Abstract: The present technology pertains to a organization directory hosted by a synchronized content management system. The corporate directory can provide access to user accounts for all members of the organization to all content items in the organization directory on the respective file systems of the members' client devices. Members can reach any content item at the same path as other members relative to the organization directory root on their respective client device. In some embodiments novel access permissions are granted to maintain path consistency.

Abstract: The present technology enhances workload management of data storage systems by using an internal copy function and/or a zone-append technology. The internal copy function is used, e.g., in merge operations to move data between locations on a disk without using off-disk resources (e.g., processing or memory of a CPU). Zone-append technology uses nameless writes (e.g., write instruction without an assigned destination address on the disk) to combine IO units from different threads to be written to a common zone of a disk (e.g., a shingled magnetic recording (SMR) disk). Sequential addresses are assigned to IO units from different threads based on their order in the write queue, reducing the latency and seek time typically associated with random writes. The zone-append technology, e.g., uses sequential write operations within specified zones, allowing the disk to determine the actual write location and to report post-write logical block addresses (LBAs).

Abstract: A content management system synchronizes content items across client computing systems connected by a network. Each client device has a storage allocation for synchronized shared content items. If the storage allocation for shared content items on a client device is exceeded by the request to add or edit a content item such that it is enlarged, or open a large content item remote to the client device, a client application or the host of content management system selects content items to remove from residence on the client device but keep remotely on content management system. Upon removal of the selected content items, the client application creates shadow items, representing the content item but only containing the metadata of the content item. This creates sufficient space for the initial request to be completed while maintaining user access to all synchronized shared content items.

Abstract: The present technology is directed to providing a multi-root architecture for managing devices associated with their own root namespaces. The system can include backing up multiple computing devices to a single user account such that the content items from each computing device are stored in their own respective root namespace at the user account. With each user account having multiple root namespaces, the system can identify which root namespace is receiving changes, identify which changes are necessary to synchronize the system, and then communicate those changes to the content management system. This process is facilitated by running a synchronization engine for each root namespace, thereby controlling synchronization for each root namespace. Each synchronization engine is able to retain its own parameters of operation, which provides the ability to tailor synchronization modes on each synchronization engine.

Abstract: The present disclosure is directed toward systems, methods, and non-transitory computer readable media for generating content clusters from topic data and focus data, generating content collections from content clusters, storing and restoring desktop scene layouts, and storing and arranging video call scenes. In some embodiments, the disclosed systems generate content clusters based on topic data and focus data associated with content items within a content management system and/or accessed via the internet. The disclosed systems can also generate content collections for a user account of the content management system from the content clusters. In some embodiments, the content scene system can further store and restore desktop scene layouts for arranging application windows presenting content items. Further, the disclosed systems can store and arrange particular desktop scene layouts for video call scenes.

Abstract: This disclosure describes systems that aggregate activity data items for one or more data sources for a user account. The disclosed systems can utilize a large language model to transform the activity data items to synthesized activity items. For example, the disclosed systems can utilize the large language model to transform the activity data items to the synthesized activity items by extracting the contextual content from the activity data items. The disclosed systems can utilize the large language model to process the synthesized activity items to generate an activity meta summary for a specified timeframe. Based on the activity meta summary, the disclosed systems can further identify a subset of synthesized activity items for the specified timeframe. Thereafter, the disclosed systems can provide the subset of synthesized activity items for the specified timeframe for display on a client device associated with the user account.

Abstract: The present disclosure is directed toward systems, methods, and non-transitory computer readable media for generating a model application comprising one or more discrete model functions classified into a function category comprising a sense category, a reason category, or an act category. In some embodiments, the disclosed systems can combine a discrete model function with one or more additional model functions to generate a model application that defines data processing for a customized instance of a large language model. The disclosed systems can surface a model interaction interface comprising selectable application elements to instantiate respective applications of a large language model and modify the model interaction interface to surface a recommended source content item within a source selection window to utilize with the large language model.

Abstract: The present technology provides a solution that can bring organization to the accumulated data and can surface insights derived from an analysis of stored objects. More specifically, the present technology is directed to an object analysis and classification service which is a configurable platform for identifying objects relevant to a topic and providing analysis of the objects relevant to that topic. The object analysis and classification service can include one or more technologies for identifying objects relevant to a topic. The topic can be configured by a user account or can be a predefined topic selected by the user account. The topic can include parameters that can be used by the one or more technologies for identifying objects relevant to a topic and to provide insights defined by the topic.

Abstract: The systems, methods, and non-transitory computer-readable media relates to the intelligent generation and completion of digital documents. For example, the disclosed systems provide a document creation interface for entering inputs to generate and modify digital documents. In some instances, the disclosed systems receive a document generation prompt, generate a digital document (e.g., using a large language model), and provide an indication of the generated digital document within the document creation interface. Moreover, the disclosed systems can further generate a document summary of the digital document and provide the document summary for a recipient device via a document review interface. Additionally, the disclosed systems can further generate a suggested document modification element and modify the digital document in response to a user interaction with the suggested document modification element using the large language model.

Abstract: The present disclosure relates to systems, non-transitory computer-readable media, and methods for generating content-item-specific large language model responses from content items by segmenting a content item and selecting relevant sections of the content item to provide to a large language model to generate a corresponding output. In particular, in one or more embodiments, the disclosed systems can generate a text representation that includes a plurality of text segments each comprising a number of tokens of the text representation. Further, the systems can extract, from the plurality of text segments, segment-specific text embeddings that correspond to respective portions of the text representation of the content item. Additionally, the systems can determine a segment-specific text embedding corresponding to a model output request.

Abstract: A computing system receives a prompt to be provided as input to a large language model. The computing system generates generating an input string to the large language model by appending a plurality of contexts to the prompt. The plurality of contexts defines rules for the large language model to follow when generating the prompt. The plurality of contexts includes a negative context. Based on the prompt and the plurality of contexts, the computing system generates an attention matrix representing relationships between the prompt and the plurality of contexts. The computing system provides the attention matrix to a trained neural network to determine a likelihood that the prompt is malicious. Responsive to determining that the prompt is likely a malicious prompt, the computing system initiates a remedial action.

Abstract: The present disclosure is directed toward systems, methods, and non-transitory computer readable media for generating and maintaining tokenized assets associated with a user account of a content management system. In particular, the disclosed systems can facilitate generating tokenized assets from content items within a content management system. In addition, the disclosed systems can identify tokenized assets associated with a user account within a content management system for facilitating access to gated content. For example, the disclosed systems can generate and provide a digital passport that serves as a centralized location for accessing and managing tokenized assets associated with a user account.

Abstract: A client application sends an application programming interface call to an interface of a content management system. The call specifies one or more content item search criteria. Sending the call causes the content management system to perform a lookup in a content item index to identify at least one content item that satisfies the one or more content item search criteria. Based on sending the call, the client application receives from the content management system a suggestion to attach the at least one content item to a text being displayed by at the computing system.

Abstract: The present disclosure is directed toward systems, methods, and non-transitory computer readable media for providing a creativity web browser for content creation that integrates features of a content management system with internet browsing capabilities. Specifically, the disclosed systems provide a creativity web browser that includes specialized interface elements for restructuring web browsing to focus on content-related context, including a digital canvas, a content collection pane, an x-ray pane, and other elements. The disclosed systems can provide access to browsing across multiple systems or platforms, such as internet locations and/or locations within a content management system, within an integrated web browsing environment.

Abstract: The present disclosure relates to systems, non-transitory computer-readable media, and methods for selecting machine-learning models and hardware environments for executing a task. In particular, in one or more embodiments, the disclosed systems select a designated machine-learning model for executing a task based on workload features of the task and task routing metrics for a plurality of machine-learning models. In addition, in one or more embodiments, the disclosed systems select a designated hardware environment for executing the task based on workload features for the task and task routing metrics for a plurality of hardware environments. In some embodiments, the disclosed systems select a fallback machine-learning model and a fallback hardware environment for executing the task if the designated machine-learning model or designated hardware environment are unavailable. Moreover, in one or more embodiments, the disclosed systems can pause and initiate tasks based on bandwidth availability.

Abstract: A client application sends an application programming interface call to an interface of a content management system. The call specifies one or more content item search criteria. Sending the call causes the content management system to perform a lookup in a content item index to identify at least one content item that satisfies the one or more content item search criteria. Based on sending the call, the client application receives from the content management system a suggestion to attach the at least one content item to a text being displayed by at the computing system.

Abstract: The present disclosure is directed toward systems, methods, and non-transitory computer-readable media for generating responses to multi-order text queries using a context orchestration engine. For example, the disclosed systems generate context-defining query subcomponents from a multi-order text query, where the context-defining query subcomponents indicate contextual data sources pertaining to their respective portions of the multi-order text query. In addition, the disclosed systems provide or transmit the context-defining query subcomponents to a large language model for domain-specific computer code pertaining to each respective context-defining query subcomponent. The disclosed systems can further execute the generated computer code for each context-defining query subcomponent to access indicated contextual data sources for generating component-specific results. The disclosed systems can also generate a multi-order result to the multi-order text query from the component-specific results.

Abstract: One or more embodiments of a document organization system quickly and conveniently provide digital documents to a user on a client device based on a physical object. In particular, the document organization system can receive an image of a physical document and an identifier from a first client device, identify digital documents that match the physical document, and provide the matching digital documents to a second client device, which displays the identifier. In another embodiment, the document organization system allows a user to bind digital documents to a physical object and later recall the digital documents using the physical object. In addition, the document organization system can store and recall the layout arrangement of digital documents on a client device when binding and recalling the digital documents to the physical object.

Abstract: The present disclosure is directed toward systems, methods, and non-transitory computer readable media for generating and providing intelligent insights for video calls and other virtual meetings. In some embodiments, the disclosed systems analyze stored meeting data from past video calls and other virtual meetings to generate intelligent insights for an upcoming video call. The disclosed systems also generate and provide intelligent insights or coaching tools for ongoing video calls. As part of the intelligent coaching tools for ongoing video calls, the disclosed systems can generate predictions for accomplishing target goals for the video calls. Further, the disclosed systems can generate intelligent insights or coaching tools after video calls take place.

Abstract: A collaborative content management system identifies an application installed on one or more client devices that is susceptible to an attack by using the API calls of the application. The collaborative content management system obtains API calls made by the application and derives API call features. The collaborative content management system inputs the API call features into a machine learning model and receives, as output from the model, a determination of whether the set of API calls represents a C2 attack. In some embodiments, the collaborative content management system, responsive to determining that the set of API calls represents a C2 attack, may take a security action.