Abstract: Disclosed are systems, methods, and non-transitory computer-readable storage media for identifying accounts having shared credentials. In some implementations, a content management system can collect user login context data when a user logs in to or accesses a user account of the content management system. For example, the content management system can collect client device data, client application data, internet protocol (IP) address data, and/or other data from the user's device when the user logs in to the user account. The content management system can analyze the login context data to determine patterns that indicate that the user account login credentials are being shared among multiple users.

Abstract: The present technology pertains to restoring a directory to a state prior to a past synchronization event. A content management system can present an interface including history of synchronization events for a directory, and a user can select to restore a directory to a state prior to any one of the past synchronization events.

Abstract: The present disclosure is directed toward systems, methods, and non-transitory computer readable media for generating and providing coaching insights using a large language model to process coaching prompts. In some embodiments, the disclosed systems generate a coaching prompt from a knowledge graph encoding data from data sources, such as an observation layer and a world state. The disclosed systems also determine a pulse status of a user account to inform a coaching prompt. Additionally, the disclosed systems provide the coaching prompt to a large language model for generating a coaching insight to improve the pulse status.

Abstract: A computer-implemented system and method of sharing files between a link sharer and a link recipient over a network. A folder sharing link is generated in response to a request by a link sharer, where the link provides a link recipient the ability to modify the contents of the folder. In response to receiving an indication that the generated link has been activated by a link recipient, the system either automatically grants modification rights to the folder or requests manual approval from the link sharer to grant modification rights to the link recipient. Once modification rights have been granted, the system adds the shared folder to the link recipient's account within the context of a document management system.

Abstract: The present disclosure relates to methods, systems, and non-transitory computer-readable media for recursively generating extract content items from a content item. Specifically, the disclosed systems can identify a content item and generate a hierarchy of extract content items, recursively adding each layer in the hierarchy by breaking the content of previously generated extract content items down into increasingly primitive segments. The extract content items include lineage metadata that indicates extract lineage generation-by-generation (for each iteration in the recursive process) all the way back to an originating content item. In some instances, the disclosed systems generate a hierarchical extract structure (represents the lineage of extract content items through generations), provides the structure for display on a client device, shares the structure with one or more user accounts, and/or provides the structure to one or more generative models to utilize in answering user queries.

Abstract: This disclosure describes systems that evaluating generative queries and modifying generative models (e.g., large language models or LLMs) and/or their system prompts based on the evaluation. The disclosed systems can generate and maintain a generative model functionality codex that stores tags for prompts and responses (to capture behaviors and intents of users and LLMs or other generative models). Using the codes, the disclosed systems can update evaluation model generation and can update queries and/or model parameters based on the evaluation. The disclosed systems can further distill models using updated queries and/or learned model parameters.

Abstract: The present disclosure relates to systems, non-transitory computer-readable media, and methods for transforming unstructured data from an application running locally on a client device and indexing the data in a local index for use by a generative model. In particular, in one or more embodiments, the disclosed systems receive unstructured local data from an application running locally on a client device and transform the unstructured local data to formatted local data that matches a data format of a connector for the application. The disclosed systems index the formatted local data in a local index and access the formatted local data for use by a generative model to generate a generative response. In some embodiments, the disclosed systems generate additional contextual data from the application running locally on the client device and index the additional contextual data with the formatted local data for use by the generative model.

Abstract: This disclosure describes systems that implement document editing pipelines to generate document modifications by utilizing a generative model to modify digital documents. The disclosed systems can generate and provide for display, via a client device, a generative document editor that includes document editing pipeline elements that correspond to document editing pipelines that are instantiable to modify composition attributes of digital documents, such as grammar attributes, tone attributes, to add quotes to the digital document, to generate a review simulation of the digital document, and/or to autonomously complete a digital document. Further, based on determining compositional attributes, the disclosed systems can generate and provide for display spectrum slider elements. The disclosed systems can generate document modifications based on receiving indications of interactions with the spectrum slider elements.

Abstract: The present technology improves the ability of a technician to navigate within a data center by making use of lights on servers, switches, and other infrastructure devices to provide navigation queues to a technician. In particular, the present technology can instruct infrastructure devices along a path to a particular infrastructure device that a technician might need to service to operate lights, (e.g., light emitting diodes (LEDs)) in a way that signals a path to the particular infrastructure device. When a number of infrastructure devices are so instructed, an easy-to-follow and understand pattern can be created. The particular infrastructure device can be instructed to operate its lights in a way that differentiates itself from other surrounding devices that might help the technician recognize the device.

Abstract: An example computer implemented method to manage and transfer access to a synchronized content item across multiple users includes displaying a local version of the content item in an application window of a first application. The local version of the content item can be associated with a cloud version of the content item stored on a content management system. The method also includes displaying a content item control interface in the application window with the displayed content item. This content item control interface can be displayed by a second application responsible for synchronizing the local version of the content item with the cloud version of the content item and for managing content item controls through the content management system. The content item level control interface can include a content item editing permission interface that can allow a user to claim or release an editing permission to the content item.

Abstract: A system can selectively synchronize content based on synchronization settings. In some examples, a client stores a local tree representing a local set of content items associated with an account on a content management system, the local tree including respective local nodes corresponding to the local set of content items. The client stores a remote tree representing a remote set of content items associated with the account, the remote set being stored at the content management system and including respective remote nodes corresponding to the remote set of content items. The client receives a synchronization setting disabling local storage of the content item. In response, the client deletes a local copy of the content item, removes a corresponding local node from the local tree, and adds, to a remote node on the remote tree, an attribute indicating that local storage of the content item has been disabled.

Abstract: The present disclosure relates to systems, non-transitory computer-readable media, and methods that utilize a meta-prompting model to generate and store modified system prompts for agent models. For example, the system utilizes an agent model to produce a generated task output from a system prompt corresponding to a task input. The system utilizes a meta-prompting model to compare the generated task output with an expected task output and determine one or more task output differences. The system utilizes the meta-prompting model to generate a modified system prompt for the agent model to produce a task output aligning with the expected task output according to the task output differences. In some embodiments, the system iteratively adjusts the modified system prompt. In some embodiments, the system stores the modified system prompt in a system prompt repository for later use with the agent model for a particular task.

Abstract: The present disclosure relates to systems, non-transitory computer-readable media, and methods for pairing a content management system with a large language model (LLM) to intelligently query groups of related content items. In particular, the disclosed systems can generate content stacks comprising contextually relevant content items. The disclosed systems further utilize an LLM to analyze the content items within the content stack and generate one or more dynamic stack objects (e.g., summaries, timelines, decision logs, or insight views) that synthesize the meaning of content items within a content stack. The disclosed systems can automatically update the dynamic stack objects based on modifications to the content stack. The disclosed systems further generate a content stack data container comprising the content stack and the one or more corresponding dynamic stack objects.

Abstract: A system and method for displaying pinned content in a user interface is described herein. A collaborative content management system displays pins in an order personalized to a user. The collaborative content management system receives a command to output a user interface with a set of pins corresponding to a collaborative content item. For each pin, the collaborative content management system inputs data corresponding to the pin into a machine learning model and receives from the model the probability that the user is interested in the pin. The collaborative content management system ranks the set of pins based on the computed probability and outputs the user interface with at least one pin based on the ranking of the set of pins.

Abstract: The present disclosure relates to systems, non-transitory computer-readable media, and methods for preparing embedding matrices for compression, storage, and reuse. For example, the disclosed systems can receive or generate an embedding matrix comprising a plurality of embedding vectors representing a text content item, the plurality of embedding vectors exhibiting locality in a first dimension of the embedding matrix. The disclosed systems can apply a decorrelating transform along the first dimension of the embedding matrix to generate a transformed embedding matrix by compacting energy distributed across the embedding matrix in the first dimension into a sub-portion of the transformed embedding matrix. Further, the disclosed systems can compress the transformed embedding matrix by performing a compression operation to generate a compressed embedding matrix, and can store the compressed embedding matrix for the text content item.

Abstract: The present disclosure relates to systems, non-transitory computer-readable media, and methods for generating a synthesized connector that defines a single data ingestion path with shared access control for multiple users within a group or organization. In particular, the disclosed systems can identify a set of content items retrieved through individual user connectors associated with user accounts belonging to the same group. Based on the identified content items, the disclosed systems can generate a synthesized connector that unifies the individual connectors in a single, group-scoped ingestion path that governs access control for all associated users. Additionally, the disclosed systems can deduplicate shared content data across the synthesized connectors to generate a synthesized content item. The disclosed systems can update a data repository (e.g.

Abstract: The present disclosure relates to systems, non-transitory computer-readable media, and methods for retrieving digital content from external applications through a software connection. In some embodiments, the disclosed systems identify a software connector comprising a plurality of federated connections linking a content management system and an external application, the plurality of federated connections associated with respective user accounts of the content management system and connecting the respective user accounts to the external application through the software connector. The disclosed systems, use a sync engine to crawl the plurality of federated connections and generate a content graph comprising nodes representing digital content from the external application to sync with the content management system. The disclosed systems request, by the sync engine using the content graph, the digital content from the external application to the content management system through the software connector.

Abstract: The present technology provides a versatile content management system that can also support bi-directional synchronization of an organization account. The organization account is maintained as a plurality of root directories that can be mounted under an organization directory top-level folder. The organization directory top-level folder can appear in a user account's view of the organization account, by does not actually exist as a directory in the content management system. The organization directory top-level folder can be a folder created on a client device in which root directories of an organization account are mounted, but the organization directory top-level folder is not a synchronized directory. The organization directory top-level folder gives the appearance of a single directory structure for an organization even though the organization content is actually organized into the plurality of root directories.

Abstract: Methods, systems, and non-transitory computer readable storage media are disclosed for generating meeting insights based on media data and device input data. In one or more embodiments, the system analyzes media data and inputs to client devices associated with a meeting to determine a portion of the meeting that is relevant for a user. In one or more embodiments, the system generates a meeting summary, meeting highlights, or action items related to the media data to provide to the client device. In one or more embodiments, the system also uses the summary, highlights, or action items to train a machine-learning model for use with future meetings.

Abstract: The present disclosure is directed toward systems, methods, and non-transitory computer readable media for generating and visualizing effectivity scores from features extracted for different data of a time period. The disclosed systems generate a meeting effectivity score, a flow state effectivity score, and/or a work about work effectivity score that reflect measures of effectiveness or productivity. To generate the effectivity scores, the disclosed systems can leverage software connectors that extract features relevant to the respective effectivity scores for inputting into effectivity-score-generating models. Based on the effectivity scores, the disclosed systems can further generate recommendations for improving one or more of the effectivity scores. The disclosed systems can also generate and provide graphical visualizations of effectivity scores for display on a client device, together with recommendations for improving the effectivity scores.