Product Updates

The Cognite Data Fusion (CDF) Bootcamp schedule for the second half of 2026 is now available.This 4-day, instructor-led, in-person training is designed for technical professionals who want hands-on experience building solutions with Cognite Data Fusion. Participants will work through key concepts including data ingestion, contextualization, data modeling, orchestration, and application development through practical exercises.Ready to build your CDF skills?Check the calendar and register for a session that fits your schedule.Open bootcamps are offered in:Oslo, Norway Phoenix, ArizonaWe recommend registering early to allow enough time to complete the required pre-work before the session begins.  Event Name Event Date Location Cognite Data Fusion Bootcamp in Phoenix - July 2026 July 20-23, 2026 9:00a.m.  US/Pacific Phoenix, Arizona Cognite Data Fusion Bootcamp in Phoenix - Aug 2026 Aug 17-20, 2026 9:00a.m. US/Pacific Phoenix, Arizona Cognite Data Fusion Bootcamp in Oslo -  Aug 2026 Aug 24-27, 2026 9:00a.m. Europe Oslo, Norway Cognite Data Fusion Bootcamp in Oslo -  Sep 2026 Sep 7-10, 2026 9:00a.m. Europe Oslo, Norway Cognite Data Fusion Bootcamp in Phoenix - Sep 2026 Sep 21-24, 2026 9:00a.m. US/Pacific Phoenix, Arizona Cognite Data Fusion Bootcamp in Oslo -  Oct 2026 Oct 12-15, 2026 9:00a.m. Europe Oslo, Norway Cognite Data Fusion Bootcamp in Phoenix - Oct 2026 Oct 19-22, 2026 9:00a.m. US/Pacific Phoenix, Arizona Cognite Data Fusion Bootcamp in Oslo -  Nov 2026 Nov 9-12, 2026 9:00a.m. Europe Oslo, Norway Cognite Data Fusion Bootcamp in Phoenix - Nov 2026 Nov 9-12, 2026 9:00a.m. US/Pacific Phoenix, Arizona Cognite Data Fusion Bootcamp in Oslo -  Dec 2026 Dec 14-17, 2026 9:00a.m. Europe Oslo, Norway Cognite Data Fusion Bootcamp in Phoenix - Dec 2026 Dec 14-17, 2026 9:00a.m. US/Pacific Phoenix, Arizona  Confirmed for a Bootcamp?Once your participation is confirmed, be sure to join the Cognite Data Fusion Bootcamp Group on Cognite Hub.The group contains:Preparation resources and pre-work Session updates and practical information A place to ask questions before and during the bootcampStay informed, stay prepared, and get the most out of your bootcamp experience. 

Learning Paths: Build skills step by stepNew learning paths bring related topics together so you can build a solid understanding of key CDF concepts and workflows, progressing from fundamentals to practical implementation.Cognite Flows Foundation Certificate: We’ve recently introduced the Cognite Flows Foundation Certificate learning journey, designed for developers and technical builders working with industrial applications and Cognite Flows.The learning path provides a strong foundation in the Cognite Flows ecosystem, including core concepts, a human-centered approach to application building, engineering architecture, submission processes, and security guidelines — everything needed before getting started with deployment and development in Flows.Who it’s for: Developers, consultants, and solution architectsValue: Build foundational knowledge for industrial application development and Builder Certification preparationDuration: 120 minutesBy completing the learning journey, you will earn the Cognite Flows Foundation community badge and certificate - the official prerequisites for accessing Part 2: Cognite Flows for Builders, which is coming soon. Courses: Focused learning on key CDF capabilitiesNew courses dive deeper into specific CDF features and workflows, helping you strengthen your technical understanding and apply them effectively.Diagram Parsing for Data ModelingLearn how static engineering diagrams such as P&IDs, process flow diagrams, and electrical schematics can be transformed into structured, connected, and contextualized data within CDF. The course focuses on extracting engineering elements from diagrams and mapping them into data models to improve operational visibility and contextualization across industrial use cases in oil & gas and manufacturing.Who it’s for: Data engineers and users with basic understanding of industrial diagrams and processesValue: Understand how engineering diagrams become searchable, connected operational dataDuration: 40 minutes Cognite ToolkitThis course focuses on configuring, deploying, and managing CDF projects using YAML-based configurations and DevOps workflows. Participants will learn how to manage transformations, extraction pipelines, RAW tables, groups, and data models across development, staging, and production environments while gaining a practical understanding of Toolkit modules and deployment workflows.Who it’s for: Data engineers and users with coding experience and foundational CDF knowledgeValue: Simplify and standardize deployment workflows in CDF projectsDuration: 40 minutes Deployment Packs: Ready-to-Use Configurations for CDFWe’ve added a new set of deployment packs designed to simplify and standardize CDF setup. These packs provide pre-configured modules, data models, and environments that help you deploy solutions faster and more consistently across projects. Oil and Gas Data Model — CFIHOS & ISO14224 ExtensionA tag-centric model that combines data from systems such as AVEVA, SAP, OPC UA, and PI into a simplified and queryable structure. Built on CFIHOS 2.0 and ISO14224 standards, the model is optimized for easier querying, dashboards, AI tools, and search experiences through a denormalized approach to industrial data. Configuring the Infield QS Deployment Pack with the ToolkitThis deployment pack simplifies setup of the InField application in CDF by providing predefined environments, permissions, data models, and Toolkit deployment workflows aligned with InField 2.0 best practices. Once deployed, users can access contextualized operational data including assets, activities, notifications, and maintenance information directly within InField. Quickstart Deployment PackA bundled deployment package that includes Entity Matching, File Annotation, the Quickstart Enterprise Data Model, foundational components, and synthetic data for end-to-end testing. It is designed to help teams quickly deploy a complete working environment with all required capabilities included out of the box.  Coming This Quarter Hands on part for builder CertificationTake the next step in your Cognite Flows learning journey by applying your knowledge in a hands-on certification experience. This practical component will allow learners to build a live simulation application based on a real-world industrial use case, demonstrating their ability to design and implement operational applications using Cognite Flows.What you'll earnCognite Flows Builder Certificate  Builder Community BadgeStatus: Enrollment opening soonThis hands-on certification experience is designed for learners who have completed the foundational Builder learning path and are ready to validate their skills through practical application. Master Cognite Data Fusion with new training on Cognite Academy. We’d love to hear your thoughts and help you connect with other experts. Come join the conversation on Cognite Hub. 

The Q2 2026 release accelerates the democratization of Industrial AI, empowering teams with production-grade custom app infrastructure and declarative data workflows. We have significantly increased industrial workforce capacity through precision search migration, zero-code CSV exports, and immersive 360° spatial walkthroughs. To rapidly scale use cases across the enterprise, builders can now leverage lifecycle-managed Atlas AI Skills and native agent execution to deliver a unified, intelligent experience across all Cognite Data Fusion (CDF) applications. Review the core release highlights below to discover how these advancements are accelerating time to value by making Industrial AI more robust, scalable, and accessible for the entire industrial workforce.Cognite Flows - Native and Custom apps Cognite Flows -  Custom Apps Cognite Flows - Adaptive Experiences Cognite Flows - Application Certification Cognite Flows - New Developer Experience Industrial Canvas - Symbol Library Update Industrial Canvas- Unified Project Settings for Canvas and Search  InField - Field-Ready Observations Mobile Lists Search UI - Precision Results Search UI - Download Individual and List of Instances (CSV)  Data Fusion - Integrations, Data Management, Knowledge Graph, and Data Processing & Orchestration Data Workflows - Native JSON Mapping Tasks Function- Access Scoping on Execution Level Function Apps Integrated in Workflows (UI)  Hosted Extractors: Email Alerting API  Integrations - API & UI  Cognite Connector - Simulation Run Load Balancer Functions -  Support for Private Link on GCP Clusters  Data Workflows- References in Workflows  Python SDK - Asynchronous Concurrency 3D - Industrial Street-View Spatial Inspections 3D - Forced CAD Node Rendering Atlas AI Enhancements Atlas AI: Data Analysis (Code Execution)  Atlas AI - Reusable Domain Skills Packages Atlas AI - Query Tool Atlas AI - Unified Agent Experience Across Apps  Agent as Task Type in Workflows  Cognite Flows - Native and Custom apps Feature Problem (Challenge) Solution: What You Can Do Now Cognite Flows -  Custom Apps Production Rapidly scaling use cases across your enterprise Industrial builders previously struggled to rapidly deploy and scale tailored enterprise applications inside CDF without creating security flaws or running up massive operational debt. Subject Matter Experts can leverage a robust, secure, and production-grade environment to deploy custom applications directly inside CDF. Features include an upgraded command-line interface tool and a strict two-step validation safety procedure (deploy and activate) to ensure flawless code-signing. Documentation  Cognite Flows - Adaptive Experiences Private Preview Accelerated time to value with your use cases  Every single system user used to inherit the exact same uninspiring home landing page, leaving critical site-specific knowledge siloed in workers' heads. Access an intelligent landing page that builds a personalized dashboard experience via conversational AI within under 5 minutes from first login. Teams can pin custom widgets to hit an aggressive 50% Daily Active User (DAU) organizational adoption target.Documentation     Cognite Flows - Application Certification Production Rapidly scaling use cases across your enterprise   There was no formal certification process for custom applications. Teams could deploy apps without any structured validation, creating risk around data model integrity, security, and production readiness, with no consistent bar across the builder ecosystem.  Applications undergo a verification process before deployment. Automated validation covers data model integrity, app robustness, security, and performance at scale, while selective human review handles higher-risk scenarios. Certified builders follow a verified path, ensuring every app is safe to operate in industrial environments. Documentation Cognite Flows - New Developer Experience Production Accelerated time to value with your use cases Fragmented tooling and manual UI assembly slowed industrial development. Existing tools were not designed for AI-assisted workflows, making it hard to scale innovation efficiently. A standardized Developer Toolbox with SDKs, CLI, UI components, and machine-readable documentation lets developers and AI agents build consistent, production-grade applications for both Cognite Flows and self-hosted environments. Documentation Industrial Canvas - Symbol Library Update Production Increase industrial workforce capacity  Field operators editing Piping and Instrumentation Diagrams (P&IDs) lacked native tools, bottlenecking the creation of clean digital work packages. Access more than 300 new standard industrial symbols organized cleanly by category for rapid discovery during work package creation. Key workspace updates introduce proportional resizing and persistent color memory to automatically save the last-used shape properties. Documentation   Industrial Canvas- Unified Project Settings for Canvas and Search  Production Accelerated time to value  When visual presentations drift across different tools, users lose confidence that they are viewing the same modeled reality across the platform. Maintain a single, consistent story for your core data model (CDM) content. Cards in Canvas now reflect the exact same properties as Search, matching your Project Settings configuration perfectly. Furthermore, file connections and annotations now display user-friendly names and descriptions instead of confusing system IDs. Documentation InField - Field-Ready Observations Mobile Lists Production Increase industrial workforce capacity Workers in the field could generate new observations on their mobile devices but had no way to track existing entries on the go, resulting in massive duplicate work reports. View, filter, and track nearby location-contextual observations directly on mobile devices. This rollout transitions core field workflows to a unified industrial data model. Documentation       Search UI - Precision Results Production Increase industrial workforce capacity Broad keyword searches historically evaluated entries using a loose "OR" script, flooding engineers with irrelevant records when querying precise tag strings. All core search surfaces across CDF front-ends (Search, Canvas, Charts, 3D, and InField) have migrated behind the scenes to an implicit AND operator. Searching strings like "21-PT-10-19" now demands that all tokens match, improving query precision. Documentation  🔍  23 tv Before:  After:        Search UI - Download Individual and List of Instances (CSV)  Production Accelerated time to value  Domain specialists lacked a simple, zero-code mechanism to extract curated, filtered data models or associated relationships out of the platform for localized evaluation.  Download data lists from Search screens directly into a localized CSV file layout. Extracted tables fully capture user-configured column orders, filtered constraints, sorting layouts, and explicit download timestamps. Documentation     Data Fusion - Integrations, Data Management, Knowledge Graph, and Data Processing & Orchestration Feature Problem (Challenge) Solution: What You Can Do Now Data Workflows - Native JSON Mapping Tasks Private Preview Accelerated time to value Normalizing data structures or reshaping payloads between individual steps in a workflow traditionally demanded heavy, custom-coded external scripts. Evaluate declarative expressions directly against JSON inputs inside the core Data Workflows canvas. Users can apply quick transformations, string filters, and custom aliases natively without leaving the workflow interface. Documentation   Function- Access Scoping on Execution Level Production Reduce the amount and cost of downtime  Execution privileges were historically hard-tied to broad data write permissions, forcing teams to grant overly permissive access roles just to run pre-approved tasks. Grant explicit execution-only privileges through a standalone access level control. This cleanly separates operational execution from data modification roles, protecting your underlying data pipelines.     Function Apps Integrated in Workflows (UI)  Private Preview Accelerated time to value Teams could not utilize versioned Function Apps as a first-class, selectable task type within the Workflows UI canvas, limiting onboarding automation velocity.  Execute Function Apps as native tasks directly within the drag-and-drop Workflows UI editor. Built on top of CDF Functions infrastructure, this packaging model exposes strongly typed parameters, specific path methods, and versioned compute configurations.  Hosted Extractors: Email Alerting API  Public Preview Reduce the amount and cost of downtime  Hosted extractors lacked an integrated notification layer to match on-premises setups, forcing data engineering teams to manually poll pipeline statuses to detect job failures.  Proactively configure automated notification sinks and alert subscriptions for specific extraction jobs via the Signals API. Teams automatically receive email notifications when a hosted REST extractor enters a failure or error state. Documentation  Integrations - API & UI  Private Preview Increase industrial workforce capacity  Traditional extraction pipelines required data engineers to manually access physical VMs to restart service tasks whenever remote configuration files were updated.  Manage remote connectors using a single, unified Integrations surface that automatically pulls configuration adjustments and pushes live statuses without manual restarts. Provides task-level visibility into run history, errors, and system health. Documentation  Cognite Connector - Simulation Run Load Balancer Public Preview Increase industrial workforce capacity  Simulator routines were hard-bound to specific connector pipelines, triggering massive backlog timeouts during peak industrial usage periods. Routes are dynamically queued and picked up by any healthy, available connector. Scale your simulation throughput horizontally by inserting new backend connectors without touching original routine layouts. Documentation Functions -  Support for Private Link on GCP Clusters  Production Reduce the amount and cost of downtime  Enterprise customers utilizing Private Link on Google Cloud Platform (GCP) clusters for strict network isolation were unable to route traffic from Cognite Functions across the hyperscaler backbone.  Configure Cognite Functions traffic routing to natively utilize Private Link architectures on GCP setups. This ensures that all automated computing traffic remains on protected network paths, eliminating public internet interception risks. Data Workflows- References in Workflows  Production Accelerated time to value  Defining dynamic variables inside the data workflows editor was prone to errors, forcing users to type out long, highly specific syntax links manually from memory.  Assemble complex dynamic parameter strings error-free using a visual Reference Picker directly inside the workflow canvas. The UI automatically populates eligible input and output references for Transformation, Function, and Agent tasks. Documentation  Python SDK - Asynchronous Concurrency Production Increase industrial workforce capacity Handling massive batch operations asynchronously inside local workspaces required developers to manually script complex concurrency patterns. Upgrade seamlessly to Python SDK version 8, unlocking fully integrated async client features out of the box. The update optimizes concurrent data downloads and delivers significantly faster file uploads on Windows machines. Documentation 3D - Industrial Street-View Spatial Inspections Production Increase industrial workforce capacity Verticals like discrete manufacturing and pharma often have detailed indoor environments lacking heavy 3D CAD files, making traditional 3D engines feel flat and unnatural.  Walk through plants using an advanced 360° image walkthrough system. The engine introduces depth cursoring over underlying point clouds, ground-level markers to prevent view blockage, and automatic snapping to the closest capture view. Documentation    3D - Forced CAD Node Rendering Private Preview Reduce the amount and cost of downtime  Dynamic viewport engines stripped down fine-grain geometric detailing during wide facility loads, removing structural visibility for field design reviews. Explicitly target up to 100,000 critical structural nodes to remain fully rendered at high detail across all viewport zoom distances. This prevents visual rendering errors during sensitive site construction evaluations.  Documentation   Atlas AI Enhancements Feature Problem (Challenge) Solution: What You Can Do Now Atlas AI: Data Analysis (Code Execution)  Public Preview Rapidly scaling use cases across your enterprise  Context caps limited assistant queries to small, 100-row batches, causing agents to drop analytical history threads mid-conversation. Atlas AI can write, test, and run internal scripts against massive datasets up to 5,000 objects simultaneously. The system remembers the conversational state across iterative questions to build unified statistical trends. Documentation  Atlas AI - Reusable Domain Skills Packages Production Accelerated time to value  Domain knowledge and specific root-cause analysis logic had to be configured by hand inside separate agent prompts, making configurations bloated and patterns un-reusable.  Package specific domain expertise into reusable, lifecycle-managed assets called Atlas Skills. Builders can rapid-assemble virtual assistants by composing existing skills instead of starting from scratch.  Documentation Atlas AI - Query Tool Production Accelerated time to value  Agents searching through extensive graph spaces required deep upfront schema maps, slowing responses and trapping agents on pre-defined paths. Deploy an intelligent query tool that completely auto-discovers graph data models, inspects properties, and reads available views independently without manual pre-configuration. Requires updating agent runtime to V3. Documentation    Atlas AI - Unified Agent Experience Across Apps  Production Increase industrial workforce capacity AI assistants behaved differently across individual apps, forcing users to constantly re-learn chat commands depending on their screen view. Execute configured AI agents defined through Atlas AI as native steps within your data processing pipelines. This allows data teams to feed the direct output of a transformation task into an agent prompt for advanced processing. Documentation  Agent as Task Type in Workflows  Private Preview Rapidly scaling use cases across your enterprise  Currently there is no provision to execute agents from workflow which limits users to operationalize and scale their use cases.  Execute configured AI agents defined through Atlas AI as native steps within your data processing pipelines. This allows data teams to feed the direct output of a transformation task into an agent prompt for advanced processing.   For a comprehensive list of all technical updates, improvements, and bug fixes, view the full Q2 Release Notes. 

We've just released v4.5.0 of the Cognite Data Source for Grafana, with a focus on making Core Data Model (CDM) the default path for new dashboards and bringing operational context (Activities) into your time series charts. This release builds on v4.4.0 and consolidates several capabilities that make Grafana a stronger fit for teams already working in CDM, or planning to migrate.What's new in v4.5.0Activities tab - a new query tab that lets you query CogniteActivity records directly in the editor, with filtering by Asset, Equipment, or TimeSeries. No GraphQL required. Activity annotations - overlay activities as shaded time regions directly on CogniteTimeSeries charts. Useful for visualising maintenance windows, planned downtime, or operational events alongside sensor data. Numeric time series auto-detection in the Data Models (GraphQL) tab - results with type "numeric" automatically fetch datapoints from CDF, removing a manual step in dashboard setup. Config editor overhaul - settings are now reorganised into Connection and Features tabs, CDM features are enabled by default for new datasources, and CDM/legacy toggles are now mutually exclusive (no more conflicting configs).Recap from v4.4.0 (still relevant if you haven't upgraded)CogniteTimeSeries tab - search and select time series directly from CDM with a single View dropdown, including unit conversion support. GraphQL variables - use Data Model GraphQL queries to drive Grafana template variables, enabling dynamic dashboard filtering and selection patterns.How to get startedIf you're on a new datasource, CDM features are on by default. If you're upgrading an existing datasource, head to your Cognite data source configuration, open the Features tab, and enable Core Data Model features.Existing dashboards built on legacy query tabs will continue to run.Documentation: https://docs.cognite.com/cdf/dashboards/guides/grafana/getting_startedWho this is forThis release is especially useful if you're:Migrating an existing project to CDM and need your Grafana dashboards to come along Building new dashboards and want to start from CDM rather than legacy asset-centric tabs Looking to overlay operational context (maintenance, downtime, events) on your sensor charts without leaving Grafana Using GraphQL and modeled context to drive dashboard variables and filteringWhat we'd like to hear from youWe're shipping CDM support iteratively and learning as we go. If you try v4.5.0, we'd value feedback on:How the Activities tab and annotations fit into your existing dashboards Anything that feels missing compared to your legacy workflows Friction in the new config editor or the CDM/legacy switch Use cases where GraphQL variables would unlock something you can't do today

Have you seen it? In our last release, we added support for File Upload through the Search application. The goal is to remove a hindrance to daily use by allowing single-file uploads for non-technical users of Cognite Data Fusion.While most data in Cognite Data Fusion is ingested through structured pipelines, there was a need to enable individual Subject Matter Experts (SMEs) to upload data directly. This allows non-technical users, such as Reliability Engineers, to add valuable information for future reference.As of earlier this year, we have support for file uploads through the Search application and for using Search as a resource selector (e.g., in Canvas or Charts). This is available for Data Modeling Files. We ensured to add governance around file uploads: Only available through use of Location(s) The Location needs to have File Upload space added to the Location configThe individual user will see the File Upload option for Files if the location they have selected has a File Upload space and they themselves have permission to write data to that spaceThe individual user will see Files uploaded by other users if they have permission to read from the File Upload spaceWe covered this feature in the March’26 release notes:  Upload files directly in Search. The upload option is available after your admin configures the space for file uploads in Admin > Project settings > Locations > Configure > Space for file upload. Only one space can be set per location. Once you upload files, they may take a moment to appear. Upload is available only for data modeling views. Let me know if you have questions regarding File Upload - we would love to see it in use!Best, Sofie, Product Manager 

We’re deprecating the totalVariation aggregate function for numeric time series in Cognite Data Fusion, effective April 20, 2027. This follows Cognite’s 12-month deprecation policy for GA features: between now and that date the aggregate will continue to function as documented, and after it, requests that include totalVariation will return an error. The change applies to requests to the time series data list endpoint (/timeseries/data/list).If your integrations rely on totalVariation, directly via the API, through an SDK, or in dashboards and derived pipelines, please review them before April 20, 2027 and migrate to another aggregate.If totalVariation is important to a workflow in your organization and none of the alternative aggregates fit, reply to this thread or reach out to your Cognite representative. It helps us most if you can share:The use case and what the output drives (dashboard, alert, calculation, report). The time series volume and aggregation granularity involved. Whether computing the value client-side from raw data points is feasible for your setup.

Today, we released a ton of improvements to the Cognite Toolkit. Here is what you should know about the fresh v0.8 version: Better performance: Substantial refactoring has led to faster execution, especially for build, deploy, and data transfer commands. Enhanced Deployment: The deploy command is now forward-compatible with the API, ensuring proper property handling. Clearer Validation: Output is now easier to read and provides richer guidance for both you and your agents.  NEAT Intelligence: You’ll notice more recommendations and auto-fixes for Data Modeling performance and scalability, powered by built-in NEAT logic. Module Scaling: Reusing modules is easier with the module repeat feature, allowing you to deploy a module multiple times using site-specific variables. New Plugins: To streamline the experience, we’ve moved development features to cdf dev ... and data features to cdf data ....Other nice-to-knows:Auth breaking changes: In cdf auth init, we have removed the --no-verify and --dry-run flags. Command Updates: cdf dump asset/timeseries has been retired. Use cdf data download assets/timeseries instead. CSV Handling: You can no longer build/deploy assets in .csv format; please use cdf data upload for those files. GA Status: Support for Streams and Simulators is now officially GA.We continue to release improvements and new resource types frequently. Most of these features have been available behind alpha flags for a while - thank you to our eager users for the constant feedback!

Learning Paths: Build skills step by stepNew learning paths bring related topics together so you can build a solid understanding of key CDF concepts and workflows, progressing from fundamentals to practical implementation.Data Modeling: Schema Design and Structure: Learn how to create Containers, Views, and Data Models. This learning path focuses on the schema design and structure of Data Modeling in CDF. It covers the foundational concepts - how to design Containers and Views, structure your model, and apply basic naming and governance practices before writing code or ingesting data. This is a fundamental-level path for users who are new to Data Modeling but already have some familiarity with Cognite Data Fusion. Data Modeling: knowledge graph creation: Understand how to load data into an existing data model in CDF. This learning path focuses on building a Knowledge Graph by ingesting nodes and edges, covering common data sources and the tools used to populate a schema.This is a foundational path for users new to Data Modeling who already have some familiarity with Cognite Data Fusion. Courses: Focused learning on key CDF capabilitiesNew courses dive deeper into specific CDF features and workflows, helping you strengthen your technical understanding and apply them effectively.An Overview of Cognite ExtractorsUnderstand the role of extractors in Cognite Data Fusion. This course covers extractor architecture, types, security considerations, and deployment approaches used for industrial data ingestion.You’ll learn what extractors are, why they are critical for bringing industrial data into CDF, and how they work at a conceptual level. The course builds a technical foundation that prepares you for later courses focused on configuration, deployment, and operations.Mastering Extraction Pipelines in Cognite Data FusionLearn how to build, secure, and monitor reliable data integrations in CDF. This course focuses on designing extraction pipelines that provide visibility into data ingestion and help prevent silent failures.You’ll explore how extraction pipelines work, how to separate extraction logic from monitoring, and how to implement mechanisms such as heartbeat monitoring, remote configuration, and secure authentication using OIDC service principals. By the end of the course, you’ll understand how to move from basic extraction scripts to more robust, observable data pipelines. Microlearning videos: Small lessons, big impactWe’ve added a new set of bite-sized microlearning videos focused on simulator integrations in CDF. These short lessons cover key concepts and practical steps, making it easier to learn at your own pace.Activities in charts You can now add Activities to your graph view in Charts! Activities can be shown together with the time series data.Data workflows from the UI: Learn the fundamentals of creating and managing Data Workflows directly in the CDF user interface. This microlearning shows how to create, run, and monitor workflow tasks and process runs to automate data operations.How-To Guide articles: Practical Tips from the ExpertsWe’ve added a fresh set of how-to guides. These articles are designed to help you troubleshoot faster, work more efficiently, and get the most out of CDF.Enabling the alpha API Subversion and DEBUG mode in the Cognite Python SDKWith this guide, learn how to work with alpha API versions in Cognite Data Fusion and use the Python SDK debug flag to capture detailed HTTP logs for troubleshooting and testing.It also explains which configuration parameters must be set during client initialization for consistent behavior across environments (Jupyter, local scripts, CI/CD) and which ones can be safely modified later. How to capture x-request-id for debugging CDF Functions With this guide, learn how to trace and debug your CDF API calls using the x-request-id returned with every request. It shows how to log the request ID from failed calls using CogniteAPIError, how to temporarily enable client.config.debug = True to capture it for successful calls, and best practices for using debug logging. It also highlights how extractors (OPC UA, PI, PIAF) can be configured to automatically include x-request-id in logs. Master Cognite Data Fusion with new training on Cognite Academy. We’d love to hear your thoughts and help you connect with other experts. Come join the conversation on Cognite Hub.  

We are committed to continuously improving your experience in Industrial Canvas. Our goal is simple: help you generate value faster through a seamless, uninterrupted workflow—eliminating the friction of slow load times and performance lag. With our March release, Industrial Canvas now loads 50% to 70% faster on average. We’ve also reduced memory usage by up to 70% during loading, even for your most complex canvases. We look forward to hearing your feedback!Best,Andreea

While most data in Cognite Data Fusion flows through structured pipelines, we know that critical insights often live with your experts.To empower non-technical users, like Reliability Engineers, we have introduced Direct File Upload. This allows Subject Matter Experts to instantly upload resources (such as a Root Cause Analysis template, a report etc.) directly into the Search UI or through resource selectors in tools like Canvas.This feature was released on 3 March 2026.   Admin governance & setupTo ensure these files remain organized and secure, Admins must first define a destination for each Location for the File Upload option to appear. The File Upload option will only appear for users once this is configured.To enable this feature:Go to Admin Workspace > Project Settings. Select Locations > Configure. Define which Space should be used for all file uploads in that location.How permissions workAccess is controlled through the permissions of the destination Space you’ve selected:Who can upload? Users who have Write access to the configured Space. Who can view? Users who have Read access to that Space.Where is it visible? In the Search UI and as a Resource Selector (specifically for Data Modeling File Views, e.g File upload to Asset-centric File resource is not supported).Looking forward to seeing this highly requested feature in use!

We are thrilled to announce the Beta release of our new Simulator UI within Cognite Data Fusion (CDF). For too long, high-fidelity simulation runs have been locked behind a "technical wall" of complex APIs and developer-only workflows. Not anymore.We are handing the keys to the Subject Matter Experts (SMEs). Whether you’re optimizing production or running custom models, you can now orchestrate the future of your assets without writing a single line of code. From API-First to SME-ReadyPreviously, our simulator connector integrations required deep technical knowledge to configure and run. Our new interface replaces manual API calls with a sleek, intuitive workspace. Visual Logic Orchestration: The Routine BoardThe heart of this release is the Routine Board. SMEs can now visually map out simulation sequences, setting inputs, defining calculation steps, and extracting results using a drag-and-drop interface. 

Hi everyone!Finding an asset is one thing; understanding its surroundings is another. In the upcoming release on the 3rd of March, we’ve upgraded the 3D Experience in Search to give you instant spatial context without leaving your search results.What’s New?In-Preview Scenes & 360-images: View 3D scenes and 360-degree images directly in the search preview pane. Seamless Transitions: Move from a search preview to the unified 3D world while staying immersed in your 360-degree view. Direct Access: Launch full Scenes immediately from the preview card.See your asset in it’s full real-world context byviewing your 3D Scenes in Search⚡ Quick Start:Prepare - Ensure you are using Scenes for 3D and that they are connected to your Location Search for an asset or location in CDF. Preview the result to see the 3D scene load automatically. Explore by opening the unified 3D view for full immersion.The preview also supports fully interactive 360 images Pro Tip 💡: Opening the full 3D view from a 360 preview maintains your exact orientation—no more "re-finding" the asset once the viewer loads!Full Guide: https://docs.cognite.com/cdf/explore/navigate_3dGive it a spin and let us know what you think!

We are excited to announce the Q1 2026 release of Cognite Data Fusion (CDF). This quarter, our focus is on breaking down technical barriers for Subject Matter Experts and providing a more immersive, high-precision data experience.Empowering SMEs: The New Simulators UI (Beta) Simulators UI (Beta) Immersive Data Exploration & 3D Context Chart - Activities 3D - MiniMap 3D - Experience in Search Smarter Search & Precision Tools Search -  Smarter Results Search - File Upload Records - Unit-Aware (Beta) Atlas AI: Continuity & Control Atlas - Conversation History Atlas - Per Agent Granular Access Control Performance & Readiness Canvas - Performance 3D - Browser Caching Diagram Parsing - Tag Detection Transformations - Full Metrics in Run History Looking Forward: Production-Grade Tailored SolutionsEmpowering SMEs: The New Simulators UI (Beta)We are transitioning from a "Code-First" to a "User-First" approach for industrial simulations. Feature Problem (Challenge) Solution: What You Can Do Now Simulators UI (Beta) SMEs are currently blocked by a "technical wall," where simulator connectors remain locked behind APIs that can be configured only by developers. Empower SMEs to independently configure their own models and use a no-code "Routine Board" for drag-and-drop visual logic orchestration.   Routine builder  Immersive Data Exploration & 3D ContextCorrelating data trends with the physical world is critical for efficient troubleshooting. Feature Problem (Challenge) Solution: What You Can Do Now Chart - Activities Industrial users struggle to correlate data trends with real-world operational events. Overlay activity types, such as work orders, directly onto time series in Charts to instantly connect anomalies with the "why". Documentation 3D - MiniMap When navigating 3D scenes in Search, it is easy to lose orientation and real-world context. Use customizable real-time minimaps to track camera position and "fast-travel" through the scene by clicking destinations on the map. Documentation 3D - Experience in Search Users previously lacked 3D scenes in search previews, leaving them without context for model placement. View 3D Scenes and 360-degree images directly in the search preview to gain immediate context for a selected Location. Documentation  Smarter Search & Precision ToolsWe’ve overhauled our search and resource management to prioritize accuracy and ease of use. Feature Problem (Challenge) Solution: What You Can Do Now Search -  Smarter Results (opt-in beta, and GA April 2026) Standard search often fails to handle partial industrial tags or returns broad, inaccurate results for complex strings. Benefit from advanced "Industrial Parsing" and prefix matching that prioritizes result accuracy over quantity. This is applicable for search on Core Data Model. Search - File Upload SMEs lacked a simple way to upload individual files, such as RCA templates, without relying on technical pipelines. Upload single files directly through the Search UI and use Search as a resource selector in tools like Industrial Canvas. Upload is available only for data modeling views. Records - Unit-Aware (Beta) Users must manually convert sensor values when querying, filtering, or displaying data across different unit systems. Perform unit-aware querying, filtering, and aggregation within Records to eliminate the need for custom unit-conversion code. Documentation   Atlas AI: Continuity & ControlOur AI agents are now more persistent and easier to manage within your existing workspace. Feature Problem (Challenge) Solution: What You Can Do Now Atlas - Conversation History Atlas agents didn't retain session history, forcing users to restart workflows and lose context. Continue past chats where you left off or work across multiple parallel conversations without losing context. Atlas - Per Agent Granular Access Control Previously, all agents were visible to every user, and only creators could modify them. Configure specific read, write, and run rights on a per-agent basis to ensure users only interact with authorized agents.  Performance & ReadinessWe continue to optimize our core tools to ensure they are production-ready and provide full transparency into pipeline health. Feature Problem (Challenge) Solution: What You Can Do Now Canvas - Performance Increased load on Industrial Canvas caused sub-optimal load times and high memory consumption. Experience a 50% average reduction in load times and up to 40% less memory usage during loading. 3D - Browser Caching Increasing 3D data volumes led to performance degradation and suboptimal load times for large scenes. Utilize client-side browser caching for CAD and point cloud data to enable 63% faster 360 image loading and 52% less data transfer, providing instant loading for returning users Diagram Parsing - Tag Detection Tag detection wasn't ready for Data Modeling, blocking users from migrating from asset-centric models. Use production-ready tag detection for Data Modeling with built-in per-file parsing status. Documentation  Transformations - Full Metrics in Run History The UI previously showed limited metrics, making it difficult to fully debug issues or track specific rate limits. Access the full library of API-tracked metrics, including searchable selectors, directly within the Transformations Run History UI.  Looking Forward: Production-Grade Tailored SolutionsBeyond this release, we are building a future where tailored industrial applications are delivered with unprecedented speed and precision. We are evolving our platform to support production-grade, tailored solutions that can be deployed natively in just days. Rapid Delivery of Industrial WorkflowsWe are targeting the "long tail" of industrial use cases, thousands of small and medium-sized workflows that deliver significant value but were previously too complex to build.Tailored Experiences: Moving from generic interfaces to AI-first, agentic user experiences designed for specific roles like reliability engineers or maintenance supervisors. Native Deployment: Fully featured applications deployed directly as part of the Cognite product stack. Immediate Value: Targeting 100X faster deployment times and up to 80% lower costs by automating the generation of industrial applications.Compounding the Value of Your Industrial DataThis new capability is built on the foundation of the Industrial Knowledge Graph and Atlas AI. Instead of creating new data silos, these tailored applications write insights and outcomes back into the knowledge graph. This ensures that every workflow contributes to a growing, unified body of industrial knowledge that can be scaled across the entire enterprise. Ready to dive deeper? Check out our Release Notes  

Hi everyone!We know that for an end-user, finding data shouldn't feel like a "needle in a haystack" mission. Last summer, we launched Per-Project Search Configuration, and if you haven't set it up yet, your team might be missing out on a much cleaner experience.As a Project Admin, you can curate what your users see when they sign in to Search. Instead of a generic view of data that maps 1:1 to your detailed data model definition, you can hand-deliver the most relevant data. Why configure your categories? Instant clarity: Define exactly which columns appear first. No more side-scrolling to find the "Status" or "ID." Relevant filtering: Hide the noise. Display only the filters that actually matter for that specific category. Context at a glance: Tailor the "Properties" card so the most crucial metadata is front and center. Quick guide: Check access: Ensure you have appconfig:read/write and app-scope = Search. Your users also need appconfig:read to make the config take effect for them.  Finding the config options: Go to Admin workspace > Project Settings > Categories. Configure: Pick a category and hit + Add.  Pro Tip 💡: After you update the columns, remind your users to hit "Reset" in their search column selector to see your shiny new layout!Ready to clean up your search? Full Guide here: https://docs.cognite.com/cdf/configure/project_settings/  Have you tried? Let me know if you have any feedback!

Understanding the inner workings of your data pipelines is key to taking action to improve performance and debug issues. To give you better visibility, we’ve unlocked the full library of transformation metrics directly in the Transformations Run History UI.Previously, the UI only showed a few key metrics. Now, you can access everything the API tracks-from specific rate limits to granular resource updates-right from the graph.Navigate to the Run history tab in the Transformations UI, and select more metrics as shown in the images below.  What’s New?Searchable Metric Selector: Use the new dropdown to find exactly what you need. With the new search bar, you can quickly filter through long lists of metrics if you're working with multiple tables or data models. Smart Defaults: To keep things clean, your most important metrics (like reads, updates, and total requests) are still shown by default. Everything else is just a click away.Identifying Efficiency GainsThis update makes it easier to track the instances.upsertedNoop metric we recently introduced. By comparing "upserted" vs. "upsertedNoop," you can see exactly how much data is being re-written unnecessarily. High "No-op" counts are a clear sign that you can save time and compute costs by using smarter, incremental loading.Give it a try!Head over to your Transformations Run History today to explore these new insights. We’re always looking to improve, so please share your questions or feedback in the comments below! 

Microlearning videos: Small lessons, big impactWe’ve added a new set of bite-sized microlearning videos focused on simulator integrations in CDF. These short lessons cover key concepts and practical steps, making it easier to learn at your own pace.Simulator Integrations: Core ConceptsDiscover how simulator integrations support optimization and digital twins by connecting simulation tools to CDF, and how these building blocks work together to run and manage simulations effectively.The Simulators API: Key Resources and Data FlowExplore how simulations are created, controlled, and tracked in CDF- from defining models and routines to running simulations and reviewing results and logs.Configuring a Simulator ConnectorLearn how to set up a Windows-based simulator connector so CDF can securely communicate with your simulation tools, including authentication, network access, and service configuration.Configuring Simulator RoutinesLearn how to define reusable simulation setups that control how simulations run, covering inputs, outputs, scheduling, and execution logic.Data Sampling and ValidationFind out how to ensure your simulations run on reliable, high-quality data by configuring time windows, sampling strategies, and validation checks to catch unstable or invalid inputs.Triggering Simulation RunsUnderstand how to start simulations manually, automatically, or programmatically - and how to review outcomes using logs and simulation results. How-To Guide articles: Practical Tips from the ExpertsWe’ve added a fresh set of how-to guides. These articles are designed to help you troubleshoot faster, work more efficiently, and get the most out of CDF.How to fix Video Playback Failure – Error Code: 232011Get back to streaming smoothly by resolving the common 232011 playback error. This guide walks you through clearing cache and cookies, checking browser settings, and avoiding common conflicts.How to: Understanding CDF User Access and Entra ID ConnectivityTake a closer look at how CDF handles orphaned content and access loss, along with admin best practices to ensure critical data remains accessible.How to Build Efficient Transformations in CDFImprove performance and scalability in your CDF transformations using proven techniques like early data filtering, incremental logic with is_new, and CDF Workflows for complex jobs.How to Upload Large Files to CDFConfidently upload files larger than 5 GiB using Multipart Upload. Learn how to split data into chunks, enable parallel uploads with retries, and implement the flow using the Cognite Python SDK.How-To: Getting started with the CDF File Extractor from local folder to data modelSet up the CDF File Extractor on Windows to ingest local files into CDF. This guide covers downloading the extractor, configuring authentication, and mapping files to Core Data Model objects.How-To: Getting started with the CDF DB Extractor to populate staging (RAW) with records from a CSV fileIngest local CSV data into CDF Staging (RAW) on Windows using the Cognite DB Extractor. Learn how to configure authentication, treat CSVs as spreadsheet databases, and populate target RAW tables.How-To: Use a CSV file to populate data points for a time series (data model) with the CDF DB ExtractorLoad CSV data directly into CDF Time Series on Windows. This article explains how to map timestamps, values, and external IDs to create or update CogniteTimeSeries objects.How to: Transformation Execution Timestamps ExplainedDemystify CDF transformation timestamps - Created Time, Started Time, and Duration - so you can better understand performance without confusion from queuing delays.How to add filtering in Cognite OPC UA ExtractorGain more control over OPC UA data ingestion by defining effective filters. Learn how to use the Include transformation and full tag hierarchies for accurate extraction.How to resolve 'Bounding Box is empty' error when uploading a .LAZ fileFix this common 3D point cloud issue by re-saving LAZ files with compressor 2 or converting them to E57 using tools like CloudCompare.How to resolve 'This Location contains no files yet' error in diagram parsing for data modelingEnsure files are visible for Diagram Parsing in CDF Data Modeling by correctly including them in the data model and linking them to a Location with the right instance spaces. Master Cognite Data Fusion with new training on Cognite Academy. We’d love to hear your thoughts and help you connect with other experts. Come join the conversation on Cognite Hub. 

We’re excited to announce the release of Activities in Charts, a new capability that brings operational context directly into time series analysis on CDM and Hybrid Projects.With Activities in Charts, you can now visualize asset-related activities, such as work orders, maintenance events, or operational tasks, directly on your Charts, alongside your time-series data. This makes it easier to understand what happened, when it happened, and how it impacted performance.Why this mattersWhen investigating operational behavior, data alone is often not enough. Activities provide critical context that helps teams:Correlate events and activities with changes in sensor data Perform faster root-cause analysis Reduce the need to switch between tools during investigationsBy bringing activities into Charts, we enable a more unified and intuitive analysis experience.What you can do with Activities in ChartsView activities overlaid on time-series charts Align operational events with performance trends Use Charts as a central workspace for investigation and troubleshootingThis capability also replaces the legacy Events visualization previously available in Asset Centric projects, offering a more modern and integrated experience.Documentation 

We have released a new course dedicated to building reliable, industry-grade agents within Cognite Atlas AI.This course is your starting point for building, configuring, and deploying Atlas AI agents in Cognite Data Fusion. You will gain a comprehensive overview of how agents function, how they connect to industrial data, and how to leverage them for smarter automation and decision-making.This course focuses specifically on how to build agents that adhere to strict operational logic suitable for your use-case and industry standards. We demonstrate how to properly configure tools, select the right models based on reasoning benchmarks, and use the instructions box to enforce standard operating procedures.Key topics covered in this course:Tool Configuration: How to grant agents access to CDF data through specific tools, creating an assistant that genuinely understands physical assets. Model Selection: Utilizing industrial benchmarks to select the right models for complex reasoning and precise filtering. Operational Instructions: Using the instructions box to implement SOPs that keep agents methodical and verified. Advanced Querying: How to use the Query Knowledge Graph tool to customize your agent for specific data types.By the end of this course, you’ll be equipped to design agents that rely on your data and documents as a source of truth. This approach ensures your AI assistants remain predictable and safe for industrial workflows.Access the full course on the Academy here 

We are on a mission to help everyone build and deploy better, more scalable, and Atlas AI-ready data models in Cognite Data Fusion. Today, we introduce NEAT 1.0 as an official part of our product tooling. Why NEAT 1.0 Matters A reliable, scalable, and usable data model is the bedrock of the Knowledge Graph in Cognite Data Fusion. NEAT 1.0 is designed to give you confidence that your models are usable, scalable, extendable, and performant. We've taken years of deep product expertise and hard-earned experience from our most demanding customers and turned it into concrete feedback and guardrails you can leverage in your automated development workflow, giving you help exactly when you need it.What's New?  Faster & Focused: We concentrated on physical data modeling and made NEAT significantly faster than previous versions. We also adopted a clean, object-oriented approach for interacting with features. Better Quality Checks: We've included a ~2x larger library of data model validators (constantly growing at thisisneat.io/validation). This means more checks to ensure your model is viable, scalable, and maintainable. Deep Pre-Deployment Analysis: The dry-run feature was completely rebuilt to give you a deep, clear analysis of the changes that will occur, including a severity score, before you even push your model to CDF. Easy Issue Navigation: You get rich analysis of data model issues with interactive navigation, search, and clustering, making it simple to find and fix problems. Ready for Governance: We take a user-centric approach to anything graph related, offering four pre-built, configurable data model governance profiles to jumpstart your quality assurance.What About Legacy Features? NEAT 1.0 embraces focus and minimalism. To make this release more performant and easier to maintain, we removed some features from older, legacy versions that did not align with our new, streamlined approach. Legacy features continue to be available from cognite.neat.legacy import NeatSession while we continue to port these features to V1. Getting started?Watch the short video and check out https://cognite-neat.readthedocs-hosted.com/en/latest/installation.html to try for yourself!

Hi everyone,We are thrilled to announce that on Monday, December 8th, we will be releasing a highly requested feature: Search-in-file!This new functionality supports searches within individual .pdf files across Asset-centric/Classic and Data Modeling files.Problem: Finding content within a file can be a daunting task without the ability to perform a quick search across the file. Feature: Search box in the file viewer allowing the user to search for text (single word) within the file (.pdf). The text gets highlighted, and if there is more than one match, you get the option to browse through all matches. (Wanted) Impact: Faster exploration, saving time (and unnecessary frustration) working with files.We hope this update will greatly improve your workflow and productivity. As always, your feedback is invaluable to us. Please continue sharing your thoughts to help us improve.Best,Sofie, Product Manager Search

2025: Delivering Industrial Scale and ReliabilityThis year, we turned industrial potential into enterprise reality. Every release, from Q1 to Q4, was designed to help you build and run AI and data solutions at unprecedented scale.We focused on three pillars:Data Foundation at Scale: Handle billions of records with ease and accelerate model creation to eliminate bottlenecks. Workflow Automation: Introduced visual orchestration and enhanced validation for reliable, high-volume processes. AI You Can Trust: Moved AI agents from pilot to production with governance, monitoring, and precision tuning for enterprise-grade reliability.Our Q4 release completes this journey, delivering the critical building blocks that make CDF the only platform purpose-built for industrial scale.Operational Excellence & Field Work Replacement Faster, Cleaner Troubleshooting in Industrial Canvas (Beta) Get Immediate Context: Location Prompt (New User Experience) Seamless Planning for Frontline Teams (Beta) Smarter, Faster 3D Measurements (Replace Field Work) Customized 3D Experience (Reduced Friction) Full 3D Support for Hybrid Projects Unlocking Scalability and Optimization Massive Log and Event Data Storage: Records API Full-Field Production Optimization (Beta) Driving Data Quality & Model Confidence High-Quality Field Observations Reliable Workflow Development Total Process Visibility (Workflow Triggers) Transparent Document Parsing (LLM Vision) (Beta) Streamlined Annotation Review for Diagrams (Beta) Identify and Optimize Transformation Inefficiency Agent Reliability and Precision (Atlas AI) Data-Driven Agent Confidence Precise and Predictable Agent Queries Platform Foundations Enhanced Japanese Search  Operational Excellence & Field Work ReplacementStreamline field work, enable safe remote operations via 3D, and boost efficiency with improved operational context and real-time troubleshooting.Faster, Cleaner Troubleshooting in Industrial Canvas (Beta)  Problem Solution: What You Can Do Now Cluttered Canvas: Complex diagrams became unreadable due to overlapping connection lines, slowing down troubleshooting. Instant Flow Tracing: The canvas is now cleaner. Hover over any connection path to instantly highlight that specific flow, increasing user efficiency when troubleshooting. Get Immediate Context: Location Prompt (New User Experience)  Problem Solution: What You Can Do Now Overlooked Context: New users often missed the critical Location filter, leading to poor data context and frustration. Contextual Prompting: If a location is not selected, a pop-up prompt appears, directing users to the filter. This ensures a greater set of users set the right context immediately, improving the data quality of their work. Seamless Planning for Frontline Teams (Beta)  Problem Solution: What You Can Do Now Limited Schedule Transparency: Supervisors lacked full visibility into what was planned for the upcoming weeks and control over what was published to frontline teams. Full Schedule Control: The new Schedules tab provides full transparency into scheduled checklists. Supervisors can decide what is published to the frontline teams and when, ensuring priorities are met. Users can also create checklists from the schedule Smarter, Faster 3D Measurements (Replace Field Work)  Problem Solution: What You Can Do Now Manual Measuring: Measuring basic dimensions like pipe diameters was laborious. One-Click Precision: Perform one-click diameter measurements for pipes, vessels, and tanks. This increases the viability of using 3D to replace field work. Customized 3D Experience (Reduced Friction)  Problem Solution: What You Can Do Now Manual Tweaking: Hardcoded default 3D settings forced users to manually adjust them per session. Customizable Settings Per Scene: Define default Model Visibility, Rendering Quality, and Point Cloud settings. This reduces friction for regular users and improves the first-time experience. Full 3D Support for Hybrid Projects  Problem Solution: What You Can Do Now Migration Bottleneck: 3D was not fully supported by the Cognite Data Model (CDM) or hybrid projects, blocking customer migrations Seamless Hybrid 3D: New 3D service API endpoints for CDM enable a seamless 3D user experience in hybrid projects. This enables migration to CDM and allows contextualization using both asset-centric and data modeling assets. Unlocking Scalability and OptimizationHandle billions of records and unlock full-field industrial optimization that was previously impossible, accelerating enterprise value realization.Massive Log and Event Data Storage: Records API  Problem Solution: What You Can Do Now Scaling Limitations: Billions of high-volume log entries and events overburdened the Knowledge Graph node structure, limiting scalability. Records API: Store billions of structured records with seamless integration to Data Modeling. This unlocks the next level of CDF scalability. Full-Field Production Optimization (Beta)  Problem Solution: What You Can Do Now Blocked Use Cases: Current CDF capabilities couldn't handle optimizing entire offshore fields with thousands of wells simultaneously. Large-Scale Workflow Support: Run complex simulations across entire facilities, handling thousands of parameters. This unlocks true field-wide production optimization. Driving Data Quality & Model ConfidenceGuarantee reliable data ingestion, enable better decision-making with high-quality observations, and validate complex workflows before deployment.High-Quality Field Observations  Problem Solution: What You Can Do Now Incorrect Asset Links: Operators struggled to attach the correct asset to field observations, reducing data quality. Correctly Contextualized Observations: Access the asset hierarchy to browse and find the correct asset on both mobile and desktop. This ensures observations are captured with the correct asset, leading to better decision making. Reliable Workflow Development  Problem Solution: What You Can Do Now Trial-and-Error: Complex definitions made subtle errors difficult to find, leading to a frustrating configuration. Actionable Validation: Perform comprehensive on-demand validation and auto-validation on publish. Errors are surfaced, explained, and made actionable directly in the UI. Total Process Visibility (Workflow Triggers)  Problem Solution: What You Can Do Now Hidden Triggers: Managing live pipelines was risky without clear visibility into automated starting points. Workflow Triggers as Nodes: Triggers are now shown as canvas nodes for full workflow visibility. This delivers a true end-to-end visualization and enables safe management of live pipelines Transparent Document Parsing (LLM Vision) (Beta)  Problem Solution: What You Can Do Now Verification Challenge: Users couldn't easily verify LLM parsing results because extracted values weren't visually linked to the source document. Visual Verification: Extracted values are displayed with bounding boxes in the UI, making verification easier. You can also store data in a user-specified space. Streamlined Annotation Review for Diagrams (Beta)  Problem Solution: What You Can Do Now Slow and Cumbersome Review: Users lacked an efficient way to verify and reject annotations, especially in bulk, and faced limitations when dealing with multi-page files. Dedicated Review Workflow: The UI tabs are split into four distinct sections, with the first tab dedicated solely to reviewing annotations. This allows users to verify or reject file and asset annotations individually or in bulk and enables annotation management across multi-page files, significantly improving focus and workflow efficiency. Identify and Optimize Transformation Inefficiency  Problem Solution: What You Can Do Now Wasted Resources: Transformations often re-process and re-write large amounts of identical data, wasting compute resources and masking optimization opportunities. Monitor "No-Op" Performance: The Run History UI now displays a "No-Op" metric, showing how many write operations were skipped because the data was unchanged. This instantly identifies inefficiency and drives smarter, incremental data loads. Agent Reliability and Precision (Atlas AI)Move AI agents from pilot to production by providing tools to monitor performance, guarantee output precision, and build trust at scale.Data-Driven Agent Confidence  Problem Solution: What You Can Do Now Blind Deployment: Lack of visibility into agent performance or failures made it hard to build trust or troubleshoot. Agent Performance Monitoring: Create test sets and run evaluations to view pass/fail results. This enables data-driven confidence in agent reliability and supports smoother UAT. Precise and Predictable Agent Queries  Problem Solution: What You Can Do Now Unpredictable Output: Agent queries could result in slightly different data or format, creating uncertainty in production workflows. Guaranteed Precision: Configure a specific query and lock fields, operations, and output. This guarantees critical data is retrieved correctly and consistently every time.  Platform FoundationsFundamental improvements to core platform elements, including search quality and the initial user experience.Enhanced Japanese Search  Problem Solution: What You Can Do Now Poor Japanese Search: The existing search didn't account for the unique structure of Japanese, leading to poor relevance. Morphological Analysis: Enhanced search capability uses Morphological Analysis to provide relevant results, balancing precision and recall.  We believe these new capabilities will significantly accelerate your industrial data journey. Dive into the detailed release notes to explore all the new features and improvements.We'd love to hear how these features are transforming your operations! 

We've just published The Atlas AI™ Security & Trust Framework , a new white paper detailing how we've engineered our low-code industrial AI agent workbench for the rigorous security demands of industry leaders.We know security and data privacy are critical when adopting generative AI. This framework outlines how Atlas AI is secured by design, with security as an intrinsic part of its architecture, not an afterthought. Key highlights from the framework: Your Data Stays Your Data: We have contractual guarantees that your prompts and responses are not stored by third-party LLM providers and are never used to train or improve their models. Your Access is the Agent's Access: Atlas AI integrates with your corporate Identity Provider (IdP) for SSO. Critically, all API calls an agent makes are mapped directly to your user roles and permissions. The agent can only see and do what you can. Enterprise-Grade Security: The platform is built on a "defense-in-depth" architecture , employs end-to-end encryption (at-rest and in-transit) , and undergoes continuous 24/7 monitoring and third-party penetration testing. Independently Verified: Atlas AI inherits the comprehensive security and compliance of Cognite Data Fusion®, including our ISO 27001, ISO 9001, and SOC 2 Type 2 certifications.  Please find the full white paper here for a deeper dive. We're committed to being your trusted partner in Industrial AI and welcome your questions in the comments.

 Hi everyone,I'm happy to share some great news! We've just shipped a product update that hopefully makes working with Annotations more fluid and intuitive.We heard your feedback that it was sometimes hard to pinpoint which annotation you were working with, especially on a busy file. To solve this, we've made two key improvements:See what you've selected: When you click on an annotation—either in the side panel or on the file itself—it will now turn green so you can easily see what you've selected. Instantly zoom in: Clicking an annotation in the side panel will automatically zoom in to its location in your file.Check out the screen recording showing the new behavior (it is available both from the search tool, and the “add data” functionality in Canvas and Charts):We hope these improvements help you work more efficiently and accurately with annotations. As always, your feedback is invaluable—please keep sharing your thoughts to help us get better. Best, Sofie, Product Manager

This quarter’s release features new training on building industrial AI agents, enhancing 3D environments, improving data workflows and an all new and improved CDF fundamentals course to help you make better decisions and get more value from your data. Courses: Build In-Depth Skills, Step-by-StepCognite Data Fundamentals with Data ModelingA new and improved CDF Fundamentals course is now available. This introductory learning path is perfect for anyone new to Cognite Data Fusion. The updated content now covers the Core Data Model, hands-on exercises with Industrial Canvas, and an overview of Atlas AI. Complete the course to earn a shareable badge and certificate.3D Configuration and ManagementThis course is designed to give you the skills to set up and configure 3D environments and integrate them into your daily work. You will learn to use 3D solutions to improve visualization, enhance decision-making, and boost operational efficiency.This course is for anyone interested in learning about 3D technology and applying it in their professional field.Microlearning videos: Bite-Sized LearningsAI Agents in Industrial WorkflowsDiscover how Cognite's industrial AI agents, powered by Atlas AI, can help your teams act faster and with greater confidence. Learn how to leverage contextualized industrial data to streamline troubleshooting, automate routine tasks, and enhance operational efficiency.Building Agents in Atlas AIMaster the process of building and customizing industrial AI agents using Cognite Atlas AI. This low-code workbench enables you to automate complex tasks by connecting to contextualized data in Cognite Data Fusion. See how to create, configure, and deploy agents to analyze time series, retrieve documents, and support decision-making—all without writing a single line of code.Contextualizing Data for AIUnderstand how to transform raw industrial data into a scalable, connected foundation for advanced workflows. Learn to build and enrich a Knowledge Graph using tools like Data Workflows, diagram parsing, and document parsing, enabling seamless reuse of data and powering robust AI agents and applications. How-To Guide articles: Practical Tips from the ExpertsChoosing the Right Extractor for Your CDF IntegrationCognite Data Fusion offers a wide range of extractors to integrate data from industrial and IT systems into a unified platform. These are divided into two main types: Prebuilt Extractors, installed within customer infrastructure to connect with systems like OSI PI, OPC UA, databases, SAP, and document libraries; and CDF-Hosted Scalable Extractors, managed in the cloud for high-volume, real-time data ingestion from sources such as Azure Event Hub, MQTT, Kafka, and RESTful APIs.Understanding Function Memory Quotas in CDF Learn best practices to prevent memory errors in CDF functions by designing workloads with memory limits in mind. This guide explains why breaking tasks into smaller, manageable chunks is the most effective approach.How-To: Simple Streamlit App for CRUD Operations on CDF Data Model viewThis guide shows you how to create a simple Streamlit web app to manage data model instances in Cognite Data Fusion (CDF) using the Python SDK. Learn to create, read, update, and delete (CRUD) instances in real time—perfect for demos, quick data manipulation, or onboarding users to structured data modeling.Getting started with CDF performance testing This guide explains the importance of performance testing in Cognite Data Fusion (CDF) and how to assess key metrics like data ingestion rates, query performance, and API response times. It includes example notebooks to help you test efficiently, optimize costs, and ensure production readiness.Changing Your Identity Provider (IdP): Playbook and Key Information This guide explains the relationship between projects, organizations, and Identity Providers (IdPs) in Cognite Data Fusion. It covers the implications of changing an organization’s IdP, including session invalidation, user ID reassignment, and potential data access issues, and provides guidance on planning and requesting IdP changes safely.Unsubscribing from Cognite Charts AlertsThis guide explains how users can unsubscribe from Cognite Charts alerts and notifications, ensuring they no longer receive emails when time series thresholds are breached.Fixing Video Playback Failure – Error Code: 101102This guide helps troubleshoot Error Code 101102 when a training video fails to load. It covers common causes—like network restrictions, browser extensions, VPNs, or outdated browsers—and provides step-by-step solutions to restore smooth video playback.Optimizing Data Ingestion PerformanceLearn best practices to improve data ingestion performance in Cognite Data Fusion, including batching datasets, using progress monitoring, efficiently handling asset hierarchies, leveraging delta loads with the is_new function, and scheduling transformations for faster, more reliable workflows. Product Updates: What's New to Enhance Your WorkflowImproving Cognite Functions stability with new rate limitsWe're implementing new rate limits for Cognite Function calls to improve system performance and stability. A new limit of 250 concurrent running calls per CDF project will apply immediately to new projects and will go into effect on November 1, 2025, for existing projects. This change will affect a small number of projects with very high usage, ensuring a more equitable distribution of resources.Power BI REST Connector now Generally AvailableThe Power BI REST API connector is now Generally Available (GA), offering better performance and broader capabilities than the legacy OData connector, which will be retired on August 18, 2026. The new REST API connector provides broader authentication support, enhanced data access, and a significant performance boost of up to 10x, and should be used for all new reportsNew features for Cognite's GAP ConnectorWe've added two new capabilities to the GAP connector that address common challenges Petroleum Engineers face when working with GAP simulation models in production environments.Information Model Extraction The GAP connector can now automatically extract detailed flowsheet information models from simulator model revisions. This makes flowsheet data, including equipment, properties, and connections, accessible for analysis directly in CDF, eliminating the need for separate simulator access. External Dependencies Support This new feature allows you to manage simulation models by linking separate CDF files for components like well models or VLP tables, instead of repackaging the entire model. This simplifies updates, reduces storage overhead, and enables different team members to work on components independently. Explore new trainings on Cognite Academy. For guides, community, and to share your ideas, visit Cognite Hub. 

We've added two new capabilities to the GAP connector that address common challenges Petroleum Engineers face when working with GAP simulation models in production environments.Information Model ExtractionThe GAP connector now can automatically extract detailed information models from simulator model revisions, making flowsheet data accessible for analysis without requiring direct simulator access. When you upload a GAP model revision, the connector can parse the complete flowsheet hierarchy, capturing all functional blocks such as wells, pumps, valves, pipes, and tanks along with their operating parameters, physical properties, and configuration settings. The system also extracts connection relationships between equipment and graphical positioning information for visualization processes.This extraction is configurable through a modelparsing.config.yml file where you define which properties to extract from different equipment types. You can specify common properties that apply to all equipment (like oil rates, water rates, and gas rates from solver results) as well as equipment-specific properties (such as well model types, pipeline correlations, or valve characteristics). Each property definition includes the data type, units, and the GAP address suffix used to retrieve the value from the model.The parsing process activates automatically when a new model revision is created or when you reparse an existing revision, provided a valid configuration file exists. The connector loads the GAP model, discovers all equipment instances, extracts the configured properties along with unit information, analyzes equipment connections to generate material flow relationships, and stores the complete information model in CDF alongside the simulator model revision.The parsing adds some processing time during model validation, with the duration depending on model complexity. If property extraction fails for individual items (due to invalid addresses, incorrect unit quantities, or wrong value types), the connector logs the errors and continues processing rather than failing the entire operation.For configuration instructions and examples, see the public documentation.External Dependencies SupportThe GAP connector now supports a new external dependencies mode that changes how you manage simulation model components. Traditional GAP workflows require packaging everything into single .gar archive files, which can get quite large for big networks. Any change to individual components like well models or VLP tables necessitates repackaging and re-uploading the entire bundle, creating storage overhead and making frequent updates impractical.The external dependencies mode allows you to upload .zip files containing only the GAP network files (production and injection networks) while mapping individual node dependencies like .OUT, .VLP as separate CDF files. When creating a model revision, the user provides a mapping structure that links each external dependency file to its proper location in the GAP model using the simulator's address notation. For example, you can map a PROSPER .OUT file to GAP.MOD[{PROD}].WELL[{A1}].File and a .VLP file to GAP.MOD[{PROD}].WELL[{A1}].VLPFile.This approach enables to update individual well models, VLP tables, or other components without rebuilding entire model bundles. Different team members can work on different components independently, and you avoid duplicating unchanged files across model revisions, significantly reducing storage costs. The connector handles the complexity by downloading the network files, processing the external dependencies mapping, downloading each referenced CDF file separately, resolving file paths using the provided arguments, and assembling everything into a complete simulation model.For complete setup instructions covering both the new external dependencies mode and the single bundle mode, refer to the the public documentation.These features are currently in beta and subject to change based on user feedback. Access the parsed information through the existing simulator model revision endpoints, and configure external dependencies through the standard model revision creation process.