Releases

This article documents the ongoing improvements we're making to the Cognite Data Source for Grafana.

4.6.0 - October 14th, 2026

Breaking changes

• The minimum supported Grafana version is now 10.4.7 (previously 10.0.0). Grafana 10.0.0–10.4.6 is no longer supported; the GraphQL editor is built on @grafana/plugin-ui, which requires it.

Features

• GraphQL: the panel tab is rebuilt on Grafana's editor layout. The query sits beside a Response pane, and Test query sends the query as it will be sent when it runs, interpolation included, and shows the untouched response — so what a query returns can be checked without saving anything. An Examples dialog replaces the page of documentation that used to sit under the editor.
• GraphQL: a new panel query opens on the core data model with a query that runs as-is — the first 10 time series. Saved queries keep their own model and query.
• GraphQL: the panel tab now plots every row, or nested field, that the data model's schema says is a CogniteTimeSeries (or a type extending it) — selecting space and externalId is enough. Selecting type is no longer required; when it is selected, rows that are not numeric are skipped as before. Aliased fields (ts: timeSeries { … }) are recognised by their alias.
• GraphQL: the panel tab has a Label field naming each series from the response, interpolating any selected field with {{field}} — {{name}} @ {{space}}, say. Its tooltip lists the fields the current query offers, and a series reached through a relation can also use fields of the row it hangs off. Left empty, series keep the name they had. A field the response does not carry renders as :field, matching the Time Series tab's labels.
• GraphQL variables can carry whole instances. Select space and externalId on the same object in the query (the rows themselves, or one nested field), then choose the Instance ID value field; the variable then emits a JSON instance reference that can filter on a direct relation, single- or multi-valued.
• Instance references work inside GraphQL queries. In both the panel tab and a variable's query, ${asset} (or $asset, [[asset]]) is written as a GraphQL instance argument ({space: "…", externalId: "…"}, or a list for a multi-value variable, All included), and ${asset.space} / ${asset.externalId} give one part of it. Variables holding anything else interpolate exactly as before.
• The GraphQL panel tab and the GraphQL variable editor are now one editor. Data model, version, query and response are laid out identically, explained by the same tooltips, and built from one shared component, so a fix to either reaches both. The dashboard-variable tab is renamed from Data Modeling to GraphQL to match.
• Dashboard variables: a new GraphQL variable opens on the core data model with a query that runs as-is — the first 10 assets, since a variable is usually picking assets. Saved variables keep their own model and query.
• Dashboard variables: an optional Display text field chooses which field the variable picker shows, independently of the value the variable carries. Left unset, the picker shows the value itself, so what the user sees is exactly what the variable emits.
• Dashboard variables: each tab follows its feature toggle, as the panel tabs do: Assets is offered when asset-centric features are on, GraphQL when the GraphQL feature is on. A variable already saved on either tab keeps that tab even when the toggle is off, so existing dashboards stay editable, and saved variables keep resolving regardless of the toggles.
• Variable queries now run through Grafana's current CustomVariableSupport API. Saved variables are unaffected.
• Records (Alpha): a new query tab for reading records from streams, released as an alpha: it is off by default, is turned on per datasource with the Records toggle under the Core Data Model features, and may change before Beta. Pick a record view and its records come back as rows, with selectable columns, sorting and a result limit. Filters cover every property of the view: a direct relation gets a searchable instance picker when its view names a target, and filter values accept dashboard variables, including a variable holding instance references. Records are read from the dashboard time range, or from the whole stream when the time window is None (not offered on immutable streams, which require one). Request preview shows the exact request sent to CDF, dashboard variables filled in, ready to copy and replay.
• Records: an Aggregate mode returns server-side statistics instead of rows: up to four Group by levels (time histograms and unique values) and up to five metrics (count, average, min, max and sum). With a time histogram the result is one time series per group; without one it is a table.
• Records: unit conversion, to a unit system or per property, for properties whose view declares a unit. Columns and series name the unit their values are in, which can be switched off.
• Records: a Custom time window, bounded by expressions over {{startTime}}, {{endTime}}, {{now}} and the stream's {{maxFilteringInterval}}, for example {{endTime}} - {{maxFilteringInterval}}. Several queries can then each read one slice of a stream whose limit is shorter than the dashboard range.

Bug fixes

• A GraphQL variable whose selected field was an object (such as instanceId { space externalId }) resolved to the literal string [object Object]. Object fields holding an instance reference now resolve correctly, and a failed query — GraphQL errors in the response, or the request itself failing — now fails the variable with the error message instead of leaving it silently empty.
• Time Series: a label token written with spaces inside the braces ({{ name }}) was rendered as an unknown field. Whitespace is now ignored, as it is in every other label.
• GraphQL: typing in the panel tab's Label or Granularity field re-ran the whole query on every keystroke. Both are now applied when the field loses focus.
• GraphQL: a dashboard variable used in the panel tab's Label was never interpolated, unlike every other tab's label. Interpolation is skipped when no label is set, which every query starts out as.
• GraphQL: a data model whose schema could not be introspected made the panel tab throw while building autocomplete. A failed introspection now simply leaves the editor without suggestions.
• GraphQL: an edit the parser rejects is reported beside the editor, and Test query waits until it is fixed; previously the panel kept running the last valid query with no sign that the text on screen was not it.
• Dashboard variables: opening the GraphQL tab on a variable saved as an assets query overwrote the asset expression with the GraphQL text, losing it. Each tab now keeps its own query, so switching between them — and saving from either — preserves both.
• Dashboard variables: an incomplete edit no longer clears a working variable, and a refused save now says why instead of appearing to succeed.
• Dashboard variables: reopening a saved GraphQL variable left the Version dropdown empty until the data model was picked again. Versions are now loaded for the saved model as the editor opens.
• Dashboard variables: choosing a data model or a version is saved straight away, rather than only when some other field happened to lose focus.
• Dashboard variables: the value-field list offered paths prefixed with the query's root field (listCogniteAsset.name), which never matched a result row and silently fell back to another field. It now offers the fields as they appear on a row (name, instanceId.space), and no longer offers pagination fields such as hasNextPage, which describe the page rather than a row.
• An error or warning shown under a query now clears when the query runs again, so it always describes the latest run. Before, it stayed on screen until the query was edited, even after a time range change had fixed the cause.
• The query editor's tab bar no longer ends in an extra, unnamed tab with nothing behind it.
• A query run from Explore could fail with "HTTP header is larger than 8192 bytes". Explore keeps the whole query in the page URL, and Grafana forwarded that URL to CDF as the Referer header, so a long query could push the request's headers past CDF's size limit. The plugin's routes now send an empty Referer instead.
• Closing a query editor subscribed its warning listener again instead of removing it, so listeners piled up each time an editor was opened and closed, and closed editors kept receiving warnings. Each editor now removes both its error and warning listeners when it closes, and follows a change of query letter (refId) without resubscribing.
• The backend is now built with Go 1.27.2, grafana-plugin-sdk-go v0.297.0 (grpc v1.84.0) and golang.org/x/net v0.61.0, which fix vulnerabilities in the Go standard library (net/http, crypto/tls, html/template, net/textproto) and in x/net that Grafana's plugin validator reported.

4.5.2 - August 25th, 2026

Features

• CogniteTimeSeries labels now resolve {{unit}} against the unit the data points are actually returned in: the query's Target Unit when one is selected, otherwise the time series' storage unit
• All CogniteUnit properties are available on the token, e.g. {{unit.symbol}}, {{unit.name}}, {{unit.description}}, {{unit.quantity}}, {{unit.externalId}}
• A bare {{unit}} serializes the full resolved unit, so the properties available for dot notation are discoverable straight from the label
• The CogniteUnit catalog is fetched and indexed once per connection and shared between the Target Unit selector and label interpolation, so unit tokens don't add a request per query

Bug fixes

• The unit catalog fetch now follows pagination cursors, so catalogs with more than 1000 units are fully available in the Target Unit selector and in labels

4.5.1 - May 4th, 2026

Features

• Added support for mixed CogniteTimeSeries types (numeric, string, and state) in the CDM query editor
• Time series type is now displayed as a chip in the selection dropdown
• Aggregates selector is disabled for string type time series
• State time series shows a warning (support to be added in a future release)
• Added dynamic label interpolation for CogniteTimeSeries queries: the Label field now supports {{property}} tokens resolved against the selected instance's view properties at query time
• {{space}} and {{externalId}} tokens are always available; nested paths like {{property.foo}} are supported

Bug fixes

• Tightened unit validation: only structured unit direct relations are treated as resolvable storage units; free-text sourceUnit values are no longer surfaced in the Target Unit field
• Storage unit is now displayed as a badge next to the Target Unit selector instead of inline text next to the search field

4.5.0 - April 27th, 2026

Features

• Added Activities query editor tab for querying CogniteActivity records by resource type (Asset, Equipment, TimeSeries)
• Added CogniteActivity annotations overlay on CogniteTimeSeries charts, displayed as shaded regions with start/end times
• Added automatic numeric time series detection in the Data Models (GraphQL) tab, auto-triggering the CDF datapoints API for matching results
• Overhauled datasource config editor: reorganized into Connection and Features tabs, CDM features now enabled by default for new datasources, CDM and asset-centric (legacy) toggles are mutually exclusive
• Base URL field now strips http(s):// prefixes and trailing slashes automatically to prevent misconfigured proxy URLs
• Query editor dropdowns (CDM Time Series and Activities) now load options lazily on menu open
• Relationships feature is now off by default for new datasources

Bug fixes

• Fixed broken Cognite documentation URLs in annotations editor, query help, variable query editor, and shared constants
• Improved UX copy and tooltip text across config editor and query editors for clarity and consistency

Security

• Updated lodash to v4.18.1 to address CVE-2025-13465
• Updated webpack to v5.104.1 to address CVE-2025-68458

4.4.0 - November 12th, 2025

Features

• Added CogniteTimeSeries Search Tab for improved time series discovery and selection
• Reorganized feature flags UI with master toggles and improved query editor experience
• Added GraphQL variables support with tabbed interface for enhanced data model querying capabilities

Bug fixes

• Updated plugin dependencies to address security vulnerabilities and compatibility issues

4.3.0 - April 24th, 2025

Features

• Updated plugin GUI components to accomodate for Grafana v12 changes
• Preliminary work to enable backend data source in the future

Bug fixes

• Upgraded plugin dependencies

4.2.1 - November 28th, 2024

Bug fixes

• Upgrade plugin dependencies to fix CVEs

4.2.0 - June 17th, 2024

Features

• Adds time zone support for aggregated time series, the user can select the timezone using the Grafana time range options.

4.1.3 - May 7th, 2024

Bug fixes

• Fixed performance issue with parser which occurs when a fairly long custom time series query is used.

4.1.2 - March 14th, 2024

Bug fixes

• Fixed an issue with Events resource timestamp columns where it was previously displayed in epoch format and now it is displayed in human-readable format

4.1.1 - February 14th, 2024

Bug fixes

• Fixed an issue with the custom query where floats with more than 7 digits after the decimal point were causing syntax error

4.1.0 - February 7th, 2024

Features

• Added new time series property unitExternalId
• Added support to filter time series by unitExternalId and unitQuantity in custom queries
• Added support to query data points with unit conversion in custom queries
• Small UI improvement to the query field in the custom query tab

Bug fixes

• Patched plugin dependencies to fix deprecation errors

4.0.1 - November 7th, 2023

Bug fixes

• Patched regressions introduced after the migration to React:
  • Events table activeAtTime filter is now working as before
  • Annotations filters are applied correctly as before
  • Added back hints and examples for annotation query
• Multiple dependencies have been updated to fix security vulnerabilities

4.0.0 - October 12th, 2023

Features

• Migrate Annotation editor from Angular to React
• Bumped minimum Grafana version requirement to v10
• Events are returned in dataframe format

3.1.0 - May 10th, 2023

Features

• Added support for the new version of CDF Data Models (GraphQL)
• Added an option to sort Events table

3.0.1 - April 3rd, 2023

• Patched plugin dependencies
• Fixed the wording for the CDF host paramether in the plugin configuration

3.0.0 - November 21st, 2022

• The connector dependencies has been updated and the connector now requires Grafana v8 or later.
• Flexible Data Modelling is now added as a preview feature. You need enable it on the data source settings. The plugin will visualize tables and time series which has been modelled in Cognite's new Flexible Data Model
• Extraction Pipelines is added as a preview feature. You can now see the latest status of your extraction pipelines and runs inside Grafana.
• Advanced filtering of events now also supports aggregates.
• Variables created with an asset-query can now return externalId and name as value in addition to id. This is useful when using a variable inside FDM graphQl query.

2.6.0 - August 4th, 2022

Kubernetes grafana operator support

• The connector now supports being used with the Kubernetes Grafana operator as the GrafanaDataSource CRD. In the operator you specify the Cognite project in defaultProject parameter and the API URL as the clusterUrl parameter. In the operator you can specify a data source using the syntax below:

datasources:
  - access: proxy
    editable: false
    isDefault: true
    jsonData:
      clusterUrl: westeurope-1.cognitedata.com # other examples are api.cognitedata.com, az-eastus-1.cognitedata.com etc
      defaultProject: <cdf-project>
      oauthPassThru: true
    name: CDF
    type: cognitedata-datasource
name: cdf.yaml

Advanced filter support for events (alpha)

• Advanced filtering support for events have been added as separate input box in the events tab. Enable this in the data source settings.
• API documentation

Relationships updates

• All drop downs in the relationships tab are now sorted and searchable

2.5.0 - June 1, 2022

Relationships support

• The connector now supports the CDF relationships resource type. Read more about relationships at Relationships documentation.
• With relationships support you can now fetch time series based on relationships in the "Time series from asset" tab. Relationships can be filtered on data set, labels and active relationships based on the time filter selected in Grafana.
• A new relationships tab has been added which can be used to utilize Grafana's Node Graph visualization plugin, or a new (alpha) custom visualization plugin which can be found at cognite-grafana-relationships-visualization

Templates support (preview)

• A new templates tab has been added which support the CDF templates feature (preview). Enable this in the data source settings.
• Using templates with Grafana will allow you to dynamically scale your dashboards when adding equipment.
• Documentation: https://docs.cognite.com/dev/concepts/resource_types/templates

2.4.0 - August 20, 2021

Features

• OAuth 2.0 client credentials grant flow to authenticate and authorize against CDF.

2.3.0 - June 1, 2021

Features

• Ability to link variables so that the value of one variable can be used as an input to another.

2.2.0 - April 6, 2021

Features

• Tabular representation of events data.
• Event annotation query shows unfinished events by default.
• Support for OAuth authentication with CDF.
• Option to fetch only the latest data point for the time series.

2.1.1 - November 23, 2020

Bug fixes:

• Support variables in time series by asset query. In version 2.1.0 $variable used as an Asset tag worked for earlier created dashboards, but it was impossible to create a new panel with the same capability.

2.1.0 - November 19, 2020

• The data source requires Grafana version 7.0.6 or above.
• Time series are now saved to dashboards by their externalId where possible. This allows you to switch between CDF projects that contain time series with matching externalIds.
• Support for synthetic time series alignment.
• Rewrote data source GUI to React.
• Updated CDF logo to match light/dark theme.

2.0.1 - August 25, 2020

Bug fixes:

• [Custom query] Option to display legend label as a plain text instead of expression

  Example for query: ts{id=1} + ts{id=2} + 1:

  • User label custom name results in custom name label. (No specific timeseries referenced in the label)
  • User label {{name}} results in timeseries_name_1 + timeseries_name_2 + 1 label. (This works the same way as before)

• [Custom query] Improved error handling

  • HTTP 500 errors messages are visible from the GUI

• [Custom query] Default granularity falls back to 1 second if shorter value is provided

  • Previously, selecting a time span less than 10 minutes resulted in a HTTP 400 error.

2.0.0 - June 29, 2020

General improvements

• A new query language inspired by PromQL and compatible with synthetic time series in CDF. Learn more.
• Support for all filtering capabilities available in CDF API v1 for the respective resource types (assets, events, times series).
• Any future filtering capabilities added to CDF API v1 will be automatically supported by the data source (filters are passed directly to CDF).
• Query expressions resulting in errors now displays error messages in the UI.
• Regular expressions filtering can be used as part of the Query expression. The Filter field has been removed.
• Supports CDF API v1, with no dependency to older API versions.
• New documentation section for the connector, covering installation and administration, upgrades, feature documentation and getting started information.
• Compatibility with Grafana 7.0. Learn more about the new capabilities here.

Time series and custom queries

• You can now access time series that don't have the legacyName attribute populated with the data source.
• Setting Root Asset in the Custom query tab is no longer required. This filter has been removed from the UI and is now available as part of a custom query expression. For example: ts{rootAssetIds=[12335453, 3455566]}.
• Custom query aggregation requests are less likely to trigger rate limiting in CDF API.
• Functions:
  • Support for string time series through the new map function allows you to convert string values to numeric values which can then be plotted by Grafana. For example: map(ts{externalId='pump_29'}, ['OPEN', 'CLOSED'], [1, 0], -1).
  • Use the on_error function to gain control over query calculations resolving with errors, such as division by 0. The function allows you to set a default value for individual data points that resolved with errors. Without the on_error function, the chart shows empty space in those places. Example: on_error(1/ts{externalId='canBeZero'}, 0).
  • Variable length functions: max(x1, x2, ...), min(...), avg(...).
  • The power function has been renamed to pow.
  • These functions are no longer supported: acos, asin, atan, ceil, celing, cot, degrees, floor, log, log2, log10, radians, sign, tan, atan2, mod, truncate, rand, crc32, conv, div.
• These aggregates are no longer supported by custom queries: max, min, count, sum, continuousVariance, discreteVariance, totalVariation.
• Short names for aggregation functions are no longer supported. Use full names instead: average, interpolation, stepInterpolation.
• The timeseries{options} function has been renamed to ts{options}.
• Filtering:
  • New options:isString=false, externalId='my_id', id=123, externalIdPrefix='my_'
  • Relate to asset tree with: assetSubtreeIds=[{id=754173880412890},{externalId="23-TE-96148"}], assetIds=[123, 234], assetExternalIds=['pump_1', 'pump_2'], rootAssetIds=[123, 234]
  • New range filters for create and updated time: createdTime={min=0, max= 1593018651}, lastUpdatedTime={min=1493018651, max= 1593018651}
  • Select time series based on dataSet membership: dataSetIds=[{externalId='Prediction Model A'},{id=123}]
  • path is no longer supported
  • See the full list of supported filters for the ts{} expression.

Annotations

• New filters: assetExternalIds, rootAssetIds, dataSetIds, externalIdPrefix, source.
• Find active events in range with the activeAtTime filter. If endTime is null, the event is active from startTime onwards. The activeAtTime filter will match all events that are active at some point from min to max. Example: activeAtTime={min=1591018651, max= 1593018651}.
• description supports only regular expression operators =~, !~, and !=.
• minStartTime/maxStartTime has been replaced by the startTime={min=0, max= 1593018651} filter.
• minEndTime/maxEndTime has been replaced by the endTime={min=0, max= 1593018651} filter, enabling you to also filter for events without the endTime set: endTime={isNull=true}.
• minCreatedTime/maxCreatedTime has been replaced by the createdTime={min=0, max= 1593018651} filter.
• minLastUpdatedTime/maxLastUpdatedTime has been replaced by the lastUpdatedTime={min=0, max= 1593018651} filter.
• assetSubtrees has been replaced by the assetSubtreeIds=[{id=754173880412890},{externalId="23-TE-96148"}] filter which introduces support for Id/ExternalId input.
• These filters are no longer supported: sort, dir, limit, offset, sourceId.
• See the full list of supported filters for the events{} expression.

Templating

• New filters: parentIds, parentExternalIds, rootIds, dataSetIds, externalIdPrefix, root.
• The description field supports only regular expression operators =~, !~, and !=.
• Select assets based on label: labels={contains={externalId="pump"}}, labels={containsAny=[{externalId="pump_type_A"},{externalId="pump_type_B"}]}, labels={containsAll=[{externalId="pump"},{externalId="rust_detected"}]}
• assetSubtrees replaced by the assetSubtreeIds=[{id=754173880412890},{externalId="23-TE-96148"}] filter which introduces support for Id/ExternalId inputs.
• minCreatedTime/maxCreatedTime has been replaced by the createdTime={min=0, max= 1593018651} filter.
• minLastUpdatedTime/maxLastUpdatedTime has been replaced by the lastUpdatedTime={min=1592222651, max= 1593018651} filters.
• These options are no longer supported: query, sort, dir, offset, boostname, path, depth, sourceId.
• See the full list of supported filters for the assets{} expression.

Other

The Cognite Data Source for Grafana 2.0.0 is compatible with CDF API v1.

Cognite Data Source for Grafana 2.0.0 is not backward compatible with dashboards created using older versions of the connector. Follow the steps in the upgrade guide to upgrade your existing installation.

1.0.1 - May 27, 2019

• Rename Cognite Data Platform to Cognite Data Fusion.
• Allowing for templating variables to be used anywhere in custom queries.
• Allowing for the base url to be modified when adding a new data source.

Other

The Cognite Data Source for Grafana 1.0.1 is compatible with CDF API 0.5.

1.0.0 - March 26, 2019

• Fixing templating so that repeated panels/rows work properly.
• Changing checkboxes to Grafana checkboxes and adding Select All option.
• Fixing data source configuration to verify the project name.
• Breaking change: function= inside of timeseries{ ... } no longer works for custom queries. Instead, functions are now applied outside of timeseries{}[] such as:
  • (timeseries{description=~".*TEMP.*"}[avg] - 32) * 5/9
  • sum(timeseries{name=~".*VALUE"}[count,1d])
  • max(timeseries{metadata.on="True"}[count,5m]) - min(timeseries{metadata.on="True"}[count,5m])

Other

The Cognite Data Source for Grafana 1.0.0 is compatible with CDF API 0.5