Concurrent communication in distributed server system for networked content retrieval and graph network data transmission

A method for increasing relevant content transmission for a distributed network comprising:

• storing, in a networked plurality of servers comprising a first server and a second server, a distributed graph by storing first content graphs in the first server and second content graphs in the second server;
• generating a query graph based on a natural language query obtained from a client device;
• retrieving an intermediate set of graphs by concurrently using the first server and the second server to retrieve respective graph subsets in parallel using the same first graph neural network executing on respective processors of the first and second servers, wherein (1) the first server retrieves a first graph subset with the first graph neural network to detect similarities between the query graph and the first content graphs and (2) the second server retrieves a second graph subset by using the first graph neural network to detect similarities between the query graph and the second content graphs;
• using a second graph neural network to re-rank similarities between the query graph and the intermediate set of graphs to select a ranked graph subset satisfying a threshold;
• retrieving selected content comprising data from the first server and the second server based on an association between the selected content and the ranked graph subset;
• and
• generating a language model response by transmitting, to an application program interface of a large language model, an input comprising the natural language query and a query augmentation context comprising the selected content.

One or more non-transitory, machine-readable media storing program instructions that, when executed by one or more processors, causes the one or more processors to perform operations comprising:

• storing first content representations in a first server and second content representations in a second server;
• generating a query representation based on a query;
• retrieving an intermediate set of representations by concurrently using the first server and the second server to retrieve respective representation subsets in parallel, wherein the first server detects similarities based on the first content representations to retrieve a first representation subset and the second server detects similarities based on the second content representations to retrieve a second representation subset;
• using a graph neural network to re-rank similarities between the query and the intermediate set of representations to select a ranked graph subset satisfying a threshold;
• determining an input context comprising at least a portion of selected content comprising data from the first server and the second server based on an association between the selected content and the ranked graph subset;
• and
• generating a language model response by transmitting, to an application program interface of a large language model, an input comprising the query and the input context.