> For the complete documentation index, see [llms.txt](https://docs.lumidb.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.lumidb.com/release-notes.md).

# Release notes

What changed in LumiDB.

## November 2026

**Highlights:** imagesets for geotagged photos, project-scoped API keys, much faster ArcGIS streaming for very large datasets, and QGIS streaming without timeouts.

### 17 November 2026

#### Integrations

* Fixed: WMTS and ArcGIS links for rasters with spaces or special characters in their name now work in the ArcGIS Online Map Viewer. The links use the raster ID, and the **ArcGIS link** uses the `/rest/services/<name>/MapServer` form that the Map Viewer requires. If a link didn't work before, copy it again from **Share**. See [ArcGIS](https://docs.lumidb.com/integrations/arcgis.md).

#### Platform

* Access changes take effect immediately. Granting or removing project access and removing an organization member no longer lag behind. See [Security and hosting](https://docs.lumidb.com/archive-management/security-and-hosting.md).

### 16 November 2026

#### Data

* New: imagesets. Import collections of geotagged JPEG or TIFF photos, such as the photos of a drone flight, in the app (**Import** → **Images**) or with `POST /imagesets/import`. LumiDB reads the position and camera direction from each photo, shows the camera positions in the 3D view, lets you browse the previews and download the originals. See [Import photos as imagesets](https://docs.lumidb.com/import-data/import-photos-as-imagesets.md).

#### API

* Fixed: requesting points as LAZ in a coordinate system other than the dataset's no longer crashes the server. Points that can't be encoded now return an error.

### 12 November 2026

#### Viewer

* The custom basemap dialog no longer offers **Project raster**. To get the same result, hide the basemap. You can still enter a custom URL template. See [Visualization modes, display settings, and performance](https://docs.lumidb.com/explore-and-analyze/visualization-modes-display-settings-and-performance.md).

### 10 November 2026

#### Integrations

* Fixed: streaming a point cloud to QGIS no longer crashes.

#### API

* API keys can be limited to one project. When you create a key, choose a project under **Project access**, or leave **All projects**. A project-scoped key reaches only that project and has no organization-level rights. The **API keys** page now shows which project each key can reach. See [API keys](https://docs.lumidb.com/archive-management/api-keys.md).

### 5 November 2026

#### Data

* The import dialogs are simpler. The custom metadata fields are gone, each setting has its own card, and **File heights are ellipsoidal** moved under **Advanced options**. See [Importing data](https://docs.lumidb.com/import-data/importing-data.md).

#### Integrations

* Streaming to ArcGIS is much faster for very large datasets, including multi-terabyte datasets and large clips, in ArcGIS Online and ArcGIS Pro. LumiDB keeps building a scene layer in the background when a request times out, so a retry from ArcGIS doesn't start the build again. Requests that use a classification filter are still slow on large areas. See [ArcGIS](https://docs.lumidb.com/integrations/arcgis.md).
* Layers in ArcGIS map services, WMTS capabilities and I3S scene layers now use the display name of the dataset. WMTS output no longer contains characters that are invalid in XML.
* Streaming large point clouds to QGIS no longer times out. LumiDB sends the data in pages as the client explores it, so the point counts QGIS shows are estimates. Older QGIS versions no longer flag the coordinate system as invalid. See [QGIS](https://docs.lumidb.com/integrations/qgis.md).

#### API

* Organization admins can download an original file again with `GET /assets/{asset_id}/download`. It returns a link that works for one hour and keeps the original file name. It also works for files that no dataset uses.

#### Platform

* The web app shows a cookie consent banner. Analytics, session recording and the LinkedIn ad tag run only after you accept. Open **Help** → **Cookie preferences** to change your choice.
* Free links from the public upload page now last 7 days instead of 3.

## October 2026

**Highlights:** Shapefile import, export in any projection, automatic coordinate system detection, Re-process, the MCP server and the public upload page.

### 30 October 2026

#### Data

* Importing a point cloud or raster with a name that already exists no longer overwrites the existing dataset in the app. LumiDB stores each import under a unique name with a date and time added, and shows the name you typed as its display name. An import through the API with the same `table_name` still replaces the dataset. See [Import data with the API](https://docs.lumidb.com/import-data/import-data-with-the-api.md).

### 29 October 2026

#### Viewer

* Basemaps line up correctly for datasets in some coordinate systems, such as OSGB36 / British National Grid. Existing datasets are corrected without importing them again.

#### Data

* Importing LAS and LAZ files is faster, because LumiDB reads large files in parallel. Import progress is counted in points instead of files, so the percentage is more accurate. COPC LAZ, E57 and ASCII files are still read one after another.
* GeoTIFF rasters with ZSTD compression now import.

#### Integrations

* The MCP server has a new `compute_volume` tool. It calculates cut, fill and net volume between a point cloud surface and a base surface inside a box or polygon. See [Connecting AI assistants (MCP)](https://docs.lumidb.com/integrations/connecting-ai-assistants-mcp.md).

#### API

* The command-line tool works against remote LumiDB servers and uploads local files for you when you import. It has new `assets list` and `assets upload` commands, and shows upload progress and a timing summary.

{% hint style="warning" %}
**For developers:** in the command-line tool, `datasets` is now `pointclouds`, and point clouds and rasters are addressed by dataset ID instead of name. The projection switches are now `--source-crs` (`-s`) and `--target-crs` (`-t`). The old names still work as aliases. Imports now use strict mode by default, so a file without a coordinate system fails instead of being assumed to be in the target projection.
{% endhint %}

#### Platform

* LumiDB has a public upload page at `app.lumidb.com/upload`. Anyone can import a point cloud without signing in. When the import is ready, LumiDB emails a link to the project. Opening the link confirms the email address and signs the person in.

### 26 October 2026

#### Integrations

* The `render_point_clouds` MCP tool shows the rendered image inline in assistants that support interactive tool views, such as Claude.

### 23 October 2026

#### Viewer

* Datasets you rename show their new name everywhere: on the **Project data** page, the map, the sidebar, in dialogs and in confirmation prompts. Search on **Project data** finds datasets by these names.

#### API

* `GET /datasets` returns a `display_name` field.

### 21 October 2026

#### Data

* The import dialogs read the coordinate system from LAS, LAZ, E57 and GeoTIFF files. When all files agree, **Input projection** is filled in for you. When they don't, each file shows its own coordinate system and a warning appears. You can still enter another projection. See [Coordinate systems and projections](https://docs.lumidb.com/import-data/coordinate-systems-and-projections.md).

### 19 October 2026

#### Integrations

* LumiDB has an experimental read-only MCP server at `/api/mcp`. AI assistants sign in with an API key and use three tools to find datasets, read their metadata and render point clouds. See [Connecting AI assistants (MCP)](https://docs.lumidb.com/integrations/connecting-ai-assistants-mcp.md).

#### API

* `POST /render` draws a point cloud to a PNG image from a camera and projection you choose, without downloading the points. It supports point clouds with RGB or gray color, perspective and orthographic cameras, and images up to 4096 pixels.
* You can send an API key in the `apikey`, `api-key`, `x-apikey` or `x-api-key` header, as well as in `Authorization: apikey`.

### 14 October 2026

#### Data

* You can cancel imports and exports that are running. A cancelled task has the status `cancelled`, and progress responses include a `phase` such as `Calculating statistics`.

#### API

* The home page has an **API keys** page where you create, list and revoke your own API keys. Keys have a name and an optional expiry date, and a revoked key stops working. See [API keys](https://docs.lumidb.com/archive-management/api-keys.md).
* Uploading an asset accepts optional `content_sha256` and `content_length` fields. Calling `/assets/{asset_id}/complete` is optional for single uploads. Asking for a new upload URL returns `409` once the file has been uploaded or after the 24-hour upload window.

### 9 October 2026

#### Data

* An export with datasets from more than one project is rejected with a clear message.

### 8 October 2026

#### Integrations

* Fixed: scene connector links from the **Tileset** dialog, and volume measurements, work when a raster or vector dataset is also visible.

### 7 October 2026

#### Viewer

* Fixed: corrupted terrain heights on mobile Safari.

#### Data

* An import that fails because the points are far apart, which is often caused by a wrong coordinate system, now shows a clear explanation.

### 5 October 2026

#### Viewer

* The web app loads faster. It first fetches a short list of your point clouds, rasters, vectors and attachments, and loads the full details when you open the project, point at a dataset or select it.

#### API

* `GET /api/datasets` returns the list of your point clouds, rasters, vectors and attachments.

{% hint style="warning" %}
**For developers:** the `/api/rasters` list endpoint is replaced by `/api/datasets`. In the npm library, `listRasters`, `getRasterMetadata` and `listAttachments` are removed. Use `/api/datasets` and the metadata endpoints that take a dataset ID.
{% endhint %}

### 2 October 2026

#### Viewer

* The **Project data** page shows the file size of rasters, and the project and total sizes include rasters and attachments.
* Dataset, raster, attachment, measurement, feature collection and group panels share one layout: an editable name and description, a collapsible **Custom fields** section and values you copy with a click. The projection shows its unit, and its full name when you point at it. See [Browsing datasets](https://docs.lumidb.com/explore-and-analyze/browsing-datasets.md).
* A link shared with **My current view** now carries only the camera position. The person who opens it sees that spot with the default view settings of the project.

{% hint style="warning" %}
**For developers:** on feature collections, the field `bbox` is now `bounds_wgs84`. Update clients that read `bbox`.
{% endhint %}

#### Data

* You can import Shapefiles. The new **Vectors** dialog accepts loose `.shp`, `.dbf`, `.prj` and `.cpg` files, reprojects them to WGS 84 with the coordinate system of the file or a fallback you choose, shows a map preview and saves a feature collection. LumiDB doesn't keep holes in polygons. See [Importing data](https://docs.lumidb.com/import-data/importing-data.md).
* Exports as LAZ, E57 and glTF open one dialog where you choose the output projection and see the estimated file size. The export limit is now 10 GB, or about 4 GB for glTF, up from 5 GB. The local offset option is gone. A grayed-out export button shows a tooltip that explains the limit. See [Merging, cropping, and filtering datasets](https://docs.lumidb.com/explore-and-analyze/merging-cropping-and-filtering-datasets.md).
* Point cloud and raster panels have a **Re-process** action. It imports the dataset again from its original files with the current import pipeline, and the existing version stays available until the new one is ready.
* The web import dialog accepts ASCII point cloud files (`.xyz`, `.asc`, `.pts`, `.txt`).
* Uploads in the import dialogs are more robust. A failed upload retries for about 7 minutes with a fresh upload URL each time, and the page warns you before you close it during an upload. **Import** stays disabled until every file has uploaded. Vector imports also store the original files with the feature collection.
* Fixed: GeoTIFFs that only have a world file, and ArcASCII grids, import with their `.prj` file or the fallback projection. Mosaics in coordinate systems with northing before easting, such as the Finnish GK zones, are no longer placed with swapped coordinates. Some E57 files that other tools wrote, for example with PDAL, now import.

#### API

* `POST /api/tables/{name}/reimport` and `POST /api/rasters/{name}/reimport` do the same as **Re-process**.
* `GET /assets/{asset_id}/upload-url` returns a new upload URL, and you can call `/assets/{asset_id}/complete` more than once safely.
* Import manifests no longer need a `proj` for each input. LumiDB reads the coordinate system from the file. A new `input_proj` covers files that declare none, and `strict_mode` makes a file without a coordinate system fail the import.
* Table responses include `bounds_wgs84`, and raster responses include `table_proj` and `proj4_string`.
* `POST /search` returns the point clouds and rasters you can access whose footprint overlaps a WGS 84 bounding box. Results aren't paged, and a box that crosses the 180° meridian must be sent as two searches. See [API quickstart](https://docs.lumidb.com/api-reference/quickstart-your-first-api-request.md).

#### Platform

* Project members and share links can now list, view and stream rasters only from projects they can access.
* Guests with a company email address whose domain has no admin in the organization see a prompt to set up their own GIS archive. One click creates a new organization with them as admin.

## September 2026

**Highlights:** the organization **People** page and clearer invitations, streaming clips to QGIS, richer dataset metadata with AI summaries, and a guided product tour.

### 17 September 2026

#### Viewer

* Large point clouds use less memory while you pan and zoom. **High** now draws at most 10 million points instead of 20 million, **Medium** 7.5 million and **Low** 5 million. Detail on **High** is somewhat coarser.

#### Data

* In the import dialogs you can find a coordinate system by its name or alias, for example `TM35FIN` or `YKJ`, and see suggestions as you type. The new CRS endpoints also return the area of use of each coordinate system.
* Fixed: an E57 import no longer hangs on a file with unreadable point records. It stops with an error message.

#### Platform

* Project invitation links open a landing page with the project name, the organization logo and a **Sign in** button. Email link scanners, such as Microsoft Safe Links, can no longer use up the one-time invitation ticket before the person clicks. Invitation emails show the project thumbnail and stay valid for 365 days instead of 30.

### 10 September 2026

#### Viewer

* New users get a guided tour the first time they open a project. It asks you to zoom, pan and rotate, and then points out the **Layers** panel, the details panel and the toolbar. It replaces the old mouse controls dialog. Replay it from **Help** → **Restart tour**. Fixed: clicking a notification now closes it.

#### Data

* Imports create a short AI summary of each dataset, with a guessed place and capture method. The dataset dialog shows them in a **Summary** block, marked as generated by a model. If creating the summary fails, the import still completes. Raster dialogs now show the same details as point clouds: resolution, footprint area, data coverage, capture dates, megapixels, compression, source resolution and band statistics. You can download the metadata as JSON.

{% hint style="warning" %}
**For developers:** point cloud metadata fields that were inside the `derived` object are now at the top level of the table metadata, and the `summary` block is renamed `inferred`, with `text` now `summary` and `place` now `location`. Update clients and SDK code that read these fields. The `status_url` returned by `POST /tables/import` now contains the table version ID, so polling it works.
{% endhint %}

#### API

* Point cloud imports can use PROJ transformation and geoid grids that PROJ doesn't include. Upload the grid file as an asset, list it in the `custom_grids` map of the manifest, and use its file name in the `proj` of an input.

### 4 September 2026

#### Viewer

* Points behind the basemap terrain fade and lose their edge shading, so depth is easier to read. The **Terrain offset** slider accepts negative values, and the default terrain sink is now −5 m instead of −10 m.

#### Data

* Point cloud datasets get richer metadata at import: density, mean point spacing, footprint area, capture dates, height, scan angle and intensity ranges, number of flight lines, classifications and details about the source files. The dataset dialog shows a summary and expandable details, and you can export them as JSON.

#### Integrations

* You can stream point clouds, including the result of a clip with several datasets and filters, to QGIS as Entwine Point Tiles (EPT). See [QGIS](https://docs.lumidb.com/integrations/qgis.md).

#### Platform

* Organization admins and members have a **People** page under **Admin**. It lists everyone in the organization with their role, status, last seen time and number of projects, and has a search box. Admins can change a person's role between **Admin**, **Member** and **Guest**, except their own, and remove people. Removing a person also removes their invitations and project access. See [Administer your organization](https://docs.lumidb.com/archive-management/administer-your-organization.md).
* Invitations go through the organization invitation flow. Pending invitations appear in their own **Invitations** section in the project members menu, and LumiDB revokes them when the person loses their last project access. Project managers can resend an invitation, and the link opens the project after sign-in. The role options in the menu have descriptions. See [Sharing projects and inviting collaborators](https://docs.lumidb.com/sharing-and-access/sharing-projects-and-inviting-collaborators.md).

## August 2026

**Highlights:** hillshade, contours and opacity for rasters, a raster link for ArcGIS, file attachments, and vertical datums with correct height units.

### 26 August 2026

#### Viewer

* Each raster has an **Opacity** slider from 0 to 100 percent. The raster properties have a **Rendering** menu with **Default**, **Hillshade** and **Contours**. Contours draw lines at an interval you choose, in a color you pick, with every fifth line heavier. The hillshade controls are now a knob and sliders without typed values.
* Rasters can be placed on either side of the **Compare** slider, next to point clouds, and the **Compare** dialog lists them.

#### Data

* LumiDB transforms heights properly. A coordinate system can name a vertical datum, for example `EPSG:3067+3900` or `+ellipsoidal`, and the server uses PROJ datum and geoid grids. Datasets in feet no longer come out about three times wrong in measurements, cross sections, queries and Cesium, glTF and ECEF output. The point cloud import dialog has a **File heights are ellipsoidal** option, and the viewer shows distances, areas and volumes in the unit of the dataset. See [Coordinate systems and projections](https://docs.lumidb.com/import-data/coordinate-systems-and-projections.md).

### 13 August 2026

#### Data

* You can attach any file to a project, for example a drawing or a report, and download it later. The file appears as its own layer in the sidebar with its name, size, lock, move and delete actions. The API can upload, list, rename, tag, move and create download links for attachments.

### 10 August 2026

#### Integrations

* Rasters have an **ArcGIS link** in the raster **Share** menu. It serves Web Mercator tiles. Paste it into ArcGIS Online with **Add layer from URL**, or into ArcGIS Pro with **Add Data** → **From Path**. Send the API key as a custom parameter, or as the username, instead of setting up a WMTS connection. See [ArcGIS](https://docs.lumidb.com/integrations/arcgis.md).

### 6 August 2026

#### Viewer

* Polygon fills and area measurements are correct for sloped and vertical polygons. Tiled feature collections keep the full geometry of features with changing elevation.

#### Data

* Elevation rasters have hillshade. The raster properties have a **Hillshade** section for the light direction, the sun elevation and the exaggeration.

{% hint style="warning" %}
**For developers and QGIS users:** the XYZ and WMTS links of elevation rasters now return Terrarium-encoded values, not a grayscale picture. External clients such as QGIS show them differently. To get the real values, use the **COG link**.
{% endhint %}

## July 2026

**Highlights:** a new 3D terrain, GeoJSON stored as feature collections, snapping between features, and support for ArcInfo ASCII grid rasters.

### 27 July 2026

#### Viewer

* The data export flow is redesigned. Vector styling has more controls: color and opacity, marker sizes and feature title labels, with a short live preview when you edit a selected feature or collection. Crop regions have a striped fill, and imported shapes hide their markers by default.
* While you draw or edit a feature, you can snap its nodes onto the nodes and lines of nearby features. A point snapped onto another feature takes that feature's dataset.

#### Data

* GeoJSON you import can be stored as a feature collection. LumiDB sends it to the map as tiles that load and thin out by zoom level, so large collections stay smooth. You can list, create, update, hide and delete collections through the API, and lock their features.
* You can import ArcInfo ASCII grid (`.asc`) rasters in the web import dialog and the command-line tool. When an `.asc` grid or a raster with only a world file has no coordinate system, LumiDB applies the input projection you choose.

### 9 July 2026

#### Platform

* Organization guests can sign in and view the resources they have access to. Guests can't create projects.

### 7 July 2026

#### Viewer

* Fixed: rasters zoom in fully instead of stopping at the maximum zoom of the basemap, and they are visible when there is no basemap.

### 6 July 2026

#### Viewer

* The 3D terrain is rebuilt. LumiDB now uses an elevation model from MapTiler terrain tiles and drapes one shared surface under the basemap and all raster layers, without gaps. The 3D terrain toggle, the z-offset and the custom basemap URL work as before. Terrain sits 10 m below other data at the same height, so it stays under point clouds, and tiles no longer flash or fight with each other at the seams.
* Annotations support the simplestyle opacity and marker size properties, and projects with many annotations draw faster.

#### Data

* Coordinate transformations use PROJ again. This fixes missing EPSG definitions and datum shifts that put data from non-WGS 84 sources 100 to 200 m off when you imported it.

#### API

* A bulk endpoint creates many features in one request.

#### Platform

* Authorization is checked in one place for every kind of sign-in: sessions, API keys, share links and table tokens. Access across organizations and read or write rights beyond what a key or link should have are now denied. Share links and table tokens reach only their project or the datasets they name.

## June 2026

**Highlights:** elevation rasters and the COG link, the new **Layers** panel, ArcGIS Pro **Extract LAS**, and rasters imported through GDAL.

### 30 June 2026

#### API

* A new command-line tool that talks to LumiDB over HTTP replaces the old one.

### 24 June 2026

#### Viewer

* The **Layers** panel is rebuilt. You can drag layers to reorder them, put them in groups, and show, hide or lock each layer. LumiDB saves all of this per project. You can show several rasters at the same time. Datasets that aren't in a project are hidden from people who aren't admins. See [Browsing datasets](https://docs.lumidb.com/explore-and-analyze/browsing-datasets.md).
* **Filters** has a **Height above ground** range slider. It appears only for datasets that have a height above ground range.

#### Data

* Raster imports now use GDAL. LumiDB reads more input formats, and rasters with an alpha channel.
* Uploads in the import dialogs are more resilient. File parts upload in parallel, failed requests retry with increasing delays, stalled transfers are stopped and retried, and uploads pause while your browser is offline. The dialogs show how many times a file failed, with a **Retry** button. Canceling or removing a file stops its requests.

{% hint style="warning" %}
**For developers:** the `band_count` field is gone from raster metadata responses. When you import individual raster files, list their sidecar files (`.prj`, `.wld`, `.tfw`) in the manifest yourself.
{% endhint %}

#### Platform

* API keys can belong to a user, with permissions that follow the user's organization role. A member's key can no longer write, export, create tokens or manage projects unless the member has a project role that allows it. Fixed: an organization admin could open projects in other organizations by knowing the project ID.

### 11 June 2026

#### Integrations

* The ArcGIS Pro **Extract LAS** tool works on published scene layers. The point counts in the scene now match the points LumiDB serves, including with a crop or classification filter. First requests on a filtered scene can be slower while LumiDB counts and caches. See [ArcGIS](https://docs.lumidb.com/integrations/arcgis.md).

### 10 June 2026

#### Data

* The raster properties panel has a download button for the original files, as point clouds already had. It needs export permission on the project. The raster metadata includes details for each source file: projection, bounds, dimensions and size.

{% hint style="warning" %}
**For developers:** the raster download endpoint now returns one file at a time, by file ID.
{% endhint %}

#### Platform

* A task that is still processing after 15 minutes because its worker died is marked as failed.

### 8 June 2026

#### Data

* Elevation rasters (DEM, DSM and DTM) import and display as height data. LumiDB stretches the values between their minimum and maximum, handles NoData at edges and cliffs, and draws the basemap without banding.

#### Integrations

* The raster **Share** menu has a **Cloud-optimized GeoTIFF (COG) link**. Use it to stream the real values, such as elevation, into QGIS and other GDAL tools instead of 8-bit tiles. The link works for 12 hours. See [QGIS](https://docs.lumidb.com/integrations/qgis.md).
* I3S point cloud layers support **Query** and **Extract** in ArcGIS.

### 4 June 2026

#### Data

* Imported GeoJSON lands at the correct height. GeoJSON features with coordinates outside the valid WGS 84 range are rejected, and a message says how many were skipped. Measurements you took before this fix keep their old height. Contact <help@lumidb.com> if you need them moved.
* Imports no longer fail when only some of the files have RGB or extra attributes.

### 3 June 2026

#### Viewer

* Raster basemaps load at their full native resolution when you zoom in, instead of stopping at zoom level 16.

#### Data

* You can import from S3 buckets in any AWS region. Before, only `eu-west-1` worked.
* Fixed coordinate transformations. Reprojecting data from non-WGS 84 systems to Web Mercator no longer skips datum shifts, which had put data such as EPSG:28992 (RD New) 100 to 200 m off. Several EPSG definitions are corrected, and many coordinate systems have the right axis order again in WMTS output.

### 2 June 2026

#### Viewer

* You can hide a single annotation from the measurements list with the eye button.

#### Data

* You import GeoJSON from an **Import** item in the measurement tools menu, where you can choose several `.geojson` files. Dragging files onto the viewer no longer works for this.

#### Integrations

* Fixed: ArcGIS Pro refused to stream any points when a crop polygon was far from the boundary of a dataset, and grayed out the layer.

## May 2026

**Highlights:** GeoTIFF rasters, importing rasters in the app, GeoJSON annotations, custom basemaps, and the Object ID and Height Above Ground attributes.

### 28 May 2026

#### Data

* LAS imports accept the `Change`, `ClusterID`, `TreeID` and `HeightAboveGround` extra bytes in any numeric data type, and with names in any letter case. LumiDB converts the values to the type of the column.

#### API

* The OpenAPI specification declares the five ways to authenticate, describes the `p` parameter of the 3D tile endpoints, and has typed responses for `/tables`, `query_spatial` and the tileset routes. It also warns that the bounding volumes of the 3D Tiles output are not compliant, which breaks renderers other than Cesium. `/export/status/{export_id}` is a `GET` request, as the specification now says.

### 24 May 2026

#### Data

* You can import rasters in the web app. A new import dialog for orthophotos replaces the command-line tool for this. It shows a note that the importer is tuned for orthophotos so far. Raster imports use 10 to 20 times less memory, so much larger rasters import on the same hardware. See [Importing data](https://docs.lumidb.com/import-data/importing-data.md).
* LumiDB writes cloud-optimized GeoTIFFs with the metadata at the front of the file. They stream efficiently from S3 and work in QGIS, ArcGIS and GDAL `/vsicurl/`.
* Point cloud and raster dialogs have a key and value editor for custom metadata. You can add, rename, edit and delete entries.

#### Integrations

* Fixed: the WMTS capabilities write coordinates in the axis order of each coordinate system. Projections that put easting first, such as EPSG:3159, now appear in the right place in ArcGIS Pro and QGIS.

#### API

* `PATCH /rasters/{name}/metadata` and the SDK method `patchRasterUserMetadata` change the custom metadata of a raster.
* The C SDK uploads files larger than the single-request limit with multipart uploads, and handles files over 2 GB on Windows. The server rejects a single upload over 5 GiB and tells you to use multipart.

### 19 May 2026

#### Viewer

* Annotations are stored as GeoJSON features and you can export and import them as GeoJSON files. Several crop regions can be active at the same time. GeoJSON features without a geometry import as a **properties** entry in the measurements list, with their key and value pairs shown in the measurement dialog.
* You can set a custom basemap for a project. Choose **Custom...** in the basemap menu and paste a slippy-map URL template, for example OpenStreetMap tiles. LumiDB saves it on the project and uses it whenever the project opens.

#### Data

* LumiDB imports GeoTIFF rasters and shows them in the viewer. For now you import them with the command-line tool.
* Two new point attributes, **Object ID** and **Height Above Ground**, come from the extra bytes of LAS files. `ClusterID` and `TreeID` become **Object ID**, and `HeightAboveGround` becomes **Height Above Ground**. Both are color modes in the viewer, and you can filter by **Object ID** next to classification and height range. The filter button shows an indicator when a filter is active. See [Supported formats and file requirements](https://docs.lumidb.com/import-data/supported-formats-and-file-requirements.md).
* Imported data keeps only the attributes that exist in your files. Files without RGB no longer get a gray color stored. LumiDB adds a color only when an older client or a LAZ export needs it. A LAZ export uses point format 0 when the source had no color.

#### Integrations

* Rasters stream as XYZ tiles and as a WMTS capabilities document that you can paste into QGIS. Cloud-optimized GeoTIFF links are available too. See [QGIS](https://docs.lumidb.com/integrations/qgis.md).

#### Platform

* All users can draw annotations and use the measurement tools. Before, only admins could.

### 7 May 2026

#### Data

* E57 imports no longer fail on files with long runs of points that have identical coordinates, such as the no-return points of Faro scanners.
* LumiDB recognizes more coordinate reference systems when you import or reproject point clouds.

### 6 May 2026

#### Viewer

* Imports and exports are kept for each project. The **Tasks** panel shows the 10 most recent, and you can retry a failed import from it.

#### Data

* Imports of point clouds with very heavy or uneven point density no longer fail when an indexing result is larger than 5 GiB.

## April 2026

**Highlights:** a map view for datasets, project thumbnails, and I3S links that work in ArcGIS Online.

### 28 April 2026

#### Viewer

* The **Project data** page has a map view. It plots each dataset at its location with markers colored by project and groups close markers. Switch between the list and the map, and click a marker to open its menu: open the project, move the dataset or delete it.
* Projects on the home page are cards with thumbnails. LumiDB captures a thumbnail the first time you open a project. Cards show the project members as colored avatars, and the creation date.
* You can rename projects and datasets on the home page and in the dataset list. A new **Copy link** action on the home page copies the address of a project.
* Fixed: the basemap loads tiles across the whole zoom range (6 to 16), not only up to level 8.

#### Data

* You can enter a proj4 string as the projection in the import dialog, as an alternative to an EPSG code.
* Fixed: volume queries across datasets with different or incompatible projections are more reliable. LumiDB calculates in the projection of the boundary, and skips a dataset it can't reproject, instead of failing the whole request.

#### Integrations

* I3S links work in ArcGIS Online. The scene settings are in the link path, and you can send the API key as an `?apikey=` parameter, because ArcGIS Online removes authorization headers. The I3S share dialog has a **Convert to Web Mercator (EPSG:3857)** option for the output coordinate system that ArcGIS Online expects. See [ArcGIS](https://docs.lumidb.com/integrations/arcgis.md).

#### Platform

* Organization admins can edit and export in every project, whatever their role in the project is.

Earlier changes aren't listed.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.lumidb.com/release-notes.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
