Abstract: A method by a router component in a multi-tenant on-demand serving infrastructure to route scoring requests to scoring containers. The method includes receiving a scoring request, determining a machine learning application associated with the scoring request, determining whether a router instance for the machine learning application exists, and responsive to a determination that a router instance for the machine learning application does not exist, obtaining a configuration object for the machine learning application and instantiating the router instance for the machine learning application based on the configuration object for the machine learning application. The method further includes invoking the router instance for the machine learning application to route the scoring request associated with the machine learning application to a scoring container that provides scoring functionality for the machine learning application.

Abstract: A multi-tenant system receives a request to generate a first workspace according to at least a subset of tenant data stored in a plurality of data spaces. The system determines a collection of components of the first workspace and determines a respective data and/or metadata source for each component of the collection of components. The respective data and/or metadata source corresponds to a data space of the plurality of data spaces. The system receives, via the plurality of data spaces, data and/or metadata for each component of the collection of components and stores the data and/or metadata of the first workspace to each component. The system also stores, in a multi-tenant junction table, relationships between the first workspace, the collection of components of the first workspace, and respective objects from the plurality of data spaces.

Abstract: Techniques for providing an interactable meeting transcript associated with a meeting via a virtual space of a communication platform are described herein. In at least one example, a user requests to conduct an audio or video meeting associated with a virtual space of the communication platform. The communication platform receives data representative of the meeting, generates a meeting transcript associated with the meeting based on the data, and causes presentation of the meeting transcript via the virtual space. In at least one example, the communication platform receives, from a user of the virtual space, an indication of a selection of a section of the meeting transcript, and causes an input interface to be presented, enabling the user of the virtual space to comment on the section of the meeting transcript. In response to receiving the comment, the communication platform updates a file associated with the meeting to include the comment.

Abstract: A computing device receives user interaction with a portion of a data visualization that is displayed on the computing device. In response to receiving the user interaction, the computing device generates a first prompt that includes parameters specifying a visualization type of the data visualization, a data array corresponding to data marks of the data visualization, and a title of the data visualization. The computing device inputs the first prompt into a large language model (LLM), and obtains from the LLM a text narrative for the portion of the data visualization. The device generates a second prompt in accordance with the text narrative, inputs the second prompt into the LLM, and obtains from the LLM a title for the text narrative. The device generates a self-contained story comprising the text narrative and the title.

Abstract: Techniques are disclosed relating to the protection of secrets within a software development lifecycle. Developers can use an encryption service to encrypt a secret to be used by an application within a package. The secret can be associated with the application, and then encrypted and included in a package that is signed and passed through a software automation pipeline to a data center that hosts the production server for the application. The application executing on the production server can request that the secret be decrypted by a decryption service after package verification. A developer can also specify, in a manifest file, a set of secrets needed for applications executing in the same data center. The manifest file may be passed from the software development environment to the data center, where the specified secrets are created and used by the applications without ever residing or being accessible outside the data center.

Abstract: A machine learning model hosted on a cloud platform may be used to proactively predict if a maintenance procedure should be performed for a vehicle. In some examples, to support the prediction, the machine learning model may be connected to a different cloud platform that includes a customer relationship management (CRM) system and receives data from sensors of the vehicle. As such, the cloud platform with the CRM data may transmit the CRM data and the sensor data of the vehicle to the cloud platform hosting the machine learning model to aid in generating the maintenance procedure predictions. Further, the maintenance procedure predictions may also include the generation of a prediction score associated with a maintenance procedure. In some examples, the prediction score may satisfy a prediction score threshold, thus a notification may be transmitted to a computing device that indicates the maintenance procedure to be performed for the vehicle.

Abstract: A database with virtual partitioning. A computer system used to implement the database receives a request for a database operation to be performed on data stored in a portion of the database, where the data has a group of logical partitions (e.g., tenants of a multi-tenant database), and where the portion of the database is not further physically partitioned. The computer system identifies a virtual partitioning scheme for the database operation, the virtual partitioning scheme defining sub-groups of the group of logical partitions. The computer system then performs the database operation for the sub-groups defined by the virtual partitioning scheme. Multiple virtual partitioning schemes can be active at the same time, and virtual partitioning schemes can be created dynamically in some instances.

Abstract: A semantic query processing system enables data analytics through a metadata-driven architecture. The system processes analytical queries through a layered execution path that maintains consistent logic across system components. For incoming semantic queries, the system coordinates processing between semantic and storage layers instead of, or in addition to, requiring pre-materialized calculations. A gateway service receives semantic queries while a query preparer generates execution plans based on semantic model metadata defining relationships, measures, and dimensions. A query generator transforms these plans into optimized SQL operations, with complex calculations handled through post-processing. The system reduces analytical complexity by eliminating pre-calculation requirements, enables real-time metric computation through coordinated query processing, and maintains semantic consistency through metadata-driven execution.

Abstract: A system generates consistent copies of data stored in a source database in a plurality of target database systems. The system sends a change data stream from the source database system to the target database systems for performing data updates. The system receives timestamps of a latest transactions processed by consumer processes updating the target database system. The system determines a pause timestamp that occurs after the received transactions. The updates to the target database systems are paused once a transaction corresponding to the pause timestamp is processed. The pause timestamp is used to generate a consistent copy of a source database system in a target database system using multiple partitioned change data streams, each partitioned change data stream processing transactions associated with a partition of the source database system.

Abstract: Techniques are described for expediting communications between a first person of an organization associated with a communication platform and a second person not associated with the organization. The first person requests for the communication platform to generate an invitation to communicate with the second person. The first person provides the invitation to the second person directly or via the communication platform. Responsive to receiving an indication that the second person accepts the invitation, the communication platform identifies whether the second person is associated with the communication platform. If the second user is associated with the communication platform, the communication platform modifies an existing user interface associated therewith to enable communications between the first person and the second person.

Abstract: Techniques for masking data based on annotations are discussed herein. A system may receive a request to perform an action and leverage an LLM to assist in performing the requested action. When generating the input data to input to the LLM, the system can use a template to organize the input data. The template may include static data and/or slot(s) which can include a reference to data to input into such slots. The system may retrieve data to input to the slot based on the reference, retrieve annotations that define a classification of the data, and receive a policy that defines which types of data classifications are to be masked. Based on the data classification and the policy, the system can determine whether to mask the data. The system can generate the input data using the template, the data, and/or the mask(s) and input such data into the LLM.

Abstract: Techniques are disclosed relating to index metadata that is usable for accessing multi-level data structures. A computer system may operate a database, including maintaining a set of records having a set of corresponding keys. The computer system may create multi-level data structures that facilitate key range lookups against those records. A given multi-level data structure may store key information indicative of a subset of the corresponding keys. The computer system may create separate index metadata that is usable for accessing the multi-level data structures. The index metadata may specify indications of key information that is stored in the multi-level data structures and locations of the multi-level data structures. The computer system may perform a key range lookup that includes using the index metadata to determine a particular set of the multi-level data structures whose key information corresponds to a key range of the key range lookup.

Abstract: Methods, systems, apparatuses, devices, and computer program products are described. A system may obtain a set of documents associated with a knowledge base for retrieval-augmented generation (RAG). The system may generate multiple representations of the information included in the documents using multiple knowledge extraction pipelines. For example, the system may generate a set of metadata-based vector embeddings based on the documents, a set of knowledge graphs based on the documents, and a set of hierarchical tree representations based on the documents. The system may receive a user query and may retrieve contextual information from the set of vector embeddings, the set of knowledge graphs, and the set of hierarchical tree representations to augment the user query for a large language model (LLM) prompt. The system may input the prompt to the LLM, and the LLM may output a response based on the user query and the contextual information.

Abstract: An application server may receive user input indicating a plurality of provisioning parameters for provisioning resources on a cloud substrate. The application server may transmit, to a first artificial intelligence (AI) model, the plurality of provisioning parameters and a request to generate, based on the plurality of provisioning parameters, provisioning code associated with the cloud substrate. The application server may transmit, to one or more second AI models, the provisioning code generated by the first AI model, the plurality of provisioning parameters, and a request to analyze the provisioning code based on the plurality of provisioning parameters. The application server may update respective reputation values associated with the first AI model and the one or more second AI models based on one or more analysis results associated with output of the one or more second AI models.

Abstract: Methods, systems, and devices for data processing are described. A server host may receive a request to access a web domain associated with a tenant of a multi-tenant cloud platform. The server host may retrieve a digital certificate and an encrypted private key associated with the tenant, where the digital certificate includes a public key associated with the tenant. The server host may obtain a symmetric key that is wrapped using an asymmetric public key associated with the server host. The wrapped symmetric key may be provisioned by a symmetric key distribution service. The server host may transmit a request that includes the encrypted private key and the wrapped symmetric key. The server host may receive a response that includes a cryptographic signature associated with the private key of the tenant. The server host may provide the digital certificate and the cryptographic signature to a client device associated with the request.

Abstract: Techniques are disclosed in which a computer system implements schema changes for existing database objects by maintaining a logical name and a current physical name corresponding to an existing database object, where the logical name points to the current physical name. The system receives a schema change request specifying the logical name and format changes for the existing object. The system performs a schema change operation in response to the request, including: creating a new database object having a new physical name, copying and transforming data from the existing object to the new object according to the format changes, mirroring new writes directed to the logical name to both the existing and new objects, and causing the logical name to point to the new physical name instead of the current physical name after completing the copying. The system responds to subsequent queries, specifying the logical name, from the new object.