Recently active topics

Inspiration“Context is king in the world of AI.”Across research, publications, and industry discussions, one theme consistently stands out — AI without context lacks true intelligence. To unlock the full potential of Industrial AI, we must ground AI solutions in process context.VisionIntroduce Process-Aware Knowledge Graphs (PAKGs) that integrate process understanding directly into the Cognite Data Fusion (CDF) ecosystem. By capturing and structuring the interconnections, interdependencies, and material flows from Process Flow Diagrams (PFDs), we enable context-driven intelligence for Agentic AI solutions built on Atlas AI and CDF.Core Capabilities System Model Extraction Automatically extract process metadata from P&IDs and PFDs (PDF/Image formats). This removes the dependency on CAD files, which are often unavailable or inconsistent. Process-Aware Knowledge Graph Generation Translate the extracted system model into a Knowledge Graph enriched with process semantics.

Problem StatementCognite Data Fusion (CDF) offers a powerful suite of tools for industrial data operations, but its adoption remains limited to highly technical users such as data engineers, data scientists, and developers. Today, creating data transformations, writing functions, deploying models, and generating insights in CDF typically requires:Knowledge of Spark SQL for transformations Python programming for custom functions Understanding of data modeling concepts Manual deployment and orchestrationThis steep technical barrier restricts broader usage, particularly among domain experts like production operations engineers, maintenance supervisors, or process owners who possess deep contextual knowledge but lack coding skills. As a result, CDF usage and ROI are throttled by dependence on a small pool of technical resources. VisionEmpower every domain expert to become a CDF power user — without writing a single line of code. Proposed Solution: Cognite vision – AI-Powered No-Code Experi

Hi Team,This request is to include time series’ UnitExternalId on UI/front-end under the time series information in the data explorer/search.Currently, it is visible using transformations, APIs or SDK. But someone who is not adept to use transformations, APIs or SDK i.e. the target persona is an SME, displaying the UnitExternalId on the UI is more suited.Thanks,Akash

One of the customer would like to see the preview of 2d plot plants or PIDs in Search just like we see the panel for 3d models. 

I hear that the Charts monitoring feature includes logic to prevent excessive and unnecessary alerts.Specifically, time series data uploaded at intervals longer than one minute is excluded from monitoring, and alerts are not triggered for such data.The reason for this is that the system cannot determine whether a lapse in recent values over a certain period is due to “legitimate low-frequency data” or “a system error causing missing data.” On the other hand, we have some time series with very low update frequencies.For example, we have data that is ingested only once every half day (see figure below).With such data, the Charts monitoring feature cannot detect anomalies and alerts are not triggered.*We have no plans to increase the number of acquisitions due to wireless system specifications. Could you consider adding a switch to the Charts monitoring feature that allows users to disable the current specification that excludes data uploaded at intervals longer than one minute from monit

Is it possible to have a set of pre-built templates for various disciplines or generic use cases, which a new user can directly utilize for preliminary work, until he get trained enough to start building his canvas. Some of the users like to use the industrial canvas like a dashboard and for them having a pre-built template gives an idea of how to use the canvas.

We recently did a backup of all configurations for a solution in both our dev and test project. For this we used `cdf dump`. The command gave us alot of options on what to dump. And for each resource, we needed to provide an id. However most of the resources we used were scoped to a cdf-group.So the tool was built with a group as a basis for fetching all the resource id’s and constructing the `cdf drump` commands. So my feature request is to have an ability to dump all resources scoped to a cdf group with the `cdf dump` command. So you would only specify the name of the cdf-group to base the resource dumping on. This would simplify backup and dumping alot

Today, the PI extractor can only write time series instances to a single target space per deployment, even when the underlying PI server contains data for multiple sites or governance domains. The only practical workaround is to run many PI extractor instances against the same PI server, each with different tag filters and a different target space, which is hard to scale and operate. [Governance & spaces; Multi-space limitation]I would like the PI extractor to support multiple target spaces from a single deployment, where the space is selected per time series based on configurable filters. Typical examples would be routing by tag name prefix or pattern (for example, ABU* to space site-abu, RUW* to space site-ruw), or by PI point attributes / metadata mapped to specific spaces. [Filter-based routing idea; Enterprise scaling concern]. This capability would avoid both the operational overhead of many parallel extractor instances.

The OPCUA extractor can only write time series instances to a single target space per deployment, even when the underlying OPCUA server contains data from different sources, governance domains or folders which should be handled individually. The only practical workaround is to run many OPCUA extractor instances against the same OPCUA server/hub, each with different tag filters and a different target space, which is hard to scale and operate. [Governance & spaces; Multi-space limitation]I would like the OPCUA extractor to support multiple target spaces from a single deployment, where the space is selected per time series based on configurable filters. Typical examples would be routing by tag name prefix or pattern (for example, ABB* to space site-abb, VAL* to space site-val), or by attributes / metadata / folder structure mapped to specific spaces. [Filter-based routing idea; Enterprise scaling concern]. This capability would avoid both the operational overhead of many parallel extr

Celanese is asking for a UI that they can see data and views and their fields within CDF. They currently use the data modeling instances page as the UI to see the data but as they migrate to S&R, they do not want to lose that functionality all together (i.e. they mentioned a year is too long to have this functionality). While it was mentioned that the streamlit app is available, they would like to have a counterpart UI to what we have in data modeling for S&R.

In our 360 images, the circles indicating shooting positions are quite large, and when multiple markers are displayed, they tend to obstruct the view.I have several ideas, but the most practical and system-friendly suggestion would be to make the “invisible distance” setting stricter.It would be even better if this parameter could be configured on the system side. Currently, I feel like markers are visible from a very long distance.Other ideas include enabling size adjustments for the circles or adding an option to toggle their visibility on and off. However, I believe the first suggestion would be the easiest to implement.  Note: Due to internal security reasons, we cannot share 360 plant images online. For illustrations, I’m using a black image during loading. Please imagine your own plant with your mind’s eye! 😊