Product User Community

I follow the Docments for binding files in CDF(client.files.upload_content)After uploading, AtlasAI is indeed able to use the content.However, this file seems to be a chunked TXT document generated from the customer document. I believe this TXT file should follow a certain structure, but I could not find any documentation describing the expected text content structure in the help docs.Is this an industry-standard format, or is the file_content structure arbitrary?It seems that the completeness and quality of the chunk file directly determine the upper limit of AtlasAI’s intelligence and response quality. 

Hello everyone ! Since this morning (3AM Paris time ) we encounter serious problem with our MQTT hosted extractor. Setup- CDF region: West Europe 1- 1 Source (Azure Event Grid MQTT broker, port 8883, TLS cert auth) controlled externally- 1 Destination (single sink, single CDF session)- 2 Jobs (topic filters), both using the same source and same destination:  - xxxx/OUT/xxxxxxx/AA/#  - xxxx/OUT/xxxxxxx/BB/# ProblemSince this morning (~03:00 UTC+1), both topic filters are stuck in connection_error with the following error looping in the event log:Could not establish connection: Mqtt state: Server sent disconnect with reason 'None' and code 'SessionTakenOver'Both jobs were running fine until today. No configuration changes were made on our side. What we tried1. Paused one job to isolate — the other still fails alone2. Paused both, waited 10 minutes, resumed — connects briefly then gets kicked again3. Deleted 10 legacy/unused sinks and 1 old paused job (cleanup) — no change4. Checked Azure

When hovering over the dropdown, the legends are lingering making it hard to find and click on correct time series. Happens when creating calculations in Charts. 

Team, I am trying to schedule math functions, specifically addition of 4 tags and averaging of 4 tags.These need to run specifically at 10:05am as these are totals that are reported once a day, at approximately 10am EST.How can this be done and make these values available for use in Grafana? Thanks for helping!

KepServerEx 6.10 OPC UA Configuration Manager : KepServerEx 6.10 OPC UA Node ID :config.yml :source:    endpoint-url: opc.tcp://127.0.0.1:49320      #This is your OPC UA server url    auto-accept: true    queue-length: 10    username:     password:     browse-chunk: 1000    attributes-chunk: 1000    x509-certificate:            browse-throttling:        max-node-parallelism: 10   extraction:    id-prefix:     # Delay in ms between each push of data points to targets    # Alternatively, use N[timeunit] where timeunit is w, d, h, m, s or ms.    data-push-delay: 5000        # Source node in the OPC-UA server. Leave empty to use the top level Objects node.    # If root-nodes is set, this is added to the list of root nodes.    root-node:        # Full name of the namespace of the root node.        namespace-uri: "Simulation"        # Id of the root node, on the form "i=123" or "s=stringid" etc.        node-id: "s=Simulation"            # List of proto-node-ids similar to root-node.    # The

I had the honor of attending NVIDIA GTC 2026 alongside our Principal AI Architect, Peet Cremer, our Head of Cognite Atlas AI, Justin Lawyer, and our SVP of Partners & Alliances, Trygve Ronnigen. Between Jensen Huang's visionary keynote and the buzzing energy at the Microsoft Pavilion, one thing is clear: we aren't just talking about "cool AI" anymore - we are talking about AI that works in the physical, messy, and complex world of heavy industry.Our team spent the week diving deep into the next generation of compute and intelligence. Here are the core insights we’re bringing back from the ground. Key Takeaways from the GTC FloorThe pace of innovation is accelerating, and NVIDIA's latest announcements are shifting the goalposts for what's possible in industrial AI:Security for the Agentic Era: As we move toward autonomous AI agents, security is non-negotiable. NVIDIA's launch of NemoClaw, featuring the OpenShell sandbox and enhanced privacy, is a big step forward in making AI safe f

I am not able to login/access CDF workspace in the last few days using my work email id which I used for registering. I am able to access the Cognite courses though. I was able to access workspace during early days of last week.PFB error message I received during my attempt to workspace accessAADSTS90072: User account 'sivaramat@virtusa.com' from identity provider 'https://sts.windows.net/0d85160c-5899-44ca-acc8-db1501b993b6/' does not exist in tenant 'Cognite Hub' and cannot access the application '9f039f43-cd64-4d32-abc1-333ca411e5f9'(Cognite Data Fusion) in that tenant. The account needs to be added as an external user in the tenant first. Sign out and sign in again with a different Azure Active Directory user account.

I have a data model in CDF (for sales/demo purposes) that is in SpaceA and i populate with data in “CustomerASpace”, “CustomerBSpace” etc.I’ve used this approach for only 1 customer and has worked well. (used “cornprocessing” space)Today I populated data (new assets/timeseries) assigned to a different space (ketchupspace).I use a custom streamlit app to ingest CSVs and write to the pre-existing model.The new entities can be seen in both the streamlit app and the data model UI directly. However these new entities do not show in Search using a new Location pointed to the data model and “ketchup” space. If I switch the location in Search to the previous/1st location, it works fine.Any ideas are appreciated. If this is not clear and you need more info, let me know.Chris

When using the industrial tools → search functionality. it’s not so untutitive that apply is not clickable when i want to remove filters on datamodels and search in all models.One would expect the user to be able to apply, cancel implies i do not select a filter on “all” and instead go back to my previously selected filters

Hello, Can someone help with the below issue? Customer installed the FT DataMosaix SQL Extractors last year and the subscription ran out in November.The customer purchased a new entitlement in February. For some reason my extractors have not been able to push data from the site SQL server to the Cognite RAW database since Feb. 8th 2026.What can I do to correct this connection problem?Please see the attached file word doc with the customer’s ODBC check results. Regards,Henwill BalladaresRockwell Automation    

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

I'd like to inquire about the current status of the monitoring functionality in the workflow orchestration. Specifically, I'm interested in the setup for notification emails. It would be helpful to know if we are able to receive notifications in the event of a workflow failure. Could you please provide an update on this?

Working with a hybrid model and creating node mappings using the following API:https://api-docs.cognite.com/20230101/tag/3D-Asset-Mapping/operation/create3DMappingsImplemented the CreateAssetMapping3DDmsBody to create the node mappings.However, when attempting to retrieve the list of node mappings along with their DMS instances, the response consistently returns an empty list. (Get 3D asset mappings, with query parameter getDmsInstances=true)Has anyone encountered this behavior or knows why mappings created through this endpoint are not appearing when queried? Any insights or guidance would be greatly appreciated.Thank you!

Innovation happens best when everyone has a voice. Instead of relying on one person to funnel feature requests, Cognite Hub solves the bottleneck by enabling all eyes in your organization - or our global community, depending on your preference - to tap in. Whether you are an admin, a developer, or a daily end-user, this is your official space to share frontline insights and directly influence the Cognite Data Fusion product roadmap. Why Submit Your Ideas? Public Ideas: The Product Ideas board is your dedicated space to share requests with the Cognite product team and the community. Submitting publicly allows others to validate, vote on, and build upon your idea! Private Ideas: Submitting privately ensures that feature requests tied to your unique business goals, confidential internal initiatives, or proprietary workflows stay strictly between your organization and Cognite. Track Your Impact: Want a top-level view of your contributions? Upon request, the community team will share an

Hi! We are building a Bayesian model using PyMC with a JAX backend, and want to run this model continuously on new incoming data inside a Streamlit App hosted in Cognite Data Fusion. As the library depends heavily on compiled extensions, there are no pure Python wheel for this. Is there an alternative approach for running PyMC models?Thanks!

Hi,I tried to create timeseries using SDK and i want to create using the instanceID as below: but i got this error:What is the correct way to create timeseries using instance? Thanks

I have a problem with retreiving data with the Python SDK. In a time series I have a data point for 2026-03-11 13:45:00 with the value “0.0”. If I retrieve it without any aggregation, it is (as it should be) “0.0”. If I retrieve it aggregated to a 15min resolution with “average” aggregation, it becomes “0.5”. Both “max” and “min” aggregations however give “0.0” - as should be expected. So there seems to be something off with the “average” aggregation. See screenshot for the case:  Is this a known behaviour?

Button to close window disappears when browser window is too small. It is quite inconvienient that the “close” button is one that disappears. Is this by design or a UI bug?

Hi everyone, At Cognite, we strive to ensure that every tool in the Cognite Data Fusion workspace meets our high standards for security, performance, and user experience.The Image and Video Management interface (found under the Data Management workspace) has reached a stage where it can no longer be maintained to those standards. To focus our efforts on more robust, integrated data exploration tools, we are initiating a 12-month deprecation period for this specific interface. This means that the Image and Video Management interface will be deprecated on 17 February 2027.What is changing?The standalone Image and Video Management app is entering its sunset phase. This includes: The specialized Map view for finding image and video data. The Annotation tool for images. The integrated Video streaming/viewing interface within this specific app. Note: This deprecation only affects the management interface. Your raw image and video files stored in CDF remain safe and accessible via the API

Cognite documentation powers human and agent workflows with new AI functionalityCognite now offers several ways to bring our official CDF product documentation into your AI-powered workflows.A docs MCP server for live documentation access in your IDE.  SKILL.md capability files so AI assistants understand what Cognite Data Fusion (CDF) can do. An integrated experience on docs.cognite.com that includes an AI assistant that answers questions and cites sources, plus one-click options to send any page to Cursor, VS Code, ChatGPT, or Claude.To use the MCP server and the SKILL.md file, you'll need a compatible AI-powered IDE or assistant (for example, Cursor, Claude, Windsurf, or VS Code with Copilot). To install the SKILL.md file with the skills CLI, you also need Node.js. Docs MCP server: live documentation in your IDEThe Model Context Protocol (MCP) is an open standard that lets AI assistants securely access external data and tools. Cognite's docs MCP server gives your IDE live access to

On the platform, when we have more than one annotation on a file, if they are very close to each other (overlapping by a pixel or two), the platform shows the overlapping box so we can select the correct annotation.I understand this as a bug.See the images below:The tags PB2885C and XA2885C are very close to each other. The overlap between them is barely noticeable.However, when you click on either one, both tags appear in the overlapping box:I’ve seen this happen every time we have annotations that are very close to each other.For a better user experience, it would be ideal if, at the moment of the click, the platform could understand that only one annotation is actually inside the click area.This would prevent the overlapping box from appearing and would solve the issue.Is this something we can have fixed?

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 have the power to curate what your users see when they sign into 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 Start Guide: Check Access: Ensure you have appconfig:read/write and app-scope = Search. Your users also need appconfig:read to make the config

Dears, How can we get the filled checklist in Infield exported for example in excel format for reporting purposes or archiving, InField's filled Checklist reports (including messages, Notes, and checkboxes) as excels is there a way to have it though an API, or is this feature coming soon? Appreciate you help in this since it’s urgent need.  Thanks!Fatima  

I have a bit of a conundrum here. I am trying to visualize the daily and weekly averages of a time series in Azure Managed Grafana v11.6.3 using the CDF datasource. Raw data is strictly on hourly resolution, and provided in CET/CEST. Grafana is set to browser tz (i.e. CET/CEST). I need the daily average to align with CET midnight (=23:00 UTC). I use these synthetic series parameters: aggregate='average', granularity='24h', alignment=82800000 (82800000 aligns the datapoints to the correct hour if I do aggregate='step_interpolation', granularity='1h'). The returned value is neither the UTC day average or the CET day average. It is exactly the average of the two. So from what I see it seems to be a 30-minute shift/interpolation in the backend. Case in point: CET day average for a specific day is 158,67, while its UTC equivalent is 159,83 (see screenshot). Neither value is found in Grafana for that day at 00:00 - here I see 159,25. Which coincidentally is equal to (158,67+159,83)/2. The ti

As part of the hands on part of the CDF Fundamentals course, I have parsed the pdf but cannot preview it or see its contents when adding it to canvas. After parsing I could also not see the Mapping tab and hence also not “verify” the matches in the diagram. Can anyone assist?