Data and views

The Data surface shows your devices in a grid: one row per device, one column per field (name, type, serial number, last activity, 24-hour consumption...). A view saves one way of looking at that data: the filters, the sort order, the visible columns and their order. Same data, several angles.

This surface is rolling out gradually: if the "Appareils" entry does not appear in your sidebar, it is not enabled for your account yet (or it was hidden via Settings → Navigation).

The grid

• Quick search: the top-left field searches name, slug and serial number.
• Filter: add conditions ("type equals...", "litres 24h greater than...") combined with AND or OR. "Add a condition group" nests a sub-group with its own AND/OR join (for example: type equals gateway OR (active AND litres 24h > 100)); groups nest over several levels and are removed in one click. The Filter button badge counts every condition, groups included. "Copy from another view" takes the filter of another view of the same source, after confirmation (the current filter is replaced).
• Sort: click a column header, or stack several sort levels via the "Trier" button.
• Column menu: every header shows its field-type icon, and a chevron (visible on hover) opens the column menu: sort both ways, filter by this field (the Filter panel opens with a ready-made condition), group, hide. On your own databases, owners also manage the structure from there: edit the field, duplicate it (definition and values), insert a field to the left or right, delete the field. Clicking the title still sorts.
• Field editor: "Edit field" opens a single panel that gathers everything: the name, the type, the type configuration, the description (collapsed behind "+ Add a description", shown as the header tooltip and as the default form help) and the permissions. The Description and Permissions menu entries open this same panel, focused on the right section. The type is changed in the same searchable list as at creation, restricted to convertible types (conversion never blocks: whatever cannot be converted is cleared); a field of a non-convertible type keeps its type, locked with a tooltip. The configuration of computed fields (formula, button, lookup, count, rollup) can be edited after creation from this panel. When creating a field, or configuring a rich type (Select, Formula, Button, Rollup, Lookup, GPS position), the panel opens as a wide centered dialog for room to fill it in; quick edits of a simple field (rename, description, permissions) stay anchored to the column.
• Add a field: on your own databases, a "+" column at the far right of the grid creates a field, in the same panel as editing. The type is chosen from a searchable, grouped list (standard fields, links and lookup, computed fields, system fields, action), each type with its icon, with the same per-type configuration area.
• Group: the "Grouper" button groups rows by a field (text, select, boolean or number: no dates in v1). Each group gets a collapsible header whose count is the full filtered total for that value, aggregated server-side: not just the rows on the current page. The collapsed state is remembered per view. Single grouping level in v1.
• Fields: show, hide, reorder and resize columns. The panel is searchable (search box at the top), "Show all / Hide all" toggles the listed rows at once (the search subset, if any), and each row carries a visibility checkbox, a "⋯" menu (the same options as the column header menu: sort, filter, pin, footer aggregate, edit, delete…) and a drag handle to reorder. A column can also be reordered by dragging its header directly in the grid (a colored edge marks the insertion point). Each view keeps ITS configuration: changing it in one view does not affect the others. On your own databases, "Edit field" in the ⋯ menu opens the field editor: a field hidden in the view stays editable without showing its column again.
• Column footer: pick an aggregate (sum, average, min, max, count) in the Fields panel. It covers the whole filtered set, not just the visible page.
• Pinned filters: pin up to 3 fields (📌 in the Fields panel) to change their value in one click from the filter bar, without opening the editor.

Telemetry fields

Three fields computed from telemetry can be filtered and sorted like any other column:

• State: alarme (an alarm is active on the device), silencieux (no data received for more than 48 hours, or ever: the meter went quiet, which is neither OK nor an alarm), or ok. An active alarm always wins over silence. The 48-hour threshold is fixed. The state renders as a colored chip (red for alarm, slate-violet for quiet) and can drive a "Colorer" rule on the map: a map of alarmed devices is just another saved view.
• Last index (L): the meter's last known volume index, in litres.
• Vu le and Litres (24 h): last activity and rolling consumption, as before.

A "quiet devices" view is therefore one filter ("State equals silencieux"), sorted by "Vu le" to surface the oldest first; the "count" footer aggregate on the State column counts alarmed devices over the whole filtered set.

Editing in the grid

On the devices source, what belongs to you can be edited straight in the grid (organisation owners only):

• Name: double-click the cell to set the nickname visible to your organisation (the same one as "Rename for my organisation" on the device page). Clear the cell to remove the nickname.
• Tags: the + button in the cell assigns or removes the organisation's static tags. Smart tags compute their own membership and cannot be edited here.
• Placement: place the device in a zone or remove it; the primary placement is highlighted.
• Dashboards: pin the device to your personal dashboards.

The registry (serial number, type, source...) and telemetry stay read-only: a lock in the column header says so, with the reason in a tooltip. Read-only members (viewer role) see the whole grid with no edit affordance at all.

The grid and the device page write to the same place: a tag assigned here shows up on the device page, and vice versa.

Multi-select and bulk actions

Tick the row checkbox to select several records; Shift + click selects a range, and the header checkbox (de)selects the whole page. An action panel appears on the right with "N selected" and, when the view is filtered to more than the current page, "Select all N matching".

What the panel offers depends on the source:

• Your databases: set a field across every selected row at once, duplicate rows, or send them to the trash. Each gesture is a single operation (not a row-by-row loop), and an undo banner lets you reverse it (see below).
• Devices: tag or place the selection in a zone, just like on the map. Registry fields stay read-only.

Set a field across rows

In the panel's "Edit a field", pick the field, type the value (or tick "Clear the field"), then apply: the value lands on every selected row in a single write. A row you are not allowed to edit is simply skipped, never the whole batch.

Copy, paste, keyboard

The grid is keyboard-driven, like a spreadsheet:

• Arrows / Tab move the active cell, Enter edits it, Esc cancels, Space opens the record, Delete clears the cell.
• Cmd/Ctrl + C copies the selection (or active cell) as spreadsheet text; Cmd/Ctrl + V pastes a range from Excel or Google Sheets: cells fill across rows and columns from the active cell, and rows that overflow past the last one are created. A value that does not match the column type leaves the cell empty instead of failing; a locked column is skipped without blocking the rest of the paste.
• Cmd/Ctrl + Z undoes the last bulk action.
• ? opens the shortcut list.

Undo (banner)

After a bulk edit, a duplication, a deletion or an import, an "Undo" banner shows for a few seconds at the bottom of the screen. It replays the inverse action through the normal paths (same permissions): a deletion restores from the trash, an edit puts the previous values back. Single-level undo: a new action replaces the previous one.

CSV import and export

Export to CSV

Export lives in the selection actions panel: check some rows (or "Tout sélectionner les N correspondants" for the whole view), then Exporter CSV. The file contains the selected rows, in the view's sort order, with only the visible columns, in the French convention (";" separator, comma decimals, ISO dates): it opens directly in Excel.

On a public share link, export only exists when the link's creator ticked "Autoriser le téléchargement CSV": an Exporter CSV button then appears on each device list of the shared view and downloads the whole view (same bounds as the display, up to 10,000 rows), with no selection panel.

Import a CSV (your databases)

"Import CSV" (the "⋯" menu at the right of the toolbar, or the empty-database card) opens the wizard:

1. Choose the file. The separator (";", "," or tab) is auto-detected; say whether the first row is a header.
2. Map the file's columns to your fields. Columns whose name matches a field are pre-mapped; the rest are yours to map or ignore.
3. Optionally choose an idempotency key: a field whose value identifies the row. A re-import then updates existing rows instead of creating duplicates.

The import runs in batches. Each value is converted without failing (an unreadable value leaves the cell empty); a fully invalid row is skipped and listed in the final report with its line number. Accepted formats: CSV and TSV, up to 5,000 rows per file.

The map view

A Map view shows the devices on a map, with the view's filters and search applied. Create one from the "+ Créer" button on a zone of the views rail. Every geolocated device shows as a point; a device without a known position does not appear. The displayed position is the registry one, or the latest GPS position reported by the device when it is fresher.

Your own databases too. One of your databases becomes mappable as soon as it carries a GPS position field (see below). The Map view then appears in the database's "+ Créer" menu; every row with a position shows as a point. With no default GPS position field, the Map view shows a message inviting you to add one.

• Color: the button opens a choice between two modes.

  • Select a field: pick a single-select field (status, type...) then a color and an icon per value. They carry across every view: point (and icon) on the map, row stripe in the grid, legend.
  • Conditions: define ordered rules ("type equals water meter", "24-hour litres above 500"...), each with its color and optional icon: the first matching rule wins. Drag a rule's handle to change its priority.

  Colors come from a curated palette: alarm red and the "quiet" violet stay reserved for semantic states. The legend floats bottom-right on the map, with each value's or rule's icon.

• Map fields: show up to 3 stacked fields next to each point (name, serial number...).

• Selection: the pointer tool selects point by point; the rectangle and the lasso select in bulk (lassos add up). The right panel shows the exact count and the bulk actions: tag, place in a zone.

• Beyond ~500 visible points, the map automatically groups points into numbered clusters: zoom in or click a cluster to expand it.

Layers

The Calques button on the map overlays your own geographic data under the devices:

• Charger un calque… accepts a GeoJSON, a KML or a zipped Shapefile. The file is converted in your browser then saved (up to 5,000 features); water network, parcels, districts...
• WMS/WMTS layer: paste the https URL of a tile service (orthophoto, cadastre...). Tiles are loaded by your browser straight from the service, never by our servers.
• Per layer: visibility (specific to each view), color, opacity slider; "activate / deactivate all".

Layers belong to the organization: every member sees them, only administrators manage them.

The "Geo zone" filter

The Zone géo toolbar button (available on the grid as well as the map) filters devices by drawn geographic areas: click to drop the vertices of a polygon, "Terminer ce polygone" to validate it, "+ Nouveau polygone" to draw more areas. A device located in any of the areas passes the filter. The count updates while you draw; the filter then applies like any other condition (exact pagination and totals, saved views included). Only the registry position counts for this filter.

The calendar view

The Calendar type lays rows out over a month or a week, by a date field of your choice: the interventions planning is the typical case. The view's filters and search stay applied: an "urgent interventions" calendar is a saved view like any other.

• Date field: pick the field that places the event (and, optionally, an end field for multi-day ranges).
• Reschedule: drag an event to another day: the row's date is updated (the time of day is kept). It is the same write path as editing in the grid: same rights, same rules. In read-only contexts (viewer role, shared view), dragging is disabled.
• Record: click an event to open its record in the right panel.

The kanban view

The Kanban type stacks rows into columns by a select field: the interventions status board ("To plan", "In progress", "Done"...) is the typical case.

• Columns: the field's declared choices, plus an "(empty)" column for rows without a value. Each column shows its count and collapses in one click (the collapsed state is remembered in the view).
• Move a card: drag it to another column: the field value is written, like a double-click in the grid. On the devices source, system fields (type, source, state...) are read-only: cards do not move and the affordance says why.
• Record: click a card to open its record.

In both views the surface footer stays honest: beyond 200 loaded rows, the number of rows past the cap is shown, never silently truncated. Both types also exist on your own databases (minus the map: a database has no position).

Views

• Edit a view: every setting (filters, sort, columns, colour, grouping, map fields) stays in session memory until you save: nothing is written on its own. As soon as a view is modified, a line appears under it in the views rail: Enregistrer (save) applies the current state to the view, Dupliquer (duplicate) saves it as a new personal view and switches to it, Annuler (cancel) reverts to the saved configuration. Navigating to another view does not lose your changes: they wait for you until the end of the session. A rejected save (someone else's view, view locked in the meantime) shows up as an error in the rail, never silently.
• Locked view: in the rail line, the save link gives way to a padlock whose tooltip carries the lock reason; duplicate and cancel remain available (duplicate to keep your settings in a personal copy).
• Opening a database: you always land on a view: the last one you opened on that database (remembered by your browser), otherwise the first view in the rail. There is no implicit "All devices" / "All rows" view any more: if the database has no view yet, the surface invites you to create the first one via the rail's "+ Créer" button.
• Personal view: visible to you alone, even within your organisation.
• Team view: visible to every member of the organisation. A locked view can only be edited by its creator; a collaborative view is editable by the organisation's administrators.
• Moving a view between zones: whether through the ⋯ menu ("Partager avec l'équipe", "Rendre personnelle") or by dragging the view from one zone to the other, nothing switches silently: a dialog explains the consequence (the view becomes visible to everyone, or stops being so) and offers Move, Duplicate (the copy goes to the target zone, the original stays put) or Cancel.
• Duplicating a view: on a personal view, the copy is created right below the original, no questions asked. On a team view, a dialog asks where to create the copy: in your personal views or in the team views.
• Deep link: the URL carries the view, the search and the page; share it as-is internally.
• Create from the navigation: the "+ Créer" button on a zone ("Mes Vues", "Vues de l'équipe") or a section opens the menu to create a typed view (grid, map, kanban): pick the type, name the view, it opens right away. On a zone, the same menu creates a section. A view's type is fixed at creation: for another display of the same data, create another view.
• Favourites: a view's ⋯ menu adds it to the rail's Favourites mode (the star in the rail's top bar), so you can reach it in one click whatever database is open. Personal: your favourites are yours alone.
• Open elsewhere: a view's ⋯ menu can also open it in a new tab (a Drexo tab, created in the background without leaving the current view) or in a new browser window, just like navigation rows.
• Sections: group your views under named, collapsible headers, beyond the fixed Personal views / Team views blocks. Drag a view onto a section's header (or onto a view it contains) to file it there; drag it onto a fixed block to pull it out, or drop it on the band that appears under each section while dragging (between two sections, or at the very bottom below the last one) to pull it back out at the end of the zone. To reorder sections, hover a header and drag its handle onto another section: the handle only shows on sections you can manage. A team section's order is shared across the whole tenant; your personal sections' order is yours alone. The collapsed state is remembered per user. A section can be personal (visible to you alone) or team-wide. From the section's ⋯ menu, an administrator can publish a personal section to the team; its creator can make it personal again (it returns to their own views). Deleting a section never deletes its views: they fall back to the fixed blocks.

The record panel

Click a row: the record opens in a right-side panel, with the device summary (recent consumption, alarms), a link to its full page and all its fields. The ↑ / ↓ arrows step through records in the view's order, across pages. The URL follows: share it to reopen the same record in the same view.

The devices widget on dashboards

The devices-list widget on zone dashboards uses the same grid, scoped to the devices of the zone and its sub-locations.

From a dashboard's edit mode, the ⊞ "Link to a view" button makes the widget target the devices of a saved view: the widget then follows the view (its filters, its sort) over time. "Detach" reverts to the previous behaviour.

Your own databases

Create business databases (work orders, contracts, readings...) from Settings → Database management (owners of the organisation only): the "+ Nouvelle base de données…" button. That page gathers all database administration: create, rename, change the icon, move to the trash, restore or purge. Each database has typed fields (text, number, date, checkbox, select, currency, percent, duration, star rating, barcode...), rows edited directly in the grid (double-click), and saved views, just like devices.

Five system fields are also available: Created time, Last modified time, Created by, Last modified by, and Autonumber (a per-database counter, never reused). They fill themselves, filter, sort and group like any other field, but are never editable. The "by" fields show the member's name; "Last modified" also moves when a row is reordered or goes through the trash.

Each database also appears in the sidebar, like dashboards and tools: in Settings → Navigation, move it, nest it under a section, or hide it. Hiding removes the menu entry for every role and every user (including you), without deleting the database or restricting access: the page stays reachable through its address. Per-role visibility is configured in Settings → Roles.

Create a database

"+ Nouvelle base de données…" is a single flow: name the database, pick its icon, then its model, and click "Créer". Blank database is pre-selected (you add your fields afterwards); the ready-made templates are Work orders (field work tracking, with status, priority, linked device and zone), Sites (assets: reservoirs, stations, metering points) and Contacts (operations directory). Picking a template suggests its name and icon, both replaceable. A template creates the database, its typed fields, a pre-sorted default view and, if you wish, a few sample rows to picture the result.

Sample rows are flagged as such: the "Remove sample data" banner sends them all to the trash in one click, without touching rows you entered yourself.

Forms

A form is the write path for people who never open a grid: a field technician scans a QR code on a meter and files a work report with the device already filled in.

• Create: the "+ Créer" button on a zone of the rail, "Formulaire" option (or the "Entry form" card of an empty database). A form is a view like any other: it shows up in the navigation.
• Compose: pick the fields, their order, a label and help text per field, whether it is required, and a display condition ("show this field when the title contains leak").
• Prefill through the URL: every field can be prefilled through the form's address, in three modes: editable (suggested value), locked (displayed, not editable: the QR-on-a-meter case) or hidden (never displayed, submitted as-is).
• Share: create a public link from the form's panel. The generator produces a per-device prefilled URL and its QR code, ready to print and stick on the equipment. A link can be suspended (reversible) or revoked (final) at any time; submissions are rate-limited to absorb abuse.
• Receive: every submission becomes a row of the database, visible in the grid with the linked device resolved. Anonymous submissions are flagged as such (no author, originating link kept).

A still-empty database

A database with no rows offers the next steps directly: add a first row, define the fields, import a CSV or create an entry form. Once the database has rows, two gestures add one: the floating "+ Ajouter" button at the bottom right opens a form to fill the row field by field before creating it (default values are prefilled there); the "+ Ajouter une ligne" line at the bottom of the grid inserts a blank row to fill in directly in the grid.

The gear in that form's header configures it per view: which fields to offer, in what order, with which label, and which are required. Each view gets its own entry form (a short "field" view, a full "admin" view). Configuring the form follows the view's edit rights, like its filters or sorts (a locked view is only reconfigurable by its author); the configuration applies when you save the view.

• Changing a field's type never fails: convertible values are converted ("42" becomes the number 42), the rest are simply cleared.
• Default value: every fillable field (text, number, select, checkbox, date...) can carry a default value, set in the field editor ("Default value" section). Every new row is born with it: rows added in the grid, at the bottom of the grid, or submitted through a form. In a public form the field shows up prefilled with its default, editable before sending (a URL prefill always wins). Setting or changing a default never touches existing rows.
• Display format: the "Format" section of the field editor tunes the display without changing the stored value: decimal places on a Number, the currency of a Currency field (symbol rendered automatically: €, $, £...), how a Percent shows (plain "50 %" or a progress bar, with decimals), the Duration format (h:mm:ss or h:mm), Date with or without time, and the number of stars on a Rating. Filters, sorts and exports keep seeing the raw value.
• Select fields (single or multiple): options are managed in the field editor: add, rename, delete, reorder by drag and drop (the order you set is the display order everywhere), pick a colour per option from the palette, or "Sort alphabetically". Each value renders as a coloured chip in the grid, the record panel, the kanban board, group headers, public forms and the filter value picker; the "Colour code" toggle switches every chip back to neutral grey. Renaming an option updates every cell that carried it, in one pass; renaming to a label a multi-select cell already carries merges (no duplicate chip). Deleting an option never touches the rows: existing values stay, shown as a neutral chip. In the grid, editing a cell opens an option menu (chips); an administrator can add an option on the fly from there.
• Formula field: a computed value per row, written like in a spreadsheet: {Field name} references a column, and the formula accepts arithmetic, concatenation (&), comparisons, AND/OR/ NOT, IF, ROUND, ABS, UPPER, LOWER, TRIM, LEN, COALESCE and BLANK(). The result filters, sorts, groups, aggregates in the column footer and lands in CSV exports like any other column. Renaming a field never breaks the formulas that use it; deleting a referenced field blanks the formula (the grid never breaks); dividing by zero yields an empty cell.
• Rollup field: aggregates the rows linked through a "Link record" field (sum, average, min, max or count of a numeric field on the linked database, e.g. the total litres logged on a worksite). Linked rows in the trash are excluded.
• Count field: counts the links a row carries through a given link field (attached devices, linked readings…).
• Button field: a per-row button that opens a URL built from the row itself (an https template where {key} inserts the column's value, encoded: for instance an external tool's page prefilled with the reference). An "Insert field" menu drops a column's {key} (or the row ID) at the cursor, so you never have to know or type the keys by hand. An optional icon shows to the left of the label. If a template field is empty or was deleted, the button disables itself instead of opening a broken address. Nothing is stored: the button never shows up in filters, forms or CSV exports.
• Link fields: link a row to a device, a zone or an organisation member. The picker only offers what you can access. If access is later revoked, the cell shows an "unresolved" link: nothing leaks, nothing breaks.
• Lookup: display, next to a device link, its name or zone, computed live (never copied, so never stale). It is now a first-class column: it filters, sorts and groups like any other. Rollup, Count, Lookup and Formula share a cap of ten computed fields per database.
• Appears in: a device's record panel lists the databases referencing it, with direct navigation to the row.

Attachments

The Attachments field attaches photos or PDFs to a row: a meter installation photo taken with a phone, a scanned work order... The cell shows thumbnails; the row's record panel offers the full gallery (preview, download, delete, add).

Good to know:

• Accepted formats: images (JPEG, PNG, WebP, GIF) and PDF, 10 MB max per file, 20 files per cell.
• Storage: 1 GB per organisation. Beyond that, uploads are refused with an explicit message; delete attachments to free up space.
• Privacy: files are private to your organisation (same access rules as the database) and served through signed links that expire within minutes.
• No antivirus scanning: files are stored as-is. Only open attachments uploaded by people you trust.
• Deletion: deleting an attachment (or its row) deletes the file, permanently.

GPS position field

The GPS position field gives each row a position (latitude, longitude) and unlocks the database's Map view. Add it like any field (field menu, "GPS position" type). The cell shows the coordinates and is edited by entering latitude and longitude; a row with no coordinates does not appear on the map.

Two settings when you create the field:

• Default position for the database: the Map view and the "Geo zone" filter use this field. A database may hold several GPS position fields (billing address, work site...) but only one default: ticking another one moves the setting.
• Projection (EPSG): the projection code you enter and read coordinates in. Defaults to 4326 (WGS84, the usual latitude and longitude). Set another code (for example 2154 for Lambert 93) to work in a local projection; the conversion is done for you and storage stays in WGS84.

The Geo zone filter (drawing polygons) works on a tenant database just like on devices, against the view's GPS position field.

Commenting a row

A row's record panel (click the row) carries its comments: the team discusses a work order, a contract or a reading right next to the data. Every member can read; whoever can enter rows can comment. Everyone can edit or delete their own comments; organisation administrators can delete any of them.

• Notify a member: mention them with @ followed by their email address (or the part before the @), e.g. @camille@utility.com or @camille. The mentioned member, and only them, gets a bell notification with a direct link to the row.
• In the grid, a discreet column shows each row's comment count.

Locking a field

Reserve editing a column to administrators: an intervention's status no longer gets changed by mistake during field entry. The lock is set in the Permissions section of the field editor (administrators only): open it from the column header menu ("Edit field permissions") or the pencil in the Fields panel, then tick "Seuls les owners modifient ce champ"; untick to reopen it.

Good to know:

• It's a write lock, not a secret. A locked field stays visible and readable to every member; only its editing is reserved. Do not use it to hide a sensitive value.
• The lock holds everywhere, not just on screen. The rule is enforced at the heart of the database: neither the grid, nor a bulk paste, nor an external tool can bypass a locked field. This is stronger than competing collaborative spreadsheets, where the lock is only an interface courtesy.
• In the grid, a locked field shows a padlock in its header and a greyed-out cell for members who cannot edit it, with the reason in a tooltip.
• Fillable via a form: tick this option on a locked field so that an entry form can still fill it (the field-survey case), while keeping grid editing reserved to administrators.
• Each member only edits the rows they created: the option of the same name lives in the toolbar's "⋯" menu, under the "Database" section (administrators only). When ticked, it restricts each member to the rows they created; administrators edit everything.

The activity log

A row's record panel carries its activity log: who changed which field, from which value to which value, and when. Every cell change leaves a trace, form entry included (credited "Formulaire" when the submission is anonymous).

The log is tamper-proof: a written entry cannot be edited or deleted by anyone. It gives the reliable history of a sensitive row (an intervention, a contract), right next to the data.

The trash

Deleting is never instantly irreversible: deleted rows and databases go through a trash, purged automatically after 30 days.

• Deleting a row ("Move to trash" in the record panel) removes it at once from the grid, views, aggregates and widgets. Its links and comments stay attached, dormant: a restored row comes back exactly as it was. While it sits in the trash, links pointing at it show as "unresolved".
• A database's trash ("Corbeille" in the toolbar's "⋯" menu, administrators only) lists deleted rows: restore them one by one, purge them for good, or empty everything.
• Deleting a database (in the same panel, or from Settings → Database management) sends it to the trash: it disappears from navigation and all views. Find it under Settings → Database management, Trash section, to restore it losslessly (fields, rows, links, comments, views) or purge it for good.

Automations

A database can react on its own: "when a row is created", "when a field changes" or "when a row enters a view", Drexo notifies members in the bell or calls a webhook you wire to any tool. The Automations entry in the toolbar's "⋯" menu (organization administrators only) opens the manager, next to fields and trash.

• The trigger says when: Row created, Field changed (you pick the watched field) or Enters a view (you pick a team view; its filter defines membership). "Enters a view" only fires on the transition out-of-view → in-view: a row that stays in does not re-fire. "Field changed" only fires when the watched value actually changes; editing another field does not count.
• The condition (optional, on the first two triggers) narrows with the same filter as views: e.g. "only when Status = urgent".
• Notify sends a bell notification to the chosen members. Editing many rows at once rings only once.
• Webhook calls a third-party https URL. Each call carries the old and new row values, and a signature (X-Drexo-Signature header) computed with a secret you set. The secret lives server-side: it is never shown again. A trashed row triggers nothing.
• Reliability: webhooks are delivered in the background (writing a row never waits on a network call), retried on failure, and the automation disables itself after repeated failures. The delivery log keeps a copy of the latest requests and responses to diagnose an endpoint that stopped answering.
• Test sends a test event to the webhook (payload flagged "test": true, with a sample row from the database), through the same circuit as real events: signature, retries and delivery log included. The result shows up in the delivery log.
• Security: the URL must be public over https; internal addresses (private IPs, localhost, internal domains) are refused.

Build a view with AI (experimental)

Describe the view you want, in English or French ("meters silent for more than 48h, sorted by litres descending"), and the AI proposes filters, sorts and columns. You review the proposal before saving it: nothing is created automatically. If the AI proposes a field or operator that does not exist, the proposal is rejected and corrected, never saved as-is.

The AI lives inside view creation: open the "+ Créer" menu on a zone of the rail, pick a type, then describe the view in the "Configurer avec l'IA" block: it is optional, without a description the view is created empty. The block only appears when the feature is enabled for your account (progressive rollout) and when AI is allowed for your organization. The AI only receives the field names, never your row data.

Classify a field with AI (experimental)

On a database, "Classify with AI" (toolbar "⋯" menu) fills a "Select" field (for example intervention type) from the text of another field (for example notes). You set the list of categories; the AI files each row into one of them. A row the AI cannot confidently classify stays unlabeled (counted separately) rather than filed at random. You review the result, then apply: labels are written through the same paths as manual editing.

Like the view-builder, classification is governed by the organization's AI setting; only the text of the column being classified is sent to the model.