Abstract: A system comprising a computer-readable storage medium storing at least one program and a computer-implemented method for providing personalized content sharing service is presented. Consistent with some embodiments, the method may include receiving a request to share a content item with a member of the social network. The method may further include determining an interest of the member of the social network based on social data about the member of the social network. A customized posting to share the content item with the member of the social network is then generated using the interest of the user.

Abstract: Systems and methods are provided for identifying a first user account based on an indication of purchase of a first item; identifying a second user account associated with the first user account; determining a shared characteristic between the first item of the first user account and a second item in an online container associated with the second user account; determining a user interest of the second item based on user attribute data associated with the second user account and the shared characteristic; removing the second item from the online container based on the determining of the user interest; and causing display of a notification on a user interface of a device associated with the second user account, the notification including an indication of the second item being removed from the online container based on the indication of purchase of the first item.

Abstract: Techniques are described, as implemented by computing devices, to control access to digital content through use of nonfungible tokens (NFTs). This is performed by leveraging a blockchain such that digital content associated with an item is made available to supplement use of the item (e.g., to supplement use of a physical item, digital content, and so forth) or make other functionality available based on a user's possession of the item.

Abstract: Some aspects relate to technologies for generating and validating content providing summary information for parts on a listing platform, such as compatibility information and item characteristic information for the parts. In some aspects, input data comprising attribute-value pairs based on one or more items is accessed for content generation. The input data and one or more pre-defined generation prompts are provided to a first generative model, which generates content comprising natural language text as output. In some aspects, the content is validated by providing the content and one or more pre-defined validation prompts to a second generative model. If approved, the content can be provided for presentation on a user device. If rejected, the content is not provided for presentation. In further aspects, the rejected content is used by a third generative model for tuning the pre-defined generation prompts.

Abstract: Automatically updating cryptographic keys using a learning model is described. A computing device receives a notification from another computing device to update at least one cryptographic key corresponding to a cryptographic key pair (e.g., a public cryptographic key and/or a private cryptographic key) associated with an exchange of data between the computing device and the other computing device. The notification includes first criterion associated with the cryptographic key. The computing device generates an updated cryptographic key by the second computing device based on providing the first criterion and second criterion associated with the cryptographic key as input to a learning model. The computing device exchanges the data between the first computing device and the second computing device based on the updated cryptographic key.

Abstract: A method for constructing a point-in-time global consistent graph is described. The method includes accessing a stream of data from a transactional graph database, receiving, at a scalable persistent computational platform that is separate from the transactional graph database, a graph analytics query that indicates a time attribute value, in response to receiving the graph analytics query, constructing, at the scalable persistent computational platform, a point-in-time graph snapshot based on the stream of data and the time attribute value, and processing, at the scalable persistent computational platform, the graph analytics query with the point-in-time graph snapshot.

Abstract: A federated application programming interface (API) gateway system is described. In one or more examples, an application programming interface (API) request is received from an application executed by a client device and a client gateway module associated with a digital service provider (DSP) system is selected based on the API request from a federated API gateway system. A transformed API request is formed by applying one or more transformations using the selected client gateway module on the API request. An API response is generated by applying the one or more transformations using the selected client gateway module on a DSP API response received from the DSP system responsive to processing the transformed API request. The API response is transmitted for receipt by the application executed by the client device.

Abstract: An apparatus and method for real-time audio processing employs a gaze detection sensor to detect a direction of a user's gaze and output a gaze signal corresponding to the detected direction of the user's gaze. A digital signal processing unit responds to a plurality of signals corresponding to a plurality of sounds received at the apparatus, and the determined direction of gaze to identify a signal of interest from the plurality of signals using the gaze signal. The signal of interest is processed for output to the user. In embodiments, a microphone array provides the plurality of signals. An imaging sensor may work with either the microphone array or the gaze detection sensor to identify the signal of interest.

Abstract: An item sorting system is described. The item sorting system includes at least one bin that holds items to be sorted and a light source that illuminates the items stored in the at least one bin. In implementations, the at least one bin includes walls that direct light, received from the light source via a first surface, for emission via a second surface facing an interior of the at least one bin. The item sorting system further includes a recognition device that identifies the items for use in sorting the items. The item sorting system further includes an end effector that manipulates the items during sorting. In some implementations, the end effector includes tips that apply a gentle vacuum force to individually manipulate items in a manner that allows for efficient sorting without damaging the items.

Abstract: Various aspects of the present disclosure relate to rule-based system management using large language models (LLMs). A rule-based system management device receives a first user input indicating to create a rule in a rule-based system. The rule-based system management device searches rules of the rule-based system based on the first user input. The rule-based system management device forms a prompt for a first LLM based on a result of the search and the first user input. The rule-based system management device generates the rule based on inputting the prompt in the first LLM. The rule-based system management device simulates the rule against historic data of the rule-based system to determine a performance metric associated with the rule. The rule-based system management device adds the rule to the rule-based system based on the performance metric.

Abstract: A system for automatically generating visual representations of documents is described. The system may be provided documents that include information about strategic planning of an organization, such as roadmap documents. Information in the document may be organized according to a template format. The system may chunk and encode content (e.g., text) of the documents to generate a corresponding vector embedding index. A large language model (LLM) may execute to distill sections of the documents, for example, based on categories of information, and the LLM may generate a file based on the vector embedding index and the distilled sections. The file may include a syntax for generating a visual representation of the information. In some examples, a visualization tool may be used to generate the visual representation. The visual representation may be sent for display via a user interface of a computing device.

Abstract: Automated workflow generation techniques are described. In one or more examples, a user interface is provided having at least one option configured to enter plain-language text specifying generation of a workflow to achieve a result and a prompt is generated based on the plain-language text. The prompt references information describing a component repository of reusable workflow modules for executing tasks, one or more instructions for creating the workflow, and usage criteria for the reusable workflow modules based on the plain-language text. A workflow is received from a language model as describing a logical order and dependencies between a set of the reusable workflow modules.

Abstract: Swipe-based personalized style generation is described. A personalization system presents different instances of digital content along with configurable aspect options in a swipeable user interface. Users provide detailed feedback through swipe gestures or other inputs to indicate interest or disinterest in both the overall item and specific attributes represented by configurable aspect options displayed in the swipeable user interface. Based on this input, the personalization system generates weighted aspects for each configurable option of an item. Using this weighted information, the personalization system creates a comprehensive representation of user preferences and leverages it to generate a personalized style for the user. This personalized style is used to identify and present new instances of digital content tailored to the user's preferences across various digital platforms.

Abstract: Systems and methods for managing an in-person transaction process are provided. In example embodiments, a network system receives, from a first user device, information regarding a service or an item. Using the received information, the network system generates an unpublished listing and associates the unpublished listing with a session identifier. The unpublished listing is temporarily stored to a datastore. The network system then generates a scannable code that provides a uniform resource identifier (URI) associated with the session identifier. The scannable code is presented on the first user device, whereby the scannable code can be scanned by a second user device to access, via the URI, a transaction user interface associated with the session identifier. In response to a request to access the transaction user interface triggered by the second user device scanning the scannable code, the transaction user interface is displayed on the second user device.

Abstract: Image duplicate detection using a learning model is described. A computing device obtains multiple images. The computing device generates respective representations of the images based on providing the images as input to a learning model, which can be stored at the local data storage of the computing device. In some examples, the computing device receives one or more parameters representative of the learning model from a server device. The computing device outputs an indication of one or more images based on the respective representations of the images matching at least one representation of the representations of the other images within a threshold value. The representations of the other images are stored at the local data storage of the computing device.

Abstract: An artificial-intelligence (AI)-based system for generating search results is described. The system may receive a search query from a user device, the search query associated with an item on an online marketplace. The system may obtain a set of search results including one or more items that correspond to the search query based on the search query. The set of search results may also be provided to a large language model (LLM). The system may receive a prompt for the LLM from the LLM based on the set of search results. Based on the prompt, the system may execute the LLM to generate a subset of the one or more items based on information associated with the items and data associated with the online marketplace.

Abstract: Systems and methods for multi-signal fault analysis are described. The system receives signal message information, over a network, from a collection device comprised of a plurality of mobile devices including a first mobile device and a second mobile device that are associated with a first user. The first signal message information includes a first maintenance message including characterization information that was received by the first mobile device from a component that includes a sensor that operates to sense a first part that is assembled into a vehicle. The second signal message information includes signal information that was received by the second mobile device including an audio signal that a microphone in the second mobile device sensed. The system analyzes the multi-signal information to diagnose a problem and communicate a message to the first user with a diagnosis of the problem.

Abstract: Some aspects relate to technologies for employing metadata identifiers for part item listings on a listing platform to generate inventory intelligence. In accordance with some aspects, textual analysis of listing data for each of a plurality of part item listings on a listing platform is performed to assign a metadata identifier to each part item listing. The metadata identifier for each part item listing comprises a combination of an inventory segment identifier, a category identifier, and fitment data. For a first metadata identifier, a set of part item listings having the first metadata identifier is identified. Demand data and/or supply data is aggregated for the set of part item listings to provide aggregated data. A user interface with one or more metrics is generated using the aggregated data, and the user interface is communicated over a network to a computing device for presentation.

Abstract: Various embodiments described herein assist in identifying one or more product items of interest (e.g., high demand product items) based on surge activity, such as user buying surges, user selling surges, and user product listing (e.g., electronic or online listing) surges on an online marketplace system, and can further notify one or more users of selling opportunities of the identified products items of interest.

Abstract: Artificial intelligence (AI)-based evaluation of a seller-contested dispute is described. A system may receive user input from a first user (e.g., a seller) indicating information related to a transaction dispute on an online marketplace that the seller is contesting. The information includes image and text evidence provided by the first user. A vision language model (VLM) capable of processing image and text data may generate a summary of the information, including a text description of the image evidence. A prompt may be generated based on the summary of the information, the prompt instructing a large language model (LLM) to evaluate the relevance of the evidence provided by the first user. Based on the prompt, the LLM may evaluate the relevance of the evidence and output a relevance score indicative of how relevant the evidence is to the first's user's dispute defense.

Abstract: A machine learning (“ML”) pipeline that includes unsupervised learning, supervised learning, and Bayesian learning is utilized to train a ML classifier that can classify machine metrics as being indicative of an anomaly. A boosting process can be utilized during the unsupervised learning portion of the ML pipeline that scores clusters of training data for completeness, and further splits clusters of training data based upon the completeness scores in order to optimize the clustering of the training data. Supervised learning is then performed on the cluster-labeled training data. Bayesian learning can also be utilized to assign incident probability inferences to the clusters of training data. Once the ML classifier has been trained, the ML classifier can be utilized in a production environment to classify multi-dimensional machine metrics generated by computing devices in the production environment as being indicative of an anomaly.

Abstract: Detecting compatibility mismatch by generative artificial intelligence is described. Compatibility data is obtained (e.g., by accessing a database). The compatibility data is associated with a compatibility between items (e.g., items and categories of vehicles or an item and another item) and includes a list of recommended compatibilities between the items and a user reported compatibility for at least one item. A machine learning model is generated for detecting a compatibility mismatch between a first item and a second item and/or between an item and a category of vehicle. At least a portion of the compatibility data is provided as input to generative artificial intelligence to generate the machine learning model. An update to the list of recommended compatibilities is determined based on the detected compatibility mismatch.

Abstract: An automated system for allocating traffic for content delivery networks (CDNs) based on data is described. The system may receive performance and availability data associated with CDNs from one or more data sources. The data sources may include CDN providers, synthetic monitoring platforms, real user monitoring (RUM) platforms, and other data sources. The system may execute a model or a steering logic for allocating traffic to one or more CDNs based on the data. Using on the model, the system may determine or calculate metrics associated with at least one of global availability, regional availability, or regional performance of the CDNs, and other metrics based on which data sources are applicable. The system may allocate traffic to the one or more CDNs based on the metrics. Allocating the traffic may include costing-out (e.g., directing traffic away from) or costing-in (e.g., directing traffic to) CDNs.

Abstract: In the implementation of techniques for generating listing insights with artificial intelligence, a system receives item data corresponding to an item. The system extracts item attributes from the item data. Based on the item attributes, the system generates an item embedding representing the item attributes. Based on the item embedding, the system generates relevance scores for item listings, wherein each relevance score represents a relevance between the item embedding and item listing embeddings corresponding to the item listings. The system initiates retrieval of item listing data including attribute data of relevant item listings based on each of the relevant item listings having a relevance score exceeding a threshold amount. Based on the attribute data of the relevant item listings, the system generates a listing insight for the item. The system broadcasts the listing insight for display.

Abstract: Grid filter for images of multiple items is described. A computing device obtains a selection of a grid filter including a set of segments of a grid filter including a set of segments to apply to an image of a set of items. The set of items is associated with a same item category. The method segments the image of the set of items into respective image segments of the set of items. Respective items of the set of items align with respective segments of the set of segments. The computing device generates at least one item listing for the set of items, where the at least one item listing includes at least one image segment of the respective image segments.

Abstract: A multimodal embedding modifier generates a modified seed search selection embedding for providing a set of search results. The multimodal embedding modifier enhances the ability and accuracy of identifying a user's true intent when searching the online marketplace. For example, embodiments disclosed herein can allow a user to navigate multiple modalities for an item. In some embodiments, a user may select a search result corresponding to an initial search query, and further modify the selected search result by inputting a modifier (e.g., a textual modifier). The multimodal embedding modifier can be trained using a training dataset including a text embedding, an image embedding, another type of embedding, or a combination thereof.

Abstract: Methods, systems, and computer storage media for providing brand-focused machine learning model training in an item listing system are described. The brand-focused machine learning model training engine supports training a brand-focused machine learning model that predicts brands for item listings that do not include brand information. The training can be based on novel training techniques and training features from data associated with the multi-dimensional authenticity analysis dataset (i.e., brand protection and verification data and item listing system data) to cause generation of a brand-focused machine learning model that is subsequently deployed. In operation, a multi-dimensional authenticity analysis dataset associated with a plurality of brands is accessed. A brand-focused machine learning model using the multi-dimensional authenticity analysis dataset is trained. Training the brand-focused machine learning model is based on brand multi-dimensional authenticity features.

Abstract: The technology described herein relates to systems, methods, and computer storage media, among other things, for generating recommendations corresponding to whether requests (e.g., transmitted by computing devices) are malicious network traffic. For example, the recommendations can be generated using various historical and current network traffic trends (e.g., associated with a particular application programming interface). As another example, the recommendations can be generated based on particular models. Based on the generated recommendation, dynamic rate limiting rules can be applied for determining whether a request is malicious network traffic. Based on determining the request is malicious network traffic, the request (and additional requests associated with that particular request) can be blocked.

Abstract: Query dependent threshold generation for search result filtering is described. In one or more implementations, a user query entered via a search platform is received, and in response, items are retrieved from a storage device based on the user query. Using a first machine learning model, relevance scores are generated for the items, and the relevance scores represent degrees of relevance of respective items with respect to the user query. Using a second machine learning model, a relevance threshold is generated for the user query based on one or more features of the user query. The items are filtered based on the relevance scores and the relevance threshold, and the filtered items are communicated over a network for display in a user interface of the search platform.

Abstract: Methods, systems and arrangements are described herein for enabling viewers to engage with live media broadcasts in ways that provide rich interactive experiences. Methods, systems and arrangements are also described for enabling transactions to be integrated with live broadcasts, and for enabling viewers to experience the feeling of being part of a group. Methods, systems and arrangements are also described for integrating viewer interactions into live media broadcasts themselves.

Abstract: Technology for the improved processing of search queries is provided. In one embodiment, methods may return semantically relevant search results for a search query. During a pre-computing offline processing, an inventory semantic index may be generated and may include inventory binary hashing signatures that are associated with inventory listings, such as goods or services for sell, and the index may be partitioned by categories and shards. When a search query is received, relevant categories are determined using a relevant category recognition service, and a search query binary hashing signature maybe generated for the search query. The relevant categories are searched to determine hamming distances between the inventory binary hashing signatures and the search query binary hashing signature, where the hamming distance indicates semantic relevance.

Abstract: Bitmask encoding-based personalized style generation is described. A personalization system presents different instances of digital content in a user interface and prompts a user to indicate interest or disinterest in individual digital content items. Based on the input indicating interest or disinterest, a bitmask encoding is generated and returned to the personalization system. The personalization system queries a database to identify respective instances of digital content that correspond to each binary value included in the bitmask encoding, which is then used to generate a prompt that causes a machine learning system to generate a personalized style for the user from which the favorable or unfavorable indications were received. The machine learning system generates a personalized style for an individual and identifies different instances of digital content of interest to the individual, and the personalized style is used to generate a user interface for the individual.

Abstract: Generating an access point certificate that defines relationships between an access point and at least one domain is described. Relationship data describing how data is to be routed between an access point and domains is received by a certificate management system. The certificate management system generates a certificate is generated that individually lists each domain associated with the access point and includes information describing data routing for the domain via the access point. The certificate excludes wildcard entries that represent multiple domains via a single entry. The certificate is used to control data communication traffic via the access point and is updated responsive to changes in in domain relationship data for the access point.

Abstract: AI-driven techniques are provided for generating multi-modal vector representations that incorporate enriched, model-generated category embeddings to enable conceptual search of item listings. For example, the enhanced conceptualized search can include generating a search vector, using one or more generative artificial intelligence (AI) models, based on a received search query. Some embodiments of generating the search vector may include applying one or more generative AI models to one or more sets of prior user interactions. Based on the search vector, a particular item listing vector can be determined. The item listing vector can be generated by applying one or more generative AI models to each of an image of the item and a textual description of an item listing, and the item listing vector can be generated based on enhanced categories determined (e.g., using one or more generative AI models) for the item listing.

Abstract: A housing includes an attachment mechanism, a light source including at least one light distributed around a first surface, and a spacer extending from the first surface by a first distance to maintain a fixed focal distance between a camera of a mobile device and an object to be imaged, the spacer forming at least a partially enclosed light shield that blocks external light.

Abstract: Music generation for item listings for an online marketplace is described. A user (e.g., seller) may list an item for sale on an online marketplace. The item listing may include information about the item, including a title, a description, and one or more images, among other information. The information may be input into a large language model (LLM), which may be trained on a set of music data and a set of inventory categories of the online marketplace. The LLM may execute to generate a music sample that corresponds to the item. The music sample may be applied to the item listing standalone, or as background music for a video corresponding to the item and included in the item listing.

Abstract: Automated image generation and optimization is described. In one or more implementations, a sample of listing titles is extracted from listings maintained by the online marketplace within a particular category. The extracted listing titles are simplified using at least one large language model (LLM) to remove extraneous information for image generation. An image prompt is generated using the LLM based on the simplified listing titles and the particular category. One or more images are generated using a large vision model based on the image prompt. The generated images are scored using a vision language model based on predefined criteria. The image prompt is iteratively refined based on the scoring, and images are regenerated until at least one image meets a threshold score for the predefined criteria.

Abstract: Some aspects relate to technologies for performing attribute-aware vector searches using adjusted similarity thresholds. In accordance with some aspects, a search query is received. Items are mapped onto an extended embedding space based on encoded attributes for the plurality of items. The search query is also mapped onto the extended embedding space based on encoded attributes for the search query. An adjusted similarity threshold is generated based on a number of attributes to be matched for the search query. Using the adjusted similarity threshold, items that are similar to the search query in the extended embedding space are retrieved. Search results based on the one or more retrieved items can be provided as a response to the search query.

Abstract: Some aspects relate to technologies for performing item retrieval on a listing platform using clusters of interchangeable parts formed using fitment data. In accordance with some aspects, item data is accessed for each of a plurality of part item listings on a listing platform, where the item data for each part item listing includes fitment data. An item embedding is generated for each part item listing using the item data. The item embeddings are clustered to generate a plurality of clusters, wherein each cluster includes one or more item embeddings. Cluster data is stored for the plurality of clusters. The cluster data for each cluster associates a cluster identifier and an item listing identifier for each item embedding in the cluster. The cluster data can be leveraged to perform item retrieval for the listing platform.

Abstract: In various example embodiments, a system and method for data mesh-based environmental augmentation are presented. Attribute data associated with a user may be received from a plurality of attribute sources. A portion of the attribute data may include real-time data. A portion of the real-time data indicative of an identity of the user may be identified. The identity of the user may be authenticated with respect to the real-time data by analyzing the identified portion of the real-time data. Based on the authentication of the identity of the user, a user activity being performed by the user may be identified based on the real-time data, and the user activity may be augmented according to a user setting.

Abstract: A visual search recommendation engine utilizes an image quality indication model for a visual search recommendation. Specifically, the visual search recommendation engine receives a search image as a search query at a search engine. The visual search recommendation engine provides the search image as an input into the image quality indication model, which is trained to output an image quality indication based on image aspects of the search image. A plurality of images are identified from an image corpus. The visual search recommendation engine determines an image similarity based on a comparison between the plurality of images from the image corpus and the search image. The image quality indication and the image similarity indicate a search query performance for the search image. A first image exceeding the search query performance is identified. The first image is provided for display at the search engine.

Abstract: Methods, systems, and computer programs are presented for replacing dialogue in a video segment. One method includes operations for analyzing content of a video to extract dialogue and meaning in the video, and for determining fragments of the video to be modified based on the extracted dialogue and meaning. The modification is based on factors comprising regional differences, cultural sensitivity, and inappropriate content. For each fragment to be modified, the following operations are performed: generate replacement speech based on the regional differences, cultural sensitivity, and inappropriate content; and generate audio for the replacement speech. The generation of the audio comprises synchronizing the audio with the video to align audio with lip movements while maintaining an emotional tone and a voice profile of each speaker in the video. Further, the method includes an operation for causing presentation on a computer display of the video with the modified one or more fragments.

Abstract: A large language model (LLM) based personalized chat quiz is described. In one or more implementations, information associated with an online marketplace is clustered into a plurality of topics. A plurality of prompts is generated, each prompt causing the LLM to generate a trivia quiz about a respective topic of the plurality of topics. Each prompt is generated by embedding a natural language description of the respective topic into a predefined prompt template. The prompts are input to the LLM, and generated trivia quizzes are received from the LLM. An indication of a customer service wait time is received. A generated trivia quiz about a personalized topic for the user based on tracked information about the user is selected from the generated trivia quizzes. The generated trivia quiz is embedded into the chat interface for presentation during the customer service wait time.

Abstract: In an example embodiment, a transaction log index is used to index log records. The transaction log index comprises an in-memory transaction log index and an on-disk transaction log index. The log records are temporarily indexed in the in-memory transaction log index until the storage server can ensure that the corresponding log records are stored without any missing log records in a sequence, as well as possibly based on other correctness and performance considerations.

Abstract: A real-time augmented reality item guide is described. A service platform may receive a digital image depicting at least one item. A plurality of listings for the at least one item on the service platform is identified via at least one machine learning model of the service platform based on the digital image, each of the plurality of listings associated with a respective point in time. An aspect is fetched from each of the plurality of listings for the at least one item on the service platform. Time-dependent augmented reality digital content is generated for the at least one item based on the aspect. The time-dependent augmented reality digital content is rendered as an overlay on the digital image.

Abstract: Automated determination of web page rendering performance techniques and systems are described. In one example, a web page including multiple user-interactive elements that can be selected or navigated via user input is received. A frame rendering system is employed to automatically identify the user-interactive elements and generate scripts to simulate user input at the user-interactive elements. The generated scripts are then used to simulate user input at the user-interactive elements. The scripts may be generated to simulate user interaction via a variety of different device types, and under a variety of different network conditions. During the simulation, a rendering performance score for the web page is generated by monitoring rendering performance relative to a threshold rendering rate.

Abstract: A method for constructing a point-in-time global consistent graph is described. The method includes accessing a stream of data from a transactional graph database, receiving, at a scalable persistent computational platform that is separate from the transactional graph database, a graph analytics query that indicates a time attribute value, in response to receiving the graph analytics query, constructing, at the scalable persistent computational platform, a point-in-time graph snapshot based on the stream of data and the time attribute value, and processing, at the scalable persistent computational platform, the graph analytics query with the point-in-time graph snapshot.

Abstract: Listwise autocomplete ranking with position debias is described. A computing system can receive a user input indicative of a portion of a search query. The computing system can receive a set of candidate queries for an autocomplete list based on the portion of the search query. The computing system can generate a respective ranking score for each candidate query of the set of candidate queries by calculating, for each candidate query of the set of candidate queries, a respective click probability prediction using a position debiased listwise click-through rate (CTR) prediction model, and calculating, for each candidate query of the set of candidate queries, a respective reward per search. The computing system can order the set of candidate queries according to the respective ranking scores. The computing system can output at least one of the candidate queries based on the ordered set of candidate queries.

Abstract: In accordance with the described techniques, a configuration update is received changing configuration parameters of a data stream processing application from a source configuration to a target configuration. In addition, the configuration update specifies incremental update parameters indicating how different partitions of the data stream processing application are to be incrementally updated. The computing device is configured to incrementally update the data stream processing application. During a time interval indicated by the incremental update parameters, for instance, the target configuration is applied to a first partition, and the source configuration is applied to a second partition. Thus, during the time interval, the first partition processes events of a data stream in accordance with the target configuration, and the second partition processes events of the data stream in accordance with the source configuration.