system or from the subset of the backup data, based on the index of embeddings, context data for a query;
- generate, based on the query and the context data for the query, a retrieval augmented generation (RAG) query;
- and
- output the RAG query to a language model for processing.
A method comprising:
- generating, by a computing system, based on an input indicative of a context for one or more queries, a filter that matches data relevant to the context for the one or more queries;
- applying, by the computing system, to backup data stored at a storage system, the filter to obtain a subset of the backup data that is relevant to the context for the one or more queries;
- generating, by the computing system, an index of embeddings from the subset of the backup data;
- obtaining, by the computing system, from the backup data stored at the storage system or from the subset of the backup data, based on the index of embeddings, context data for a query;
- generating, by the computing system, based on the query and the context data for the query, a retrieval augmented generation (RAG) query;
- and
- outputting, by the computing system, the RAG query to a language model for processing.
Non-transitory computer-readable media comprising instructions executable by one or more processors to cause a computing system to:
- receive an input indicative of a context for one or more queries;
- generate, based on the input, a filter that matches data relevant to the context for the one or more queries;
- apply, to backup data stored at a storage system, the filter to obtain a subset of the backup data that is relevant to the context for the one or more queries;
- generate an index of embeddings from the subset of the backup data;
- obtain, from the backup data stored at the storage system or from the subset of the backup data, based on the index of embeddings, context data for a query;
- generate, based on the query and the context data for the query, a retrieval augmented generation (RAG) query;
- and
- output the RAG query to a language model for processing.
Dependencies are inferred from the claim text.