# US 12,762,164 B2 · Computing tool retrieval using sequence processing models > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,762,164 B2 - Document ID: US12762164B2 - Status: Granted - Kind: B2 - Application number: 18/444,094 - Filed: 2024-02-16 - Published: 2025-08-21 - Granted: 2026-10-13 - Earliest claimed: 2024-02-16 - Assignees: Google Inc. - Applicants: Google LLC - Inventors: Jinsung Yoon; Yanfei Chen; Chen-Yu Lee; Qingze Wang; Mohammad Hossein Bateni; Vincent Cohen-Addad; Devendra Singh Sachan - Web page: https://patentgrep.com/patent/US12762164B2 ## CPC - G06F 16/3334: {Selection or weighting of terms from queries, including natural language queries} - G06F 16/2438: {Embedded query languages} - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract A machine-learning system effectively and efficiently identifies computing tools that are relevant to processing a query for a sequence processing model. A system can stores, for each computing tool, data associated with at least one synthetic query generated by a machine-learned sequence processing model based on tool documentation for the computing tools. The system determines a subset of computing tools relevant to a particular user query based on the synthetic query for each of the plurality of computing tools. The system generates at least one prompt including the user query and a processing result from each of the subset of computing tools in response to the user query. The system generates a response to the particular user query based at least in part on an output of at least one machine-learned sequence processing model in response to the at least one prompt. ## Claims ### Claim 1 A computer-implemented method for improving efficiency of computing tool retrieval for a sequence processing model, comprising: for each of a plurality of computing tools, creating, by a computing system, an expanded tool documentation data encoding by encoding, via one or more machine-learned sequence processing models, at least one synthetic query into a multidimensional embedding space based on tool documentation for said each computing tool, wherein the at least one synthetic query represents a potential use case for said each computing tool; storing, by the computing system in an embedding database, a plurality of embeddings corresponding to the expanded tool documentation data encodings for the plurality of computing tools; receiving a particular user query; determining, by the computing system, a subset of the plurality of computing tools that are relevant to the particular user query by comparing an embedding of the particular user query to the plurality of stored embeddings, wherein the subset includes less than all of the plurality of computing tools; generating, by the computing system, at least one prompt for at least one machine-learned sequence processing model, the at least one prompt including the particular user query and a processing result from each of the subset of the plurality of computing tools in response to the particular user query; and generating, by the computing system, a response to the particular user query based at least in part on an output of the at least one machine-learned sequence processing model in response to the at least one prompt. ### Claim 2 (depends on 1) The computer-implemented method of claim 1, further comprising: providing, by the computing system to the one or more machine-learned sequence processing models for each of the plurality of computing tools, at least one synthetic query generation request based at least in part on tool documentation associated with said each computing tool; and obtaining, by the computing system from the one or more machine-learned sequence processing models for each computing tool, at least one synthetic query configured for processing by said each computing tool. ### Claim 3 (depends on 2) The computer-implemented method of claim 2, wherein providing, to the one or more machine-learned sequence processing models for each of the plurality of computing tools, the at least one synthetic query generation request comprises: providing at least one prompt to the one or more machine-learned sequence processing models, the at least one prompt including the tool documentation associated with said each computing tool and a request to generate at least one synthetic query configured for processing by said each computing tool. ### Claim 4 (depends on 2) The computer-implemented method of claim 2, wherein providing, by the computing system to the one or more machine-learned sequence processing models for each of the plurality of computing tools, the at least one synthetic query generation request based at least in part on the tool documentation associated with said each computing tool, comprises: providing, to the one or more machine-learned sequence processing models for at least one of the plurality of computing tools, a plurality of synthetic query generation requests; and varying a temperature of the one or more machine-learned sequence processing models for at least one of the plurality of synthetic query generation requests. ### Claim 5 (depends on 1) The computer-implemented method of claim 1, wherein storing in the embedding database, the plurality of embeddings corresponding to the expanded tool documentation data encodings for the plurality of computing tools, comprises: storing tool documentation for said each computing tool with the at least one synthetic query generated by the one or more machine-learned sequence processing models in response to the tool documentation for said each computing tool. ### Claim 6 (depends on 1) The computer-implemented method of claim 1, further comprising: encoding, into an embedding space, the at least one synthetic query generated by the one or more machine-learned sequence processing models in response to the tool documentation for said each computing tool; wherein storing in the embedding database, the plurality of embeddings corresponding to the expanded tool documentation data encodings for the plurality of computing tools comprises, storing at least one embedding of the at least one synthetic query. ### Claim 7 (depends on 1) The computer-implemented method of claim 1, wherein: the one or more machine-learned sequence processing models includes a first sequence processing model; and the at least one machine-learned sequence processing model includes the first sequence processing model. ### Claim 8 (depends on 1) The computer-implemented method of claim 1, wherein determining, by the computing system the subset of the plurality of computing tools that are relevant to the particular user query, comprises: performing at least one sparse similarity-based retrieval method to compare the particular user query with the plurality of stored embeddings. ### Claim 9 (depends on 1) The computer-implemented method of claim 1, wherein determining, by the computing system the subset of the plurality of computing tools that are relevant to the particular user query, comprises: performing at least one dense similarity-based retrieval method to compare the particular user query with the plurality of stored embeddings. ### Claim 10 (depends on 1) The computer-implemented method of claim 1, wherein: the one or more machine-learned sequence processing models include a first large language model. ### Claim 11 (depends on 1) The computer-implemented method of claim 1, wherein: the at least one machine-learned sequence processing model includes a first large language model. ### Claim 12 A computer-implemented method for improving efficiency of computing tool retrieval for a sequence processing model, comprising: providing, by a computing system to one or more machine-learned sequence processing models for each of a plurality of computing tools, at least one synthetic query generation request based at least in part on tool documentation associated with said each computing tool; obtaining, by the computing system from the one or more machine-learned sequence processing models for each of the plurality of computing tools, at least one synthetic query that is configured for processing by said each computing tool of the plurality of computing tools, wherein the at least one synthetic query represents a potential use case for said each computing tool; storing, by the computing system in an embedding database, a plurality of embeddings corresponding to an expanded tool documentation data encoding for the plurality of computing tools, the expanded tool documentation data encoding comprising an encoding of the at least one synthetic query into a multidimensional embedding space; and determining, by the computing system, a subset of the plurality of computing tools that are relevant to a particular user query by comparing an embedding of the particular user query to the plurality of stored embeddings, wherein the subset includes less than all of the plurality of computing tools. ### Claim 13 (depends on 12) The computer-implemented method of claim 12, wherein determining, by the computing system, the subset of the plurality of computing tools that are relevant to the particular user query, comprises: generating, by the computing system, at least one prompt for at least one machine-learned sequence processing model, the at least one prompt including the particular user query and a processing result from each of the subset of the plurality of computing tools in response to the particular user query; and generating, by the computing system, a response to the particular user query based at least in part on an output of the at least one machine-learned sequence processing model in response to the at least one prompt. ### Claim 14 A computing system, comprising: one or more processors; and one or more computer-readable storage media that store instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, the operations comprising: for each of a plurality of computing tools, creating an expanded tool documentation data encoding by encoding, via one or more machine-learned sequence processing models, at least one synthetic query into a multidimensional embedding space based on tool documentation for said each computing tool, wherein the at least one synthetic query represents a potential use case for said each computing tool; storing in an embedding database, a plurality of embeddings corresponding to the expanded tool documentation encodings for the plurality of computing tools; receiving a particular user query; determining a subset of the of the plurality of computing tools that are relevant to the particular user query by comparing an embedding of the particular user query to the plurality of stored embeddings, where the subset includes less than all of the plurality of computing tools; generating at least one prompt for at least one machine-learned sequence processing model, the at least one prompt including the particular user query and a processing result from each of the subset of the plurality of computing tools in response to the particular user query; and generating a response to the particular user query based at least in part on an output of the at least one machine-learned sequence processing model in response to the at least one prompt. ### Claim 15 (depends on 14) The computing system of claim 14, wherein the operations comprise: providing, to the one or more machine-learned sequence processing models for each of the plurality of computing tools, at least one synthetic query generation request based at least in part on tool documentation associated with said each computing tool; and obtaining, from the one or more machine-learned sequence processing models for each computing tool, at least one synthetic query configured for processing by said each computing tool. ### Claim 16 (depends on 15) The computing system of claim 15, wherein providing, to the one or more machine-learned sequence processing models for each of the plurality of computing tools, the at least one synthetic query generation request comprises: providing at least one prompt to the one or more machine-learned sequence processing models, the at least one prompt including the tool documentation associated with said each computing tool and a request to generate at least one synthetic query configured for processing by said each computing tool. ### Claim 17 (depends on 15) The computing system of claim 15, wherein providing, to the one or more machine-learned sequence processing models for each of the plurality of computing tools, the at least one synthetic query generation request based at least in part on the tool documentation associated with said each computing tool, comprises: providing, to the one or more machine-learned sequence processing models for at least one of the plurality of computing tools, a plurality of synthetic query generation requests; and varying a temperature of the one or more machine-learned sequence processing models for at least two of the plurality of synthetic query generation requests. ### Claim 18 (depends on 14) The computing system of claim 14, wherein storing in the embedding database, the plurality of embeddings corresponding to the expanded tool documentation data encodings for the plurality of computing tools, comprises: storing tool documentation for said each computing tool with the at least one synthetic query generated by the one or more machine-learned sequence processing models in response to the tool documentation for said each computing tool. ### Claim 19 (depends on 14) The computing system of claim 14, further comprising: encoding, into an embedding space, the at least one synthetic query generated by the one or more machine-learned sequence processing models in response to the tool documentation for said each computing tool; wherein storing in the embedding database, the plurality of embeddings corresponding to the expanded tool documentation data encodings for the plurality of computing tools comprises, storing at least one embedding of the at least one synthetic query.