## Summary
- Add `@mention` support to the AI Chat text input by replacing the
plain textarea with a minimal Tiptap editor and building a shared
`mention` module with reusable Tiptap extensions (`MentionTag`,
`MentionSuggestion`), search hook (`useMentionSearch`), and suggestion
menu — all shared with the existing BlockNote-based Notes mentions to
avoid code duplication
- Mentions are serialized as
`[[record:objectName:recordId:displayName]]` markdown (the format
already understood by the backend and rendered in chat messages), and
displayed using the existing `RecordLink` chip component for visual
consistency
- Fix images in chat messages overflowing their container by
constraining to `max-width: 100%`
- Fix web_search tool display showing literal `{query}` instead of the
actual query (ICU single-quote escaping issue in Lingui `t` tagged
templates)
## Test plan
- [ ] Open AI Chat, type `@` and verify the suggestion menu appears with
searchable records
- [ ] Select a mention from the dropdown (via click or keyboard
Enter/ArrowUp/Down) and verify the record chip renders inline
- [ ] Send a message containing a mention and verify it appears
correctly in the conversation as a clickable `RecordLink`
- [ ] Verify Enter sends the message when the suggestion menu is closed,
and selects a mention when the menu is open
- [ ] Verify images in AI chat responses are constrained to the
container width
- [ ] Verify the web_search tool step shows the actual search query
(e.g. "Searched the web for Salesforce") instead of `{query}`
- [ ] Verify Notes @mentions still work as before
Made with [Cursor](https://cursor.com)
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
As attachment files have migrated from fullPath to file files field,
need to migrate richText logic to fit to new attachment file handling +
data migration
## Reduce type leakage between GraphQL schemas
### Why
Twenty runs two separate GraphQL schemas: **core** and **metadata**.
NestJS's `@nestjs/graphql` uses a global `TypeMetadataStorage` that
accumulates all decorated types across all modules. When each schema is
built, every registered type leaks into both schemas regardless of which
module it belongs to.
This means the core schema's generated TypeScript
(`generated/graphql.ts`) contained ~2,700 lines of types that only
belong to the metadata schema (and vice versa). This creates confusion
about type ownership, inflates generated code, and makes it harder to
reason about which API surface each schema actually exposes.
### How
**1. Patch `@nestjs/graphql` to support schema-scoped type resolution**
- **(Already done)** Added a `resolverSchemaScope` option to
`GqlModuleOptions`, allowing each schema to declare a scope (e.g.
`'metadata'`)
- `ResolversExplorerService` now filters resolvers by a
`RESOLVER_SCHEMA_SCOPE` metadata key, so each schema only sees its own
resolvers
- `GraphQLSchemaFactory` now performs a **reachability walk**
(`computeReachableTypes`) starting from scoped resolver return types and
arguments, only including types that are transitively referenced —
handling unions, interfaces, and prototype chains
- Type definition storage and orphaned reference registry are cleared
between schema builds to prevent cross-contamination
**2. Register `ClientConfig` as orphaned type in metadata schema**
Since `ClientConfig` is needed in the metadata schema but not directly
returned by a resolver, it's explicitly declared via
`buildSchemaOptions.orphanedTypes`.
**3. Regenerate frontend types and fix imports**
- `generated/graphql.ts` shrank by ~2,700 lines (types moved to where
they belong)
- `generated-metadata/graphql.ts` gained types like `ClientConfig` that
were previously missing
- ~500 frontend files updated to import from the correct generated file
Closes [1744](https://github.com/twentyhq/core-team-issues/issues/1744).
This PR migrates attachments to morph relations behind a feature flag,
following the TimelineActivity pattern. it introduces the
`IS_ATTACHMENT_MIGRATED` flag, updates standard field metadata and
indexes to use morph relations, adds a workspace migration that renames
`attachment.*Id` columns to `target*Id` and converts the corresponding
field metadata to `MORPH_RELATION` with a shared `morphId`. On the
frontend, attachment read/write paths now switch to `target*Id` when the
flag is enabled.
It also fixes optimistic filtering for morph join columns. The metadata
API deduplicates morph fields, so attachments now expose a single target
field of type `MORPH_RELATION` plus a `morphRelations` array listing
each target object. Because only one `settings.joinColumnName` is
returned (e.g. `targetRocketId`), filters like `targetCompanyId` don’t
map to any field and the optimistic cache code throws.
`doesMorphRelationJoinColumnMatch` resolves this by computing all valid
join column names from `morphRelations` using
`computeMorphRelationFieldName` and comparing them to the filter key.
That makes filters like `targetCompanyId` resolvable even with a single
target field, so attachment uploads and list matching no longer crash.
<img width="477" height="474" alt="image"
src="https://github.com/user-attachments/assets/50e19418-3438-4d1e-9f1f-1bc1a03174a9"
/>
<br />
<br />
Today the metadata API returns one morph field called `target` and a
list of possible targets (`morphRelations`), but it does not tell us the
join column for each target. That’s why the Frontend had to compute join
column names.
If we want to fix this at the API level, there are two options:
- Add join column names to each target in `morphRelations` (e.g. company
→ `targetCompanyId`). This is additive and low‑risk.
- Return each target as its own field (`targetCompany`, `targetPerson`,
etc.) instead of a single target. This is a larger change because it
changes the shape of metadata and would require more UI updates.
<!-- CURSOR_SUMMARY -->
---
> [!NOTE]
> Introduces morph relations for attachments behind
`IS_ATTACHMENT_MIGRATED`, aligning server schema/metadata and frontend
behavior.
>
> - Adds `IS_ATTACHMENT_MIGRATED` flag (frontend/server) and
seeds/defaults; updates generated GraphQL enums
> - New workspace upgrade `1.17` command migrates data: renames
`attachment.*Id` → `target*Id` and converts related fields to
`MORPH_RELATION` with shared `morphId`
> - Updates standard field metadata and indexes to `target*` (attachment
+ related objects), dev seeds, snapshots, and workspace entity types
> - Frontend: switches attachment read/write filters via
`getActivityTargetObjectFieldIdName` using the feature flag; updates
hooks/components (`useAttachments`, `useUploadAttachmentFile`, editors);
expands `Attachment` type
> - Fixes optimistic cache filtering to recognize morph join columns in
`isRecordMatchingFilter` by computing valid join-column keys from
`morphRelations`
>
> <sup>Written by [Cursor
Bugbot](https://cursor.com/dashboard?tab=bugbot) for commit
f208fa23b1135af114f484f6d65041a816d6e718. This will update automatically
on new commits. Configure
[here](https://cursor.com/dashboard?tab=bugbot).</sup>
<!-- /CURSOR_SUMMARY -->
---------
Co-authored-by: Félix Malfait <felix@twenty.com>
CommandMenu did not push its own focus item to the focus stack when
focused. (See NavigationDrawerInput)
---------
Co-authored-by: Charles Bochet <charles@twenty.com>
This is first step toward separating the path between display mode and
input mode for inline-cells like we have done for tableCell.
The idea is to have 1 floating or anchored field input shared between
all cells. This should be done for field-list, board and task/note row +
for hover and edit mode.
This PR is only about field-list and hover
This PR prepares the refactor of record field definition and column
definition.
We need to separate the two concepts of FieldInput and FieldContext at
the cell level, and the new RecordField concept that will replace
FieldDefinition at the metadata and state level.
So we create a new ui sub-folder in the already existing record-field
folder, in order to separate those two concepts that are still very
close.
This PR introduces refactors and tradeoffs in the API around the events
of field input.
# Refactored usePersistField
The hook `usePersistField` has been refactored to be used anywhere in
the app, not just inside a FieldContext.
This was meant to solve a bug at the beginning but now it is just used
once in `RecordDetailRelationSection` outside of the context, still this
is better to have this hook like that for future use cases.
We also introduce `usePersistFieldFromFieldInputContext`, for an easier
API inside a FieldContext.
# Introduced a new `FieldInputEventContext`
To remove the drill-down of events, we introduce
`FieldInputEventContext`, this allows to set only once the handlers /
events. In practice it allows to have an easier time maintaining the
events for the many different field inputs, because it matches the
pattern we already use of taking everything from a context
(`FieldContext`).
# Removed drill-down from FieldInput
The heavy drill-down in FieldInput has been completely removed, since
everything can be derived from `FieldContext` and
`FieldInputEventContext`.
Also there was some readonly and other specific props, but they were all
drilled down from FieldContext, so it was easier to just use
FieldContext where needed.
# Refactored events of `MultiItemFieldInputProps`
The component `MultiItemFieldInputProps` has a contrived API, here we
just remove the complex part that was persisting from inside.
We now only give a classic API with `onChange` and `onEscape` the rest
is left to higher levels, where it should be, because this generic
component shouldn't be aware of persisting things.
# Extracted the parsing logic of persisted values
For each input field component, we now have a clear util that was before
bound to the persist call,
# Tradeoff with persist times
The tradeoff before was that persistField was called anywhere, before
exiting the component sometimes, now it is only called by the higher
levels like table or show page, which handles this abstraction.
This could be challenged, however I think that having a lot of different
events, and not just `handleSubmit` and `handleCancel`, convey enough
meaning for the higher levels to decide what to do in each case.
A `skipPersist` argument was added in events in the rare edge cases
where we want to voluntarily skip persisting even with a submit or
escape, but that could be challenged because we could also say that we
should use cancel for that and stick to that convention.
# Handling of the bug in `ActivityRichTextEditor`
Initially this refactor was prioritized for solving this bug, which was
very annoying for the users.
But while fixing it with the new persistField hook I just understood
that the problem is not just for record title cells but for anything
that is open when we click on a rich text editor.
The issue is described here :
https://github.com/twentyhq/core-team-issues/issues/1317
So right now I just let it as is.
# Stories
The stories were checking that a request was sent in some cases where
persist was called before a component exiting, now that persist is only
called by higher-levels I just removed those tests from the stories,
because that should be the responsibility of higher levels.
Also a helper `getFieldInputEventContextProviderWithJestMocks` was
created that exposes a context and jest mock functions for testing this
new API in stories.
# Miscellaneous
Deactivated tui with nx by default, because it can be annoying.
Closes https://github.com/twentyhq/core-team-issues/issues/1276
We now need to take fields permissions into account when determining if
a cell is editable or not.
a record field (= a cell) is read only if one of the condition is met:
- the object of the record is read-only (can be determined at table
level) by permission setting
- the field of the object is read-only (can be determined at column
level) - either by permission settings or by system (some workflows
fields for intance)
- the record is deleted (can be determined at row level)
we reorganized the code to avoid re-computing known information as much
as possible.
---------
Co-authored-by: Charles Bochet <charles@twenty.com>
This PR removes every legacy state management logic left in the code
base, we only keep our last component state management logic.
Removed :
- Scoped states logic
- Dropdown scope logic
- Scope id naming
- Component state v1
- Component state v2_alpha
# QA
| Component | Comments |
| --- | --- |
| AttachementDropdown | Ok |
| ObjectTasks | Ok |
| ActivityRichTextEditor | Ok |
| App | Ok dialogs work |
| AppRouterProviders | Ok dialogs work |
| RecordBoardScrollToFocusedCardEffect | Ok |
| RecordBoard | Ok |
| RecordBoardCard | OK |
| useSelectAllCards | Ok |
| RecordInlineCell | Ok |
| useInlineCell | Ok |
| RecordDetailRelationRecordsListItem | Ok |
| RecordDetailRelationSectionDropdownToMany | Ok |
| RecordDetailRelationSectionDropdownToOne | Ok |
| useHandleContainerMouseEnter | Ok |
| useResetTableRowSelection | Ok |
| useSelectAllRows | Ok |
| useSetRecordTableData | Ok |
| useRecordTable | Ok |
| useRecordTableMoveFocusedCell | Ok |
| useRecordTableMoveFocusedRow | Ok |
| useMoveHoverToCurrentCell | OK |
| useTriggerActionMenuDropdown | Ok |
| RecordTableBodyDragDropContextProvider | Ok |
| RecordTableHeaderCell | Ok |
| useSetCurrentRowSelected | Ok |
| useRecordTitleCell | Ok |
| SettingsObjectFieldActiveActionDropdown | Ok |
| SettingsObjectFieldInactiveActionDropdown | Ok |
| SettingsObjectFieldItemTableRow | Ok |
| SettingsObjectInactiveMenuDropDown | Ok |
| SettingsRoleEditEffect | Ok |
| SettingsRolesQueryEffect | Ok |
| Dropdown | Ok |
| useSelectableList | Ok |
| useSelectableListHotKeys | Ok |
| useSelectableListListenToEnterHotkeyOnItem | Ok |
| useNavigationSection | Ok |
| useCreateViewFromCurrentState | Ok |
| useDeleteViewFromCurrentState | Ok |
| useUpdateViewFromCurrentState | Ok |
| ViewPickerFavoriteFoldersDropdown | Ok |
| useCreateViewFromCurrentView | Ok |
| ViewBarDetails | Ok |
| WorkflowRunVisualizer | Ok |
| useHandleWorkflowRunDiagramCanvasInit | Ok |
| WorkflowDiagramCanvasBase | Ok |
| WorkflowRunVisualizerEffect | Ok |
| useRemoveStepFilter | Ok |
| useRemoveStepFilterGroup | Ok |
| useUpsertStepFilterSettings | Ok |
| WorkflowFindRecordsFilters | Ok |
| SettingsObjects | Ok |
resolve#13168 , #8501
This PR fixes an issue where the onChange trigger was deleting all
attachments by removing the JWT token from their paths (if it existed)
and comparing the new body with the old body to identify deleted
attachments. It ensures that only attachments actually removed from the
body get deleted, preventing unintended deletion of attachments not
added directly through the body. It also handles updating attachment
names in the body so changes are reflected in Files, with related tests
updated accordingly.
https://github.com/user-attachments/assets/8d824a24-b257-4794-942e-3b2dceb9907d
---------
Co-authored-by: Lucas Bordeau <bordeau.lucas@gmail.com>
Fixes https://github.com/twentyhq/twenty/issues/12885
This PR fixes a hotkey scope race condition happening on note/task
creation.
The problem is that `ActivityRichTextEditor` catches the click event
before the title cell.
So here we prevent this from happening by checking if the record title
cell is.
This is only temporary and should be improved after the persist logic
refactor : https://github.com/twentyhq/core-team-issues/issues/192
Moves system-level operations (auth, billing, admin) to use the
/metadata endpoint instead of /graphql.
This cleans up the endpoint separation so /graphql is purely for core
objects (Company, People, etc.) and /metadata handles all system
operations.
Part of prep work for webhook/API key core migration.
Creating manual chunk was a bad idea, we should always solve lazy
loading problem at the source instance.
Setting a 4.5MB for the index bundle size, CI will fail if we go above.
There is still a lot of room for optimizations!
- More agressive lazy loading (e.g. xyflow and tiptap are still loaded
in index!)
- Add a prefetch mechanism
- Add stronger CI checks to make sure libraries we've set asides are not
added back
- Fix AllIcons component with does not work as intended (loaded on
initial load)
# Introduce focus stack to handle hotkeys
This PR introduces a focus stack to track the order in which the
elements are focused:
- Each focused element has a unique focus id
- When an element is focused, it is pushed on top of the stack
- When an element loses focus, we remove it from the stack
This focus stack is then used to determine which hotkeys are available.
The previous implementation lead to many regressions because of race
conditions, of wrong order of open and close operations and by
overwriting previous states. This implementation should be way more
robust than the previous one.
The new api can be incrementally implemented since it preserves
backwards compatibility by writing to the old hotkey scopes states.
For now, it has been implemented on the modal components.
To test this PR, verify that the shortcuts still work correctly,
especially for the modal components.
Using useEffect triggered at ActivityRichTextEditor unmount, to delete
attachments only when note is closed (and not when file block is deleted
during note update to keep command + z shortcut)
closes : https://github.com/twentyhq/twenty/issues/11229
## What
- Deprecate overlayscrollbars as we decided to follow the native
behavior
- rework on performances (avoid calling recoil states too much at field
level which is quite expensive)
- Also implements:
https://github.com/twentyhq/core-team-issues/issues/569
---------
Co-authored-by: Lucas Bordeau <bordeau.lucas@gmail.com>
# Introduction
In this PR we've migrated `twenty-shared` from a `vite` app
[libary-mode](https://vite.dev/guide/build#library-mode) to a
[preconstruct](https://preconstruct.tools/) "atomic" application ( in
the future would like to introduce preconstruct to handle of all our
atomic dependencies such as `twenty-emails` `twenty-ui` etc it will be
integrated at the monorepo's root directly, would be to invasive in the
first, starting incremental via `twenty-shared`)
For more information regarding the motivations please refer to nor:
- https://github.com/twentyhq/core-team-issues/issues/587
-
https://github.com/twentyhq/core-team-issues/issues/281#issuecomment-2630949682
close https://github.com/twentyhq/core-team-issues/issues/589
close https://github.com/twentyhq/core-team-issues/issues/590
## How to test
In order to ease the review this PR will ship all the codegen at the
very end, the actual meaning full diff is `+2,411 −114`
In order to migrate existing dependent packages to `twenty-shared` multi
barrel new arch you need to run in local:
```sh
yarn tsx packages/twenty-shared/scripts/migrateFromSingleToMultiBarrelImport.ts && \
npx nx run-many -t lint --fix -p twenty-front twenty-ui twenty-server twenty-emails twenty-shared twenty-zapier
```
Note that `migrateFromSingleToMultiBarrelImport` is idempotent, it's atm
included in the PR but should not be merged. ( such as codegen will be
added before merging this script will be removed )
## Misc
- related opened issue preconstruct
https://github.com/preconstruct/preconstruct/issues/617
## Closed related PR
- https://github.com/twentyhq/twenty/pull/11028
- https://github.com/twentyhq/twenty/pull/10993
- https://github.com/twentyhq/twenty/pull/10960
## Upcoming enhancement: ( in others dedicated PRs )
- 1/ refactor generate barrel to export atomic module instead of `*`
- 2/ generate barrel own package with several files and tests
- 3/ Migration twenty-ui the same way
- 4/ Use `preconstruct` at monorepo global level
## Conclusion
As always any suggestions are welcomed !
Closes https://github.com/twentyhq/core-team-issues/issues/271
This PR
- Removes the feature flag IS_COMMAND_MENU_V2_ENABLED
- Removes all old Right drawer components
- Removes the Action menu bar
- Removes unused Copilot page
# Introduction
Avoid having multiple `isDefined` definition across our pacakges
Also avoid importing `isDefined` from `twenty-ui` which exposes a huge
barrel for a such little util function
## In a nutshell
Removed own `isDefined.ts` definition from `twenty-ui` `twenty-front`
and `twenty-server` to move it to `twenty-shared`.
Updated imports for each packages, and added explicit dependencies to
`twenty-shared` if not already in place
Related PR https://github.com/twentyhq/twenty/pull/9941
Replaced references to '~/generated-metadata/graphql' with
'~/generated/graphql' across multiple files. This change ensures
uniformity in import paths and aligns with the updated directory
structure.
### Solution
fullPath prop on attachement (when returned by backend) is updated to
'domain + path' (formerly 'path').
Consequently, getFileAbsoluteURI util in front is removed.
closes#8763
---------
Co-authored-by: etiennejouan <jouan.etienne@gmail.com>
Charles Bochet