## Summary
- Enable application log console output in the `twenty-app-dev` Docker
image by adding `APPLICATION_LOG_DRIVER=CONSOLE` to the Dockerfile ENV
block
- Rename `APPLICATION_LOG_DRIVER_TYPE` to `APPLICATION_LOG_DRIVER` and
`ApplicationLogDriverType` to `ApplicationLogDriver` for consistency
with all other driver config variables (`EMAIL_DRIVER`, `LOGGER_DRIVER`,
`CAPTCHA_DRIVER`, etc.)
- Add `APPLICATION_LOG_DRIVER` to `.env.example`
## Summary
This PR enhances the handling of sensitive configuration variables by
improving masking logic and user experience when editing them. It adds
metadata-aware masking for dynamically marked sensitive variables and
clears sensitive values when entering edit mode.
## Key Changes
- **Backend masking improvements**: Updated `maskSensitiveValue()` to
accept metadata parameter, enabling masking of variables marked as
sensitive via metadata (not just predefined masking config). Sensitive
non-string values are masked as `********`.
- **Edit mode UX**: When editing a sensitive variable, the value field
is now cleared on entering edit mode to prevent exposing masked values
and ensure users intentionally provide new secret values.
- **Form state tracking**: Enhanced `useConfigVariableForm()` hook to
accept an `isEditing` parameter and properly track value changes for
sensitive variables during edit operations.
- **Input placeholder**: Updated placeholder text for sensitive variable
inputs to show `Enter a new secret value` instead of the generic
database storage message, providing clearer intent to users.
## Implementation Details
- The `maskSensitiveValue()` method now checks both predefined masking
configurations and runtime metadata to determine if a value should be
masked
- Sensitive string values use LAST_N_CHARS strategy (4 characters),
while non-string values are masked uniformly
- The form's `hasValueChanged` flag is set to true when editing
sensitive variables to ensure proper validation and submission handling
https://claude.ai/code/session_01JiTckmuVJMQWpJ7TUGwsqb
---------
Co-authored-by: Claude <noreply@anthropic.com>
## Summary
- Introduces a new `application-logs` core module with a driver pattern
(disabled/console/clickhouse) to capture and persist logic function
execution logs
- Adds a ClickHouse `applicationLog` table with per-line log storage,
30-day TTL, and `ORDER BY (workspaceId, timestamp, applicationId,
logicFunctionId)`
- Surfaces application logs in the existing frontend audit logs table as
a new "Application Logs" source with dedicated columns (Function,
Timestamp, Level, Message, Execution ID)
## Details
**Write path**: `LogicFunctionExecutorService.handleExecutionResult()`
parses the multi-line log string from driver output into individual `{
timestamp, level, message }` entries, generates an execution UUID, and
passes them to `ApplicationLogsService.writeLogs()` which delegates to
the configured driver.
**Driver pattern**: Follows the exception-handler module style (Symbol
injection token + `forRootAsync` dynamic module). Three drivers:
- `DISABLED` (default) — no-op, prevents information leaking
- `CONSOLE` — structured stdout logging with level-based `console.*`
calls
- `CLICKHOUSE` — inserts rows into the `applicationLog` ClickHouse table
**Read path**: Extends the existing event-logs module by adding
`APPLICATION_LOG` to the `EventLogTable` enum, table name mapping, and
normalization logic.
**Config**: New `APPLICATION_LOG_DRIVER_TYPE` environment variable
(default: `DISABLED`).
## Summary
Fixes#19422.
Self-hosted users hitting "No AI models are available" after configuring
API keys via the admin panel were victims of a stale
`AiModelRegistryService` cache. Only the four `addAiProvider` /
`removeAiProvider` / `addModelToProvider` / `removeModelFromProvider`
mutations called `refreshRegistry()` — setting an API key through
`set/update/deleteDatabaseConfigVariable` left the registry pointing at
the pre-mutation provider state.
Rather than patch each mutation site (and re-introduce the same class of
bug on the next one), the registry now invalidates lazily based on the
LLM config-group hash, mirroring the pattern `WebSearchDriverFactory`
already uses via `DriverFactoryBase`. Any mutation to an LLM-tagged
config variable is picked up automatically on the next read — callers
never have to remember to refresh.
- Extracted `getConfigGroupHash` into a shared util reused by both
`DriverFactoryBase` and `AiModelRegistryService` (and switched the hash
to `JSON.stringify` so object-typed config vars like `AI_PROVIDERS`
actually contribute meaningfully).
- `AiModelRegistryService` gates all internal `Map` access behind
private getters that call `ensureFresh()`, so future read paths can't
accidentally observe stale state. Build path uses underscored backing
fields directly to avoid recursing.
- Dropped the now-redundant `refreshRegistry()` public method and its
four call sites in `admin-panel.resolver.ts`.
## Test plan
- [x] `nx typecheck twenty-server` passes
- [x] Existing admin-panel + ai-models specs pass (79 tests)
- [ ] Manual: on a self-hosted instance, set `OPENAI_API_KEY` (or
another provider key) via the admin panel `setDatabaseConfigVariable`
mutation and confirm AI features work without a server restart
- [ ] Manual: existing flows (`addAiProvider`, `addModelToProvider`,
etc.) still pick up new providers on the next read
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
**Optimize workflow cron jobs: partition workspaces and use raw
queries**
- Split all 3 workflow cron jobs (WorkflowRunEnqueueCronJob,
WorkflowHandleStaledRunsCronJob, WorkflowCleanWorkflowRunsCronJob) to
process only 1/10th of workspaces per invocation using minute-based
partitioning, reducing per-run load
- Replace ORM repository + workspace context loading with raw SQL
queries in WorkflowRunEnqueueCronJob and
WorkflowHandleStaledRunsCronJob, avoiding costly cache/metadata
hydration for a simple existence check
## Summary
- Introduces a pluggable `WebSearchDriver` abstraction (interface,
factory, service, module) so web search is no longer tied to native
provider tools (Anthropic/OpenAI)
- **Exa** is the first driver implementation with support for
category-filtered search (company, people, news, research paper, etc.) —
particularly useful for CRM workflows
- Per-query billing for both Exa ($0.007/query) and native provider
surcharges ($0.01/query for Anthropic/OpenAI) via the existing
`USAGE_RECORDED` pipeline
- New config variables: `WEB_SEARCH_DRIVER` (EXA/DISABLED),
`EXA_API_KEY`, `WEB_SEARCH_PREFER_NATIVE` (default false — prefers Exa
over native when both available)
- `WEB_SEARCH` operation type added for usage tracking and Stripe
metering
### Architecture
```
WebSearchDriver (interface)
├── ExaDriver — Exa neural search with category support
└── DisabledDriver — throws when search is disabled
WebSearchDriverFactory (extends DriverFactoryBase)
└── creates driver based on WEB_SEARCH_DRIVER config
WebSearchService (facade)
├── search(query, options?, billingContext?)
├── isEnabled()
└── emits USAGE_RECORDED events per query
WebSearchTool (Tool implementation)
└── registered in ActionToolProvider, available via tool catalog
```
### Native search billing gap fixed
Anthropic and OpenAI both charge $0.01/search on top of token costs. The
token costs were already billed, but the per-call surcharge was not.
Added `countNativeWebSearchCallsFromSteps` utility +
`billNativeWebSearchUsage` to `AiBillingService`, wired into both chat
and workflow agent paths.
## Test plan
- [ ] Set `WEB_SEARCH_DRIVER=EXA` + `EXA_API_KEY=...` and verify AI chat
can search the web
- [ ] Verify category parameter works (ask about a specific
company/person)
- [ ] Set `WEB_SEARCH_DRIVER=DISABLED` and verify search tool is not
exposed
- [ ] Set `WEB_SEARCH_PREFER_NATIVE=true` with Anthropic model and
verify native search is used
- [ ] Verify usage events are emitted in ClickHouse for both Exa and
native search paths
- [ ] Verify existing billing tests pass (`npx jest
ai-billing.service.spec.ts`)
Made with [Cursor](https://cursor.com)
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: claude[bot] <41898282+claude[bot]@users.noreply.github.com>
## Summary
- Add `WORKSPACE_SCHEMA_DDL_LOCKED` env-only boolean config variable
that blocks all workspace schema DDL changes when set to `true`. This is
intended for hot upgrades where logical replication cannot handle DDL
changes. Enforced at two chokepoints:
- `WorkspaceMigrationRunnerService.run` — blocks all metadata-driven DDL
(object/field/index CRUD, app sync/uninstall, standard app sync, upgrade
commands)
- `WorkspaceDataSourceService.createWorkspaceDBSchema` /
`deleteWorkspaceDBSchema` — blocks workspace creation (sign-up) and hard
deletion. Uses a dedicated `WorkspaceDataSourceException` (not
ForbiddenException)
- Add maintenance mode feature with Admin Panel UI and user-facing
banner:
- **Backend**: `MaintenanceModeService` stores maintenance window
(startAt, endAt, optional link) in `core.keyValuePair` as
`CONFIG_VARIABLE`. Validates endAt > startAt. Uses `GraphQLISODateTime`
scalar for date fields. Exposed via `clientConfig` REST endpoint and
admin GraphQL mutations (`setMaintenanceMode`, `clearMaintenanceMode`)
- **Admin Panel**: New "Maintenance Mode" section in Health tab with UTC
datetime pickers and activate/deactivate controls
- **Banner**: `InformationBannerMaintenance` displayed at the top of
`DefaultLayout` for all users, using Temporal API for timezone-aware
formatting with an optional "Learn more" link
These two features are **independent** — the DDL lock is controlled via
env var for operational use, while maintenance mode is a UI notification
mechanism controlled from the admin panel.
## Summary
- When `STORAGE_S3_PRESIGNED_URL_BASE` is configured, the file
controller returns a **302 redirect** to a presigned S3 URL instead of
proxying every byte through the server. This eliminates server bandwidth
and CPU overhead for S3-backed deployments.
- For local storage or S3 without a public endpoint, behavior is
unchanged (stream + pipe with security headers).
- Added `getPresignedUrl` to the `StorageDriver` interface (required
method returning `string | null`), with implementations in S3Driver
(uses a separate presign client with the public endpoint), LocalDriver
(returns `null`), and ValidatedStorageDriver (path traversal protection
+ delegation).
- Added a unified `getFileResponseById` method in `FileService` that
performs a single DB lookup and returns either a redirect URL or a
stream, avoiding double lookups.
- Extracted `getContentDisposition` from the header util so both the
proxy path and presigned URL path share the same inline/attachment
allowlist.
- Added MinIO service to `docker-compose.dev.yml` (optional `s3`
profile) for local S3 testing.
- Documented S3 presigned URL setup, CORS, and `nosniff` requirements in
the self-hosting docs.
## Test plan
- [x] All 63 unit tests pass across 5 test suites (util, S3 driver,
validated driver, file storage service, controller)
- [x] `npx nx typecheck twenty-server` passes
- [ ] Manual E2E test with MinIO: `docker compose --profile s3 up -d`,
configure S3 env vars, verify `curl -I` returns 302 with `Location`
header pointing to MinIO
- [ ] Verify local storage (no `STORAGE_S3_PRESIGNED_URL_BASE`) still
streams files with 200 + security headers
- [ ] Verify public assets endpoint still proxies (no redirect)
Made with [Cursor](https://cursor.com)
## Summary
- Replaces per-provider TypeScript constant files
(`openai-models.const.ts`, `anthropic-models.const.ts`, etc.) with a
single `ai-providers.json` catalog as the source of truth
- Adds runtime model discovery via AI SDK for self-hosted providers,
with `models.dev` enrichment for pricing/capabilities
- Introduces composite model IDs (`provider/modelId`) for canonical,
conflict-free identification
- Simplifies provider configuration: API keys are injected from
environment variables (e.g., `OPENAI_API_KEY`)
- Adds admin panel UI for provider management (add/remove/test), model
discovery, recommended model configuration, and default fast/smart model
selection per workspace
- Removes deprecated config variables (`AI_DISABLED_MODEL_IDS`,
`AUTO_ENABLE_NEW_AI_MODELS`, etc.)
- Adds database migration for composite model ID format
## Test plan
- [ ] Server typecheck passes
- [ ] Frontend typecheck passes
- [ ] Server unit tests pass
- [ ] Frontend unit tests pass
- [ ] CI pipeline green
- [ ] Admin panel AI tab loads correctly
- [ ] Provider discovery works for configured providers
- [ ] Model recommendation toggles persist
- [ ] Default fast/smart model selection works
Made with [Cursor](https://cursor.com)
## Summary
- Migrates `LogicFunctionModule`, `CodeInterpreterModule`, and
`CaptchaModule` from the `forRootAsync` + injection token pattern to the
`DriverFactoryBase` lazy-loading pattern (matching `EmailModule` and
`FileStorageModule`)
- Fixes#18724 where `LOGIC_FUNCTION_TYPE` was not respected in worker
processes because the driver was created at module boot time before the
DB config cache was loaded
- Removes `isEnvOnly` from `LOGIC_FUNCTION_TYPE`,
`CODE_INTERPRETER_TYPE`, `CAPTCHA_DRIVER`, `IS_MULTIWORKSPACE_ENABLED`,
and `FRONTEND_URL` — these can now be safely configured via the database
at runtime
## How it works
Each migrated module now uses a `DriverFactory` (extending
`DriverFactoryBase`) instead of a module-level async factory + Symbol
injection token:
1. **Lazy creation**: `getCurrentDriver()` creates the driver on first
call, after `DatabaseConfigDriver.onModuleInit()` has loaded the DB
cache
2. **Auto-recreation**: If config changes in the DB, the next
`getCurrentDriver()` call detects the key mismatch and creates a new
driver instance
3. **Unified config**: Both server and worker read from the same
database — driver config only needs to be set once
### Files deleted (old pattern)
- `logic-function-module.factory.ts`,
`logic-function-drivers.module.ts`, `logic-function-driver.constants.ts`
- `code-interpreter-module.factory.ts`
- `captcha.module-factory.ts`, `captcha-driver.constants.ts`
### Files created (new pattern)
- `logic-function-driver.factory.ts`
- `code-interpreter-driver.factory.ts`
- `captcha-driver.factory.ts`
Net: **-150 lines**
## Test plan
- [x] `npx nx typecheck twenty-server` passes
- [x] `npx nx lint:diff-with-main twenty-server` passes
- [ ] Integration tests pass (`npx nx run
twenty-server:test:integration:with-db-reset`)
- [ ] Verify logic functions execute in workflow runs (the original bug)
- [ ] Verify code interpreter works in workflow code steps
- [ ] Verify captcha validation works on sign-up (when captcha is
configured)
Made with [Cursor](https://cursor.com)
## Summary
- Removes the redundant `DATABASE_STATEMENT_TIMEOUT_MS` config variable
(default 15s) from `ConfigVariables`
- Updates the core TypeORM datasource to use
`PG_DATABASE_PRIMARY_TIMEOUT_MS` (default 10s) for its `query_timeout`,
aligning it with the workspace datasource which already uses this
variable
- This consolidates two separate env vars that controlled the same
concern (database query timeout) into a single one
## Summary
- **Split tsvector migration into individual per-field transactions**:
each tsvector field now runs in its own
`workspaceMigrationRunnerService.run()` call (its own DB transaction).
Since STORED generated columns trigger full table rewrites, a timeout on
one large table (e.g. `timelineActivity`) no longer rolls back the
others. Each field has its own idempotency check, so the migration is
fully resumable.
- **Add configurable `DATABASE_STATEMENT_TIMEOUT_MS` env var** (default
15000ms): controls the `query_timeout` on the core datasource globally,
allowing operators to raise it for long-running upgrade commands without
code changes.
- **Reorder 1.19 upgrade commands**: move
`fixRoleAndAgentUniversalIdentifiersCommand` first so that subsequent
commands see corrected universal identifiers.
## Summary
Implements enterprise licensing and per-seat billing for self-hosted
environments, with Stripe as the single source of truth for subscription
data.
### Components
- **twenty-website** hosts the private key to sign `ENTERPRISE_KEY` and
`ENTERPRISE_VALIDITY_TOKEN`. It communicates with Stripe to emit the
daily `ENTERPRISE_VALIDITY_TOKEN` if the subscription is active, based
on the user's Stripe subscription ID stored in `ENTERPRISE_KEY`.
- **Stripe** is the single source of truth for subscription data
(status, seats, billing).
- **The client** (twenty-server + DB + workers) saves `ENTERPRISE_KEY`
in the `keyValuePair` table (or `.env` if
`IS_CONFIG_VARIABLES_IN_DB_ENABLED` is false) and the daily-renewed
`ENTERPRISE_VALIDITY_TOKEN` in the `appToken` table.
`ENTERPRISE_VALIDITY_TOKEN` is verified client-side using a public key
to grant access to enterprise features (RLS, SSO, audit logs, etc.).
### Flow
1. When requesting an upgrade to an enterprise plan (from **Enterprise**
in settings), the user is shown a modal to choose monthly/yearly
billing, then redirected to Stripe to enter payment details. After
checkout, they land on twenty-website where they are exposed to their
`ENTERPRISE_KEY`, which they paste in the UI. It is saved in the
`keyValuePair` table. On activation, a first `ENTERPRISE_VALIDITY_TOKEN`
with 30-day validity is stored in the `appToken` table.
2. **Every day**, a cron job runs and does two things:
- **Refreshes the validity token**: communicates with twenty-website to
get a new `ENTERPRISE_VALIDITY_TOKEN` with 30-day validity if the Stripe
subscription is still active. If the subscription is in cancellation,
the emitted token has a validity equal to the cancellation date. If it's
no longer valid, the token is not replaced. The cron only needs to run
every 30 days in practice, but runs daily so it's resilient to
occasional failures.
- **Reports seat count**: counts active (non-soft-deleted)
`UserWorkspace` entries and sends the count to twenty-website, which
updates the Stripe subscription quantity with proration. Seats are also
reported on first activation. If the subscription is canceled or
scheduled for cancellation, the seat update is skipped.
3. `ENTERPRISE_VALIDITY_TOKEN` is verified server-side via a public key
to grant access to enterprise features.
### Key concepts
Three distinct checks are exposed as GraphQL fields on `Workspace`:
| Field | Meaning |
|---|---|
| `hasValidEnterpriseKey` | Has any valid enterprise key (signed JWT
**or** legacy plain string) |
| `hasValidSignedEnterpriseKey` | `ENTERPRISE_KEY` is a properly signed
JWT (billing portal makes sense) |
| `hasValidEnterpriseValidityToken` | `ENTERPRISE_VALIDITY_TOKEN` is
present and not expired (expiration depends on signed token payload, not
on "expiresAt" on appToken table which is only indicative) |
Feature access is gated by `isValid()` =
`hasValidEnterpriseValidityToken || hasValidEnterpriseKey` (to support
both new and legacy keys during transition). After transition isValid()
= hasValidEnterpriseValidityToken
### Frontend states
The Enterprise settings page handles multiple states:
- **No key**: show "Get Enterprise" with checkout modal
- **Orphaned validity token** (token valid but no signed key): prompt
user to set a valid enterprise key
- **Active/trialing but no validity token**: show subscription status
with a "Reload validity token" action
- **Active/trialing**: show full subscription info, billing portal
access, cancel option
- **Cancellation scheduled**: show cancellation date, billing portal
- **Canceled**: show billing history link and option to start a new
subscription
- **Past due / Incomplete**: prompt to update payment or restart
### Temporary retro-compatibility: legacy plain-text keys
Previously, enterprise features were gated by a simple check: any
non-empty string in `ENTERPRISE_KEY` granted access. With this PR, we
transition to a controlled system relying on signed JWTs.
To avoid breaking existing self-hosted users:
- **Legacy plain-text keys still grant access** to enterprise features.
`hasValidEnterpriseKey` returns `true` for both signed JWTs and plain
strings, and `isValid()` checks `hasValidEnterpriseKey` as a fallback
when no validity token is present.
- **A deprecation banner** is shown at the top of the app when
`hasValidEnterpriseKey` is `true` but `hasValidSignedEnterpriseKey` is
`false`, informing the user that their key format is deprecated and they
should activate a new signed key.
- **No billing portal or subscription management** is available for
legacy keys since there is no Stripe subscription to manage.
This retro-compatibility will be removed in a future version. At that
point, `isValid()` will only check `hasValidEnterpriseValidityToken`.
### Edge cases
- **Air-gapped / production environments**: for self-hosted clients that
block external traffic (or for our own production), provide a long-lived
`ENTERPRISE_VALIDITY_TOKEN` (e.g. 99 years) directly in the `appToken`
table, with no `ENTERPRISE_KEY`. The daily cron will skip the refresh
(no enterprise key to authenticate with), but the pre-seeded validity
token will be used to grant feature access. No billing or seat reporting
occurs in this mode.
- **`IS_CONFIG_VARIABLES_IN_DB_ENABLED` is false**: if the user tries to
activate an enterprise key but DB config writes are disabled, the
backend returns a clear error asking them to add `ENTERPRISE_KEY` to
their `.env` file manually.
- **Canceled subscriptions**: the `/seats` endpoint skips Stripe updates
for canceled or cancellation-scheduled subscriptions to avoid Stripe API
errors.
### How to test
- launch twenty-website on a different url (eg localhost:1002)
- add ENTERPRISE_API_URL=http://localhost:3002/api/enterprise (or else)
in your server .env
- ask me for twenty-website's .env file content (STRIPE_SECRET_KEY;
STRIPE_ENTERPRISE_MONTHLY_PRICE_ID;STRIPE_ENTERPRISE_YEARLY_PRICE_ID;
ENTERPRISE_JWT_PRIVATE_KEY; ENTERPRISE_JWT_PUBLIC_KEY;
NEXT_PUBLIC_WEBSITE_URL)
- visit Admin panel / enterprise
## Summary
For security reasons, the code interpreter and logic function drivers
now default based on `NODE_ENV`:
- **Production** (`NODE_ENV=production` or unset): Default to
**Disabled**
- **Development** (`NODE_ENV=development`): Default to **LOCAL** for
convenience
This ensures self-hosted production deployments don't accidentally run
user-provided code without explicit configuration.
## Changes
### Config (`config-variables.ts`)
- `CODE_INTERPRETER_TYPE`: Disabled in prod, LOCAL in dev
- `LOGIC_FUNCTION_TYPE`: Disabled in prod, LOCAL in dev
### Documentation (`setup.mdx`)
- Added **Security Defaults** section explaining NODE_ENV-based behavior
- Fixed variable names: `SERVERLESS_TYPE` → `LOGIC_FUNCTION_TYPE`,
`SERVERLESS_LAMBDA_*` → `LOGIC_FUNCTION_LAMBDA_*`
- Added **Code Interpreter** section with available drivers (Disabled,
Local, E2B)
### Environment files
- `.env.example`: Updated to `LOGIC_FUNCTION_TYPE` with comments
- `.env.test`: Added `LOGIC_FUNCTION_TYPE=LOCAL` for logic function
integration tests
Made with [Cursor](https://cursor.com)
---------
Co-authored-by: claude[bot] <41898282+claude[bot]@users.noreply.github.com>
Fixes#18356
## Summary
Setting `LOG_LEVELS=debug,info,error,warn` crashes with `TypeError:
logLevels.map is not a function` because `CastToLogLevelArray` silently
returns `undefined` for invalid levels.
Now it throws a clear error message listing the invalid levels and valid
options:
```
Invalid log level(s): info. Valid levels are: log, error, warn, debug, verbose
```
## Changes
- Throw descriptive `Error` when invalid log levels are provided instead
of returning `undefined`
- Updated tests to verify the error message
## Test plan
- [x] All 8 existing tests passing
- [x] `"toto"` → throws `Invalid log level(s): toto. Valid levels are:
log, error, warn, debug, verbose`
- [x] `"verbose,error,toto"` → throws listing only `toto` as invalid
- [x] Valid levels (`log,error,warn,debug,verbose`) continue working as
before
- Fixes self host application
- add new telemetry information
- add serverId to identify a server instance
- remove .twenty from git tracking
- tree-shake "twenty-sdk" usage in built logic functions and front
components
- fix "twenty-sdk" version usage
- fix twenty-zapier cli
---------
Co-authored-by: cubic-dev-ai[bot] <191113872+cubic-dev-ai[bot]@users.noreply.github.com>
## Summary
- **Admin-level filtering**: New AI tab in admin panel with server-wide
model availability controls (whitelist/blacklist via
`AI_AUTO_ENABLE_NEW_MODELS`, `AI_DISABLED_MODEL_IDS`,
`AI_ENABLED_MODEL_IDS` config variables). Dedicated
`setAdminAiModelEnabled` mutation replaces frontend config-variable
manipulation. Filter dropdown to show/hide unconfigured and deprecated
models.
- **Workspace-level filtering**: Per-workspace controls with "Use best
models only" mode (curated list backed by `isRecommended` flag), or
custom whitelist/blacklist. Separate Smart/Fast model selectors with
"Best (...)" virtual options.
- **Security enforcement**: Both layers enforced at every backend
execution point — workspace update, agent create/update, chat execution.
Model ID validated against known models before config mutation. All
admin endpoints protected by `AdminPanelGuard`.
## Changes
### Backend (`twenty-server`)
- New config variables for admin-level model filtering
- `AiModelRegistryService`: `getAllModelsWithStatus()`,
`setModelAdminEnabled()` with model ID validation,
`isModelAdminAllowed()`
- `AdminPanelResolver`: `getAdminAiModels` query,
`setAdminAiModelEnabled` mutation
- `WorkspaceEntity`: new fields (`autoEnableNewAiModels`,
`disabledAiModelIds`, `enabledAiModelIds`, `useRecommendedModels`)
- `WorkspaceService`: model validation on `smartModel`/`fastModel`
updates
- `AgentResolver`: model availability checks on create/update
- `isModelAllowedByWorkspace` centralized utility
- `isRecommended` flag on model definitions
- Two TypeORM migrations
### Frontend (`twenty-front`)
- New `SettingsAdminAI` component with search, filter dropdown
(unconfigured/deprecated), and model toggle cards
- AI tab added to admin panel navigation
- `useWorkspaceAiModelAvailability` hook for workspace-level filtering
- `SettingsAIModelsTab` redesigned: merged sections, "Use best models
only" toggle, conditional available models list
- `getModelIcon`/`getModelProviderLabel` shared utilities with GraphQL
enum casing normalization
- Updated generated GraphQL types and mock data
## Test plan
- [ ] Toggle models on/off in admin panel AI tab and verify they
appear/disappear in workspace settings
- [ ] Enable "Use best models only" in workspace settings and verify
only recommended models are selectable
- [ ] Disable recommended mode and verify whitelist/blacklist toggles
work correctly
- [ ] Verify deprecated models hidden by default, shown greyed out when
filter enabled
- [ ] Verify unconfigured models hidden by default, shown disabled when
filter enabled
- [ ] Try setting a disabled model as Smart/Fast model — should be
rejected
- [ ] Try creating an agent with a disabled model — should be rejected
- [ ] Verify admin panel AI tab requires admin access
Made with [Cursor](https://cursor.com)
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
## Summary
- **Model pricing overhaul**: All model constants updated with accurate
pricing in dollars per 1M tokens, including cached input rates, cache
creation rates, and tiered >200k context pricing
- **New providers**: Added Google (Gemini 3.x), Mistral, and AWS Bedrock
as inference providers. Bedrock serves Claude Opus 4.6 and Sonnet 4.6
via AWS, with proper credential handling following the existing S3/SES
pattern
- **InferenceProvider/ModelFamily split**: Refactored `ModelProvider`
into two orthogonal enums — `InferenceProvider` (who serves the model:
auth, SDK, metadata format) and `ModelFamily` (who created it: token
counting semantics). This eliminates growing `||` chains for token
normalization checks like `excludesCachedTokens`
- **Billing improvements**: Reasoning tokens charged at output rate,
cache token discounts applied accurately, real errors thrown to Sentry
on billing failures
## Test plan
- [x] All existing unit tests updated and passing (23 tests across 3
test files)
- [x] Lint passes for both twenty-server and twenty-front
- [ ] CI checks pass
Made with [Cursor](https://cursor.com)
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
⚠️ **AI-generated PR — not ready for review** ⚠️
cc @FelixMalfait
---
## Changes
### System prompt improvements
- Explicit skill-before-tools workflow to prevent the model from calling
tools without loading the matching skill first
- Data efficiency guidance (default small limits, use filters)
- Pluralized `load_skill` → `load_skills` for consistency with
`load_tools`
### Token usage reduction
- Output serialization layer: strips null/undefined/empty values from
tool results
- Lowered default `find_*` limit from 100 → 10, max from 1000 → 100
### System object tool generation
- System objects (calendar events, messages, etc.) now generate AI tools
- Only workflow-related and favorite-related objects are excluded
### Context window display fix
- **Bug**: UI compared cumulative tokens (sum of all turns) against
single-request context window → showed 100% after a few turns
- **Fix**: Track `conversationSize` (last step's `inputTokens`) which
represents the actual conversation history size sent to the model
- New `conversationSize` column on thread entity with migration
### Workspace AI instructions
- Support for custom workspace-level AI instructions
---------
Co-authored-by: claude[bot] <41898282+claude[bot]@users.noreply.github.com>
## Summary
- Removes unreachable dead code in `verifyJwtToken` — a legacy API key
verification block where the condition (`!payload.type && type ===
API_KEY`) was logically impossible. Also removes the now-unused
`isLegacyApiKey` parameter and `generateAppSecretLegacy` method.
- Adds explicit token type validation after JWT decode in
`verifyLoginToken`, `verifyRefreshToken`, and `verifyTransientToken`.
Each function now rejects tokens whose `type` field doesn't match what's
expected (defense-in-depth — the HMAC secret already binds the type, but
this makes the contract explicit).
## Test plan
- [x] Updated existing specs for login-token, refresh-token, renew-token
services — all 16 tests passing
- [x] Lint clean
Made with [Cursor](https://cursor.com)
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: claude[bot] <41898282+claude[bot]@users.noreply.github.com>
## Summary
- Renames `REFRESH_TOKEN_COOL_DOWN` to
`REFRESH_TOKEN_REUSE_GRACE_PERIOD` — the old name was misleading and
suggested a security mechanism rather than what it actually is: a grace
period for concurrent refresh token use (e.g. two browser tabs
refreshing simultaneously).
- Makes the token revocation in `renew-token.service.ts` conditional
(`revokedAt: IsNull()`), so if the token was already revoked by a
concurrent request, the original `revokedAt` timestamp is preserved and
the grace window stays anchored.
- Updates comments and config description to clarify intent.
## Test plan
- [x] Existing unit tests updated and passing
(`refresh-token.service.spec.ts`, `renew-token.service.spec.ts`)
- [x] Lint clean
Made with [Cursor](https://cursor.com)
Co-authored-by: Cursor <cursoragent@cursor.com>
## Summary
- Adds SSRF (Server-Side Request Forgery) protection to webhook requests
by using the same secure axios adapter already used by HTTP workflow
actions
- Prevents webhooks from making requests to private/internal IP
addresses (10.x, 192.168.x, 172.16-31.x, 169.254.x, localhost)
- Adds specific error logging when a webhook fails due to SSRF
protection
## Context
The HTTP workflow tool (`HTTP_REQUEST` action) already had SSRF
protection via `HTTP_TOOL_SAFE_MODE_ENABLED`, but webhooks were using
`HttpService` directly without this protection. This inconsistency meant
users could potentially configure webhooks to probe internal
infrastructure.
### What's protected now:
| Feature | Before | After |
|---------|--------|-------|
| HTTP Workflow Action | Protected (secure adapter) | Protected (secure
adapter) |
| Webhooks | **Unprotected** | Protected (secure adapter) |
### The secure adapter validates:
1. Protocol must be `http:` or `https:`
2. DNS resolution of hostname
3. Resolved IP must not be in private ranges
## Test plan
- [ ] Configure a webhook with an external URL (e.g.,
`https://webhook.site`) - should work
- [ ] Configure a webhook with `http://localhost:3000` - should fail
with SSRF error in audit log
- [ ] Configure a webhook with `http://10.0.0.1/test` - should fail with
SSRF error in audit log
- [ ] Configure a webhook with a domain that resolves to a private IP -
should fail
# Introduction
Followup of
https://github.com/twentyhq/twenty/pull/17001#pullrequestreview-3638508738
close https://github.com/twentyhq/core-team-issues/issues/1910
We've completely decom the `sync-metadata` in production. We're now then
removing its implementation in favor of the v2.
## TODO:
- [x] Remove sync-metadata implem and commands
- [x] Remove workspace decorators
- [x] Type each deprecated field to deprecated on their workspaceEntity
- [x] Remove the `workspace-sync-metadata` folder entirely
- [x] remove workspace migration
- [x] workspace migration removal migration
- [x] remove the `v2` references from workspace manager file names
- [x] remove the `v2` references from workspace manager modules
- [ ] Double check impact on translation file path updates
## Note
- Removed the gate logic
- Remains some service v2 naming, serverless needs to be migrated on v2
fully
- Removed workspaceMigration service app health consumption, making it
always returning up ( no more down ) cc @FelixMalfait ( quite obsolete
health check now, will require complete refactor once we introduce inter
app dependency etc )
Two challenges with error messages
- always provide a useful/meaningful error message for the end user
instead of the generic one. eg: show "Wrong password" and not "An error
occured"
- avoid technical details unless error regards a technical feature. eg:
show "An error occured" and not "Invalid post-hook payload."; but do
show "Invalid issuer URL." as it occurs while configuring SSO
What this PR does
- Make userFriendlyMessage mandatory for widely used
GraphqlQueryRunnerException and CommonQueryRunnerException, so that
developers are forced to ask themselves what the error message should
be, and as it contains very wide error codes (eg: "Bad request") which
should not be mapped to just one default message
- Keep userFriendlyMessage optional for service-specific exceptions (eg:
workflowStepExecutorException), but convert the error code to
userFriendlyMessage mapper to a switch case function with a typecheck
ensuring that all codes are mapped to a message. These default messages
are still overridable where they are thrown.
## Summary
Instead of throwing an error at server startup when LOCAL code
interpreter is configured in production, we now return a DisabledDriver
that only throws when the code interpreter is actually used.
## Changes
- Created `DisabledDriver` class that implements `CodeInterpreterDriver`
but throws an error only when `execute()` is called
- Added `DISABLED` to the `CodeInterpreterDriverType` enum
- Updated the factory to return a `DISABLED` driver config instead of
throwing when LOCAL is used in production
- Updated the module to handle the new `DISABLED` driver type
## Motivation
Many users don't need the code interpreter feature and want to deploy to
production without configuring E2B. Previously, the server would crash
at startup if `CODE_INTERPRETER_TYPE=LOCAL` was set in production.
**Before:** Server crashes at startup in production if
`CODE_INTERPRETER_TYPE=LOCAL`
**After:** Server starts fine. The error only occurs if someone actually
tries to **use** the code interpreter feature, at which point they get a
clear error message explaining how to configure E2B.
## Summary
This PR enforces that all custom exceptions must provide a
`userFriendlyMessage`, ensuring end users always see readable error
messages.
## Changes
### Core Changes
- **`CustomException` simplified**: Removed the `ForceFriendlyMessage`
generic parameter - `userFriendlyMessage` is now always required
- **Type safety**: The constructor now requires `{ userFriendlyMessage:
MessageDescriptor }` (no longer optional)
### Updated Files
- **74+ exception classes** updated to provide default user-friendly
messages using Lingui `msg` macro
- Each exception class has a sensible fallback message (e.g., `msg\`An
authentication error occurred.\``)
- Exception classes that had code-specific message maps retain their
behavior
## Benefits
- **Compile-time enforcement**: Forgetting to add a user-friendly
message now causes a TypeScript error
- **Better UX**: End users always see a localized, human-readable error
message
- **Simpler API**: No more boolean generic parameter to think about
## Testing
- `npx nx run twenty-server:typecheck` passes
- `npx nx run twenty-server:lint` passes
<!-- CURSOR_SUMMARY -->
---
> [!NOTE]
> Enforces `userFriendlyMessage` on `CustomException` and updates all
exception classes to supply localized default messages, with
filters/tests adjusted accordingly.
>
> - **Core**:
> - Enforce required `userFriendlyMessage` in `CustomException` (remove
optional generic; constructor now requires `{ userFriendlyMessage:
MessageDescriptor }`).
> - **Exceptions**:
> - Update ~70+ exception classes to set default localized messages via
Lingui `msg` maps and pass them in constructors (e.g., `AuthException`,
`ObjectMetadataException`, `FieldMetadataException`, etc.).
> - Add fallback messages where needed (e.g., `INTERNAL_SERVER_ERROR` or
domain-specific defaults).
> - **HTTP/GraphQL Filters**:
> - Ensure fallbacks create `UnknownException` with `msg` for
user-friendly text in REST/GraphQL exception filters.
> - **Tests**:
> - Adjust unit tests to pass `userFriendlyMessage` to exceptions.
> - Update Jest snapshots to include `extensions.userFriendlyMessage` or
message objects where applicable.
>
> <sup>Written by [Cursor
Bugbot](https://cursor.com/dashboard?tab=bugbot) for commit
221004fdfc0d97b7d152a258b347bf571e70f10e. This will update automatically
on new commits. Configure
[here](https://cursor.com/dashboard?tab=bugbot).</sup>
<!-- /CURSOR_SUMMARY -->
## Summary
- Add code interpreter tool that enables AI to execute Python code for
data analysis, CSV processing, and chart generation
- Support for both local (development) and E2B (sandboxed production)
execution drivers
- Real-time streaming of stdout/stderr and generated files
- Frontend components for displaying code execution results with
expandable sections
## Code Quality Improvements
- Extract `getMimeType` to shared utility to reduce code duplication
between drivers
- Fix security issue: escape single quotes/backslashes in E2B driver env
variable injection
- Add `buildExecutionState` helper to reduce duplicated state object
construction
- Add `DEFAULT_CODE_INTERPRETER_TIMEOUT_MS` constant for consistency
- Fix lingui linting warning and TypeScript theme errors in frontend
## Test Plan
- [ ] Test code interpreter with local driver in development
- [ ] Test code interpreter with E2B driver in production environment
- [ ] Verify streaming output displays correctly in chat UI
- [ ] Verify generated files (charts, CSVs) are uploaded and
downloadable
- [ ] Test file upload flow (CSV, Excel) triggers code interpreter
<!-- CURSOR_SUMMARY -->
---
> [!NOTE]
> Updates generated i18n catalogs for Polish and pseudo-English, adding
strings for code execution/output (code interpreter) and various UI
messages, with minor text adjustments.
>
> - **Localization**:
> - **Generated catalogs**: Refresh `locales/generated/pl-PL.ts` and
`locales/generated/pseudo-en.ts`.
> - Add strings for code execution/output (e.g., code, copy code/output,
running/waiting states, download files, generated files, Python code
execution).
> - Include new UI texts (errors, prompts, menus) and minor text
corrections.
> - No changes to `pt-BR`; other files unchanged functionally.
>
> <sup>Written by [Cursor
Bugbot](https://cursor.com/dashboard?tab=bugbot) for commit
befc13d02c21e5a6647bc1aa6daa2a89f60b7ef8. This will update automatically
on new commits. Configure
[here](https://cursor.com/dashboard?tab=bugbot).</sup>
<!-- /CURSOR_SUMMARY -->
## Summary
- Add latest AI models from OpenAI (GPT-4.1, o3, o4-mini), Anthropic
(Claude 4.5 Opus/Sonnet/Haiku), and xAI (Grok 4.1)
- Mark deprecated models (GPT-4o, GPT-4o-mini, GPT-4-turbo, Claude Opus
4, Claude Sonnet 4) with a `deprecated` flag
- Split AI models into separate files per provider for better
maintainability
- Support comma-separated default model lists for automatic fallback
across providers (works out of the box for self-hosters regardless of
which provider they configure)
- Filter deprecated models from dropdown selection while keeping them
functional for existing agents
## Changes
### New Models Added
| Provider | Models |
|----------|--------|
| OpenAI | gpt-4.1, gpt-4.1-mini, o3, o4-mini |
| Anthropic | claude-opus-4-5, claude-sonnet-4-5, claude-haiku-4-5 |
| xAI | grok-4-1-fast-reasoning |
### Deprecated Models
- gpt-4o, gpt-4o-mini, gpt-4-turbo (OpenAI)
- claude-opus-4-20250514, claude-sonnet-4-20250514 (Anthropic)
### Config Changes
Default model configs now support comma-separated fallback lists:
-
`DEFAULT_AI_SPEED_MODEL_ID=gpt-4.1-mini,claude-haiku-4-5-20251001,grok-3-mini`
-
`DEFAULT_AI_PERFORMANCE_MODEL_ID=gpt-4.1,claude-sonnet-4-5-20250929,grok-4`
## Test plan
- [x] Unit tests pass
- [x] Typecheck passes
- [x] Lint passes
- [ ] Verify deprecated models don't appear in model dropdowns
- [ ] Verify agents with deprecated models still work correctly
- [ ] Verify default model fallback works when only one provider is
configured
Should be merged once https://github.com/twentyhq/twenty/pull/16206 has
been released + command run to prod
In this PR
- Remove usage of viewGroup.fieldMetadataId, both in BE and FE states.
- But we still need to properly populate it until we fully remove
viewGroup.fieldMetadataId from db and ORM entity (upcoming 3rd PR out of
3). fieldMetadataId was removed from CoreViewGroup type and
CreateViewGroupInput and is determined BE-side based on the associated
view's mainGroupByFieldMetadataId. **I expect this means a downtime on
viewGroup creation, until both FE and BE are deployed and cache is
flushed.** This seems acceptable to me as it only regards viewGroup
creation.
- this information is replaced by view.mainGroupByFieldMetadataID
- Handle view group creation, update and deletion in the BE as a
side-effect of a view creation, update or deletion. Optimistic effects
are still used
- Add validation at view creation or update regarding
mainGroupByFieldMetadata
Left to do in 3rd PR
- Remove viewGroup.fieldMetadataId from db and ORM entity
- Restore feature allowing to update an existing grouped view's group by
field (already OK on BE side but need to rebuild FE optimistic)
Validations :
- relations count (in common api)
- oneToMany relation nested count (in common)
- requested fields count (in gql)
- root resolver count (in gql)
- root resolver duplicates (in gql)
- specific complexity for metadata / nesting count (in gql)
Charles Bochet