## Overview
This PR replaces the dynamic agent handoff system with a more
predictable planning-based router that decides upfront how to handle
multi-agent coordination.
## Major Changes
### 🔄 Architecture Shift: Handoffs → Planning
**Removed:**
- `AgentHandoffEntity` and handoff tracking system
- `AgentHandoffService` and `AgentHandoffExecutorService`
- Dynamic agent-to-agent transfers during execution
- Handoff tool generation and description templates
**Added:**
- `AiRouterService` with two strategies: `simple` (single agent) and
`planned` (multi-agent)
- `AgentPlanExecutorService` for executing multi-step plans
- Plan validation (cycle detection, dependency resolution)
- `UnifiedRouterResult` type with discriminated union
### 🤖 New Standard Agents
Added two new specialized agents:
- **Researcher Agent**: Web search, fact-finding, competitive
intelligence
- **Code Agent**: TypeScript function generation for serverless
workflows
### 🏗️ Router Refactoring (Latest)
Split router responsibilities into focused services:
- `AiRouterStrategyDeciderService`: Decides simple vs planned strategy
- `AiRouterPlanGeneratorService`: Generates and validates execution
plans
- `AiRouterService`: Coordinates between services (reduced from 426→275
lines)
### ⚙️ Configuration Improvements
- Added `outputStrategy` to agent definitions (`direct` vs `synthesize`)
- Removed hardcoded special cases for workflow-builder
- Added `plannerModel` field to workspace entity
- Increased `MAX_STEPS` from 10 to 25 for complex workflows
### 📝 Agent Prompt Refinements
Significantly simplified prompts for better clarity:
- Workflow Builder: 51→36 lines
- Helper: 49→28 lines
- Data Manipulator: Enhanced with sorting guidance
### 🔍 Enhanced Debugging
- Plan reasoning and step count in data message parts
- Router debug info with token usage tracking
- Better logging throughout execution pipeline
## Benefits
1. **Simpler Mental Model**: Router decides upfront vs dynamic transfers
2. **Better Predictability**: Users see the plan before execution
3. **Cleaner Architecture**: SRP with focused services
4. **Configuration Over Code**: Agent behavior via config, not hardcoded
logic
5. **Plan Validation**: Catches invalid dependencies and cycles
## Migration Notes
- Database migration removes `agentHandoff` table
- Adds `plannerModel` column to workspace table
- No API breaking changes (agent endpoints unchanged)
## Testing
- Integration tests updated to remove handoff dependencies
- Agent tool test utilities simplified
- Plan validation covered by new logic
## Next Steps (Future PRs)
- Parallel execution of independent plan steps
- Dynamic re-planning based on results
- Plan caching for common routing patterns
- Error recovery strategies in plan executor
## Context
Deprecating legacy ObjectMetadata from cache in favor of flat entities.
Introducing utils to build byName/byNameSingular/byNamePlural in
isolated cases
## Next
- I had to introduce a util to build from flat to legacy
objectMetadataMaps, we should instead use flat maps directly when needed
(datasource, schema generation, etc)
- Deprecate metadata version in the cache
- Use the new cache strategy for flat entities with permissions and
feature flags and inject in the global datasource context
## Summary
This PR adds configurable response format support for AI agents,
allowing them to return either plain text or structured JSON data based
on a defined schema.
## Key Features
### 1. Agent Response Format Configuration
- Added `AgentResponseFormat` type supporting:
- `text`: Returns plain text responses (default)
- `json`: Returns structured JSON based on defined schema
- New `AgentResponseSchema` type moved to `twenty-shared/ai` for sharing
between frontend/backend
### 2. Settings UI
- New `SettingsAgentResponseFormat` component for configuring response
format
- Visual schema builder for defining JSON output structure
- Real-time validation and preview
- Integrated into agent settings tab
### 3. Workflow Integration
- AI Agent workflow action automatically uses agent's configured
response format
- Output schema dynamically generated from agent's response format
- Workflow variable picker shows structured fields for JSON responses
- Backward compatible with existing text-only agents
### 4. Backend Implementation
- Added `convertAgentSchemaToZod` utility to validate JSON responses
- Agent executor service handles both text and JSON generation
- Automatic agent creation/cloning when adding AI agent steps to
workflows
- Unique agent naming with conflict resolution
### 5. Database Migration
- Migration `1763622159656-update-agent-response-format.ts`
- Sets default `responseFormat` to `{"type":"text"}` for existing agents
- Updated all standard agents with proper response format
## Changes by Module
### Frontend (`twenty-front`)
- 🆕 `AgentResponseFormat` type
- 🆕 `SettingsAgentResponseFormat` component
- ✏️ Updated `WorkflowEditActionAiAgent` to support response format
configuration
- 🗑️ Removed deprecated `useAiAgentOutputSchema` hook and
`AiAgentOutputSchema` type
### Backend (`twenty-server`)
- 🆕 `AgentResponseFormat` type in agent entity
- 🆕 `convertAgentSchemaToZod` utility for schema validation
- ✏️ Updated `AiAgentExecutorService` to handle both text and JSON
generation
- ✏️ Updated `WorkflowSchemaWorkspaceService` to generate output schema
from agent config
- ✏️ Enhanced `WorkflowVersionStepOperationsWorkspaceService` with agent
creation/cloning
- 🆕 Agent naming constants for conflict resolution
### Shared (`twenty-shared`)
- 🆕 `AgentResponseSchema` type
- 🆕 `ModelConfiguration` type moved to shared package
- Updated exports in `ai/index.ts`
## Code Quality
- Removed useless comments following code style guidelines
- All linter checks passed
- Type-safe implementation with proper TypeScript types
## Testing
- ✅ Database migration tested
- ✅ Agent creation/cloning in workflows verified
- ✅ Response format switching (text ↔ JSON) validated
- ✅ Backward compatibility with existing agents confirmed
## Migration Notes
- Existing agents will have `responseFormat: {type: 'text'}` set
automatically
- No breaking changes - all existing functionality preserved
- Agents can be updated to use JSON format through settings UI
# Introduction
Remove the v2 feature flag for view-field field-metadata and
object-metadata metadata entities
## Some details
- Disabled nestjs-query for object metadata creation and explicitly
calling it
- removed all v1 integration tests files
## Remarks
Not remove v2 referencing in both filenaming right now will handle that
globally later
## Breaking change
Due to object metadata resolver createOne standardization had to rename
the input from `CreateObjectInput` to `CreateOneObjectInput`
Adds intelligent routing system that automatically selects the best
agent for user queries based on conversation context.
### Changes:
- Added `routerModel` column to workspace table for configurable router
LLM selection
- Implemented `RouterService` with conversation history analysis and
agent matching logic
- Created router settings UI in AI Settings page with model dropdown
- Removed agent-specific thread associations - threads are now
agent-agnostic
- Added real-time routing status notification in chat UI with shimmer
effect
- Removed automatic default assistant agent creation
- Renamed GraphQL operations from agent-specific to generic (e.g.,
`agentChatThreads` → `chatThreads`)
---------
Co-authored-by: Félix Malfait <felix.malfait@gmail.com>
Co-authored-by: Félix Malfait <felix@twenty.com>
Implements permission intersection (AND logic) to prevent permission
escalation when agents act on behalf of users.
### Changes:
- **Permission Intersection**: Operations requiring both user AND agent
permissions
- **RoleContext Type**: Unified type supporting single `roleId` or
multiple `roleIds` for intersection
- **CRUD Services**: Updated to accept `roleContext` for granular
permission control
- **Agent Integration**: Chat agents now use user + agent role
intersection for all operations
- **ORM Layer**: Enhanced `getRepository` to support multi-role
permission checks
### Related:
- Part 2 of ["Acting on behalf of user" concept
PR](https://github.com/twentyhq/twenty/pull/15103)
[Closes#1661](https://github.com/twentyhq/core-team-issues/issues/1661)
---------
Co-authored-by: Félix Malfait <felix.malfait@gmail.com>
## Context
- All flatEntity should extend SyncableEntity
- SyncableEntity should now have applicationId and application relation
- Fix syncApp deletion, should now properly use migration v2 to delete
syncable entities
We want to update UserWorkspacePermissions from
```
export type UserWorkspacePermissions = {
permissionFlags: Record<PermissionFlagType, boolean>;
objectRecordsPermissions: Record<PermissionsOnAllObjectRecords, boolean>;
objectPermissions: ObjectsPermissionsDeprecated;
};
```
to
```
export type UserWorkspacePermissions = {
permissionFlags: Record<PermissionFlagType, boolean>;
objectsPermissions: ObjectsPermissions;
};
```
`ObjectsPermissionsDeprecated` and `ObjectsPermissions` are actually
very similar, they only have different key names (`canRead` vs
`canReadObjectRecords`)
To avoid brutal breaking changes, we will proceed in multiple steps:
1. This PR: adapt FE so it does not call objectRecordsPermissions
anymore + add the new objectsPermissions to UserWorkspacePermissions
without calling it yet in the FE
2. Remove objectRecordsPermissions in BE + use objectsPermissions in FE
instead of objectPermissions
3. Remove objectPermissions
- Renamed `WorkflowActionAdapter` to `ToolExecutorWorkflowAction`
- Renamed `settingPermission` table to `permissionFlag` and `setting`
column to `flag`
- Decoupled the send email logic from workflows to tools
- Add new `Tools Permission` section in FE
---------
Co-authored-by: Félix Malfait <felix.malfait@gmail.com>
Co-authored-by: Félix Malfait <felix@twenty.com>
# Introduction
Following `FieldMetadataInterface` deprecation in
https://github.com/twentyhq/twenty/pull/13264
As for the previous PR will rename and remove all the file in a
secondary PR to avoid conflicts and over loading this one
## Improvements
Removed optional properties from the `objectMetadataEntity` model and
added utils to retrieve test data
## Notes
By touching to `ObjectMetadataDTO` I would have expected a twenty-front
codegenerated types mutation, but it does not seem to be granular enough
to null/undefined coercion
- Added a generic HTTP request tool, allowing agents and workflows to
make HTTP requests to external APIs with configurable method, headers,
and body.
- Decoupled HTTP request workflow nodes from workflow-specific types and
factories, introducing a generic tool interface.
- Updated agent system prompts to include explicit guidance for the HTTP
request tool, including when and how to use it, and how to communicate
limitations.
### Demo
https://github.com/user-attachments/assets/129bc445-a277-4a19-95ab-09f890f8f051
## Context
Add an eventEmitter instance to twenty datasources so we can emit DB
events.
Add input and output formatting to twenty orm (formatData, formatResult)
Those 2 elements simplified existing logic when we interact with the
ORM, input will be formatted by the ORM so we can directly use
field-like structure instead of column-like. The output will be
formatted, for builder queries it will be in `result.generatedMaps`
where `result.raw` preserves the previous column-like structure.
Important change: We now have an authContext that we can pass when we
get a repository, this will be used for the different events emitted in
the ORM. We also removed the caching for repositories as it was not
scaling well and not necessary imho
Note: An upcoming PR should handle the onDelete: cascade behavior where
we send DESTROY events in cascade when there is an onDelete: CASCADE on
the FK.
---------
Co-authored-by: Charles Bochet <charles@twenty.com>
In this PR
- introduction of fieldPermission entity
- addition of upsertFieldPermission in role resolver
- computing of permissions taking fieldPermission into account. In order
to limit what is stored in Redis we only store fields restrictions. For
instance for objectMetadata with id XXX with a restriction on field with
id YYY we store:
`"XXX":{"canRead":true,"canUpdate":false,"canSoftDelete":false,"canDestroy":false,"restrictedFields":{"YYY":{"canRead":false,"canUpdate":null}}}`
---------
Co-authored-by: Charles Bochet <charlesBochet@users.noreply.github.com>
This PR introduces a significant enhancement to the role-based
permission system by extending it to support AI agents, enabling them to
perform database operations based on assigned permissions.
## Key Changes
### 1. Database Schema Migration
- **Table Rename**: `userWorkspaceRole` → `roleTargets` to better
reflect its expanded purpose
- **New Column**: Added `agentId` (UUID, nullable) to support AI agent
role assignments
- **Constraint Updates**:
- Made `userWorkspaceId` nullable to accommodate agent-only role
assignments
- Added check constraint `CHK_role_targets_either_agent_or_user`
ensuring either `agentId` OR `userWorkspaceId` is set (not both)
### 2. Entity & Service Layer Updates
- **RoleTargetsEntity**: Updated with new `agentId` field and constraint
validation
- **AgentRoleService**: New service for managing agent role assignments
with validation
- **AgentService**: Enhanced to include role information when retrieving
agents
- **RoleResolver**: Added GraphQL mutations for `assignRoleToAgent` and
`removeRoleFromAgent`
### 3. AI Agent CRUD Operations
- **Permission-Based Tool Generation**: AI agents now receive database
tools based on their assigned role permissions
- **Dynamic Tool Creation**: The `AgentToolService` generates CRUD tools
(`create_*`, `find_*`, `update_*`, `soft_delete_*`, `destroy_*`) for
each object based on role permissions
- **Granular Permissions**: Supports both global role permissions
(`canReadAllObjectRecords`) and object-specific permissions
(`canReadObjectRecords`)
### 4. Frontend Integration
- **Role Assignment UI**: Added hooks and components for
assigning/removing roles from agents
## Demo
https://github.com/user-attachments/assets/41732267-742e-416c-b423-b687c2614c82
---------
Co-authored-by: Antoine Moreaux <moreaux.antoine@gmail.com>
Co-authored-by: Lucas Bordeau <bordeau.lucas@gmail.com>
Co-authored-by: Charles Bochet <charles@twenty.com>
Co-authored-by: Guillim <guillim@users.noreply.github.com>
Co-authored-by: Charles Bochet <charlesBochet@users.noreply.github.com>
Co-authored-by: Weiko <corentin@twenty.com>
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: github-actions <github-actions@twenty.com>
Co-authored-by: Félix Malfait <felix.malfait@gmail.com>
Co-authored-by: Marie <51697796+ijreilly@users.noreply.github.com>
Co-authored-by: martmull <martmull@hotmail.fr>
Co-authored-by: Thomas Trompette <thomas.trompette@sfr.fr>
Co-authored-by: Etienne <45695613+etiennejouan@users.noreply.github.com>
Co-authored-by: Baptiste Devessier <baptiste@devessier.fr>
Co-authored-by: nitin <142569587+ehconitin@users.noreply.github.com>
Co-authored-by: Paul Rastoin <45004772+prastoin@users.noreply.github.com>
Co-authored-by: prastoin <paul@twenty.com>
Co-authored-by: Vicky Wang <157669812+vickywxng@users.noreply.github.com>
Co-authored-by: Vicky Wang <vw92@cornell.edu>
Co-authored-by: Raphaël Bosi <71827178+bosiraphael@users.noreply.github.com>
Félix Malfait