Intent-based organisation of APIs

• US 10,725,836 B2
• Filed: 04/03/2018
• Issued: 07/28/2020
• Est. Priority Date: 04/25/2017
• Status: Active Grant

• Current Assignee: Intento Inc.
• Original Assignee: Intento Inc.
• Inventors: Savenkov, Konstantin, Sapunov, Grigory
• Primary Examiner(s): Wathen, Brian W
• Application Number:
  US16/608,481
• Publication Number:
  US 20200192727A1
• Time in Patent Office: 847 Days
• CPC Class Codes: G06F 16/367 Ontology
• CPC Class Codes: G06F 8/36 Software reuse
• CPC Class Codes: G06F 8/73 Program documentation
• CPC Class Codes: G06F 9/541 via adapters, e.g. between incompatible applications
• CPC Class Codes: G06F 9/547 Remote procedure calls [RPC]; Web services

Abstract

A service platform for routing intent-based API requests to the most relevant APIs, comprising: a data ingestion subsystem, configured to retrieve API schema definitions and context information from external sources and store them in an API schema library; an ontology mining subsystem, configured to identify ontologies for said stored APIs, cluster the identified ontologies hierarchically and label said hierarchies with meaningful labels; an evaluation subsystem, configured to evaluate a set of APIs based on chosen datasets and metrics or by externally provided quality signals; an API dispatcher subsystem, configured to receive an intent-based request and identify the API schema and API Endpoint that match the request; and an API integration subsystem, configured to receive said intent-based request and said matched API schema and API Endpoint and synthesize an integration code.

Claims

1. A service platform for routing intent-based API (Application Programming Interface) requests to the most relevant APIs, comprising:

a data ingestion subsystem, configured to retrieve API schema definitions and context information from external sources and store them in an API schema library;

an ontology mining subsystem, comprising;

an API Ontology Extraction module, configured to build an ontology for a single API;

an API Ontology Hierarchical Clustering module, configured to hierarchically cluster API Ontologies to discover Latent Intents Hierarchies; and

an Intent Ontology Mining module, configured to build ontologies for said discovered intents hierarchies;

an API dispatcher subsystem, configured to receive an intent-based request comprising intent parameters and identify the API schema and API Endpoint that match the request; and

an API integration subsystem, configured to;

receive said intent-based request and said matched API schema and API Endpoint;

use said intent ontology to translate said intent parameters to the matched API schema ontology;

perform the request to the API host;

receive an answer from the API host;

translate the answer back to the intent ontology; and

synthesize an integration code.

2. The service platform of claim 1, wherein said data ingestion subsystem comprises:

an API extraction module, configured to identify API schemas in text documents;

an API Schema Lookup module, configured to checks whether an identified API schema has a corresponding API Schema in the API Schema Library;

an API Schema Updater module, configured to update an API in the API schema library with new data from the text document;

an API Host Discovery Module, configured to discover physical instances of the API Schema addressable via the Internet or Intranet;

an API Endpoints Library storing all known endpoints for discovered APIs;

an API Data Enrichment Module, configured to perform API schema enrichment using of external services;

a data Ingestion Monitoring module, configured to monitor the data ingestion pipeline and produce visualizations, alerts and system checks to detect whether all parts of the pipeline function correctly; and

a Web Server comprising a representational layer of the system.

3. The service platform of claim 2, wherein said API Host Discovery Module comprises:

an API Host Library, storing all known physical instances of the API Schemas;

an API Host Extractor module, configured to extract API host from said text documents; and

an API Host Probe module, configured to probe network hosts to check for the presence of any addressable API.

4. The service platform of claim 2, wherein said API Data Enrichment Module comprises:

an API Schema Context database storing all relevant information related to an API found on external services;

an API Enrichment module, configured to find mentions of the API in other text documents and stores them in the API Schema Context database; and

an API Presentation module, configured to process API Enrichment data from said API Schema Context database to achieve proper presentation of the API in an API Catalog.

5. The service platform of claim 2, wherein said Web Server comprises:

an API Catalog Web UI module comprising a web interface, configured to allow end-users to navigate through a catalog of the APIs; and

an API Catalog Backoffice UI module comprising a web interface, configured to allow system administrators interaction with parts of the process.

6. The service platform of claim 1, wherein said ontology mining subsystem further comprises:

an Ontology Labeling module, configured to label the Intent ontologies with meaningful names.

7. The service platform of claim 1, wherein said Intent Ontology Mining module comprises a General Ontology Building module, configured to build a general ontology for all APIs in some intent, based on clustering of previously extracted API Ontologies.

8. The service platform of claim 1, wherein said API Ontology Extraction module comprises:

a general Language Ontology that describes the general knowledge of the world;

an API Ontology Library storing all extracted API Ontologies; and

an API Ontology Extractor module, configured to construct and update the API Ontology for a particular API, based on the API Schema, General Language Ontology and API Enrichment Data.

9. The service platform of claim 1, further comprising an evaluation subsystem, configured to evaluate a set of APIs based on chosen datasets and metrics or by externally provided quality signals, wherein said evaluation subsystem comprises:

Evaluation Datasets, developed for intent fulfillment evaluation;

Evaluation Result Storage, configured to store evaluation dataset processing results, performance feedback, intermediate and final metrics;

Performance Feedback Collection module, configured to collect performance feedback from intent consumers and store them in the Evaluation Result Storage;

Intent Execution module, configured to run evaluation datasets through selected intents and providers using the API integration subsystem and store all relevant results and technical metrics in the Evaluation Result Storage;

Metrics Calculation module, configured to calculate quality metrics and their statistical significance and store them in the Evaluation Result Storage;

Reference Model Building module, configured to build reference models for APIs; and

Reference Model Storage, configured to store Reference models.

10. A method of routing intent-based API (Application Programming Interface) requests to the most relevant APIs, comprising:

retrieving API schema definitions and context information from external sources and storing them in an API schema library;

building an ontology for a single API;

hierarchically clustering API ontologies to discover latent intent hierarchies and sub-hierarchies;

building ontologies for said discovered hierarchies of intents;

receiving an intent-based request and identifying the API schema and API Endpoint that match the request;

using said intent ontology to translate said intent parameters to the matched API schema ontology;

performing the request to the API host;

receiving answer from the API host;

translating the answer back to the intent ontology; and

synthesizing an integration code.

11. The method of claim 10, wherein said retrieving API schema definitions and context information comprises:

identifying API schemas in text documents;

checking whether an identified API schema has a corresponding API Schema in the API Schema Library;

updating an API in the API schema library with new data from the text document;

discovering physical instances of the API Schema addressable via the Internet or Intranet;

storing all known endpoints for discovered APIs;

performing API schema enrichment using external services; and

monitoring the data ingestion pipeline and producing visualizations, alerts and system checks to detect whether all parts of the pipeline function correctly.

12. The method of claim 10, further comprising building a general ontology for all APIs in some intent, based on clustering of previously extracted API Ontologies.

13. The method of claim 10, wherein said identifying the API schema and API Endpoint that match the request comprises one of a monologue mode and a dialogue mode,

wherein said monologue mode comprises:

interpreting the Intent Parameters within the Intent Ontology and finding one or more API Schema with corresponding endpoints that can fulfill the request given said Intent Parameters; and

ranking said found one or more API Schema with the hosts associated with said retrieved endpoints; and

wherein said dialogue mode comprises identifying missing Intent Parameters.

14. The method of claim 13, wherein said ranking comprises using at least one of evaluation data sets processing results and performance feedback collected from users.

15. The method of claim 13, wherein said ranking comprises using a user'"'"'s bidding strategy.

16. The method of claim 13, further comprising filtering said found one or more API Schema to satisfy additional explicit or implicit constraints.

17. The method of claim 10, wherein said identifying the API schema and API Endpoint that match the request comprises at least one of:

ensembling results of multiple APIs, performing fallback to a next provider in the case of provider unavailability, and harmonizing features of said identified API schema by adding features.

18. The method of claim 10, wherein said synthesizing an integration code is performed in one of a centralized mode and a decentralized mode, wherein in said centralized mode said synthesizing an integration code runs on a cloud integration subsystem communicating bi-directionally with addressable API endpoints and with a system running on a client device, and

wherein said decentralized mode comprises:

invoking an intent method in a system SDK running on a client computer;

building synthesized Integration Code by said system SDK;

translating the intent parameters into parameters of a particular API;

making a direct request to said particular API Endpoint;

receiving an answer; and

translating it back to intent return parameters.

19. The method of claim 10, further comprising evaluating a set of APIs based on chosen datasets and metrics or by externally provided quality signals.

20. One or more computer storage media having stored thereon multiple instructions that, when executed by one or more processors of a computing device, cause the one or more processors to:

retrieve API schema definitions and context information from external sources and store them in an API schema library;

build an ontology for a single API;

hierarchically cluster API ontologies to discover latent intent hierarchies and sub-hierarchies;

build ontologies for said discovered hierarchies of intents;

receive an intent-based request and identifying the API schema and API Endpoint that match the request;

use said intent ontology to translate said intent parameters to the matched API schema ontology;

perform the request to the API host;

receive an answer from the API host;

translate the answer back to the intent ontology; and

synthesize an integration code.