LLM Query Architecture for Public and Private Data Retrieval
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional search engines struggle to leverage private data sources effectively, leading to incomplete and frustrating search results, as they primarily rely on public data indices and fail to integrate user-specific private data, such as workplace information, into search responses.
Innovation Solution
An intelligent chat system powered by a large language model that integrates both public and private data sources, using user credentials to access private data, and generates a natural language answer supplemented with relevant links, enhancing the search experience by providing comprehensive and efficient responses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If search engines rely on pre-populated public URL indices, then search results can be retrieved efficiently, but the results are incomplete and fail to include user-specific private data
Solution Approach 1:
The system segments search data sources into public indices and private data sources, querying each type appropriately. Public search indices handle general information retrieval while private data sources provide user-specific information, allowing the system to maintain simplicity for public queries while integrating private data when needed.
Solution Approach 2:
The large language model serves as an intermediary that receives queries, determines whether private data access is needed, and coordinates between public search indices and private data sources. This intermediary manages the complexity by abstracting the dual-data-source architecture behind a unified query interface.
2Loss of information
If search engines integrate multiple data sources including private data, then search results become more comprehensive, but the system complexity increases
Solution Approach 1:
The large language model performs multiple functions: it acts as a query processor, a decision-maker for data source selection, and a coordinator for retrieving information from both public and private sources. This multi-functional component handles data source integration complexity centrally, allowing the rest of the system to remain relatively simple.
Solution Approach 2:
The system automatically determines when private data access is needed and retrieves appropriate data without requiring manual user configuration. The large language model autonomously manages the complexity of integrating multiple data sources by making real-time decisions about which sources to query based on the query content.
3Ease of operation
If conventional search engines provide only public data results, then the system remains simple to operate, but user satisfaction decreases due to incomplete information
Solution Approach 1:
The large language model acts as an intermediary between the user and the dual data source system, presenting a simple natural language interface while handling the complexity of querying both public and private sources. Users interact through simple questions and receive comprehensive answers without needing to understand or configure the underlying complex data integration architecture.
Solution Approach 2:
The system dynamically changes its data retrieval parameters based on the query, switching between public-only mode and public-plus-private mode as needed. This allows the system to maintain simplicity for straightforward public queries while automatically enhancing completeness when private data would be beneficial, without requiring users to adjust any parameters.
Data Source
AI summary
Aspects of the disclosure include methods and systems for an intelligent chat powered by a large language model that leverages both public and private data to answer user questions. An exemplary method includes receiving a user query including natural language input from a user and executing the user query against at least one public data source and at least one private data source. Queries executed against a public source are retrieved using public search indices and queries executed against a private data source are retrieved using user credentials. A query rewrite and a query context including the user query and retrieved information from the public and private data sources are input to a large language model. A response is received from the large language model that includes a natural language answer to the user query and a link to the retrieved information.