Adds user feedback and error handling to LLM registry modals (add/edit creator, model, provider) in the admin UI, including loading states and error messages. Ensures atomic updates for model costs in the backend using transactions. Improves display of creator website URLs and handles the case where no LLM models are available in analytics config. Updates icon usage and removes unnecessary 'use server' directive.
Refactored the creator deletion flow in CreatorsTable to use a new DeleteCreatorModal component, providing a confirmation dialog and improved error handling. The previous DeleteCreatorButton was removed and replaced for better user experience and safety.
Refactored backend imports and test mocks to use new admin LLM routes location. Cleaned up and reordered imports for clarity and consistency. Improved code formatting and readability across backend and frontend files. Renamed useLlmRegistryPage to getLlmRegistryPageData for clarity and updated all usages. No functional changes to business logic.
Introduces a migration to add the GPT-5.2 model and updates the O3 model slug in the database. Refactors backend LLM model registry usage for search and migration logic. Expands the OpenAPI spec with new admin endpoints for managing LLM models, providers, creators, and migrations.
Replaces usage of the custom BackendApi client and legacy types in admin LLM actions and components with generated OpenAPI endpoints and types. Updates API calls, error handling, and type imports throughout the admin LLM dashboard. Also corrects operationId fields in backend routes and OpenAPI spec for consistency.
Refactored admin LLM actions and components to improve code organization, update color schemes to use design tokens, and enhance UI consistency. Updated API types and endpoints to support model creators and migrations, and switched tables to use shared Table components. Added and documented new API endpoints for model migrations, creators, and usage in openapi.json.
<!-- ⚠️ Reminder: Think about your Changeset/Docs changes! -->
<!-- If you are introducing new blocks or features, document them for
users. -->
<!-- Reference:
https://github.com/Significant-Gravitas/AutoGPT/blob/dev/CONTRIBUTING.md
-->
## Summary
This PR fixes two related issues with number/integer inputs in the
frontend:
1. **HTMLType typo fix**: INTEGER input type was incorrectly mapped to
`htmlType: 'account'` (which is not a valid HTML input type) instead of
`htmlType: 'number'`.
2. **AnyOfField toggle fix**: When a user cleared a number input field,
the input would disappear because `useAnyOfField` checked for both
`null` AND `undefined` in `isEnabled`. This PR changes it to only check
for explicit `null` (set by toggle off), allowing `undefined` (empty
input) to keep the field visible.
### Root cause analysis
When a user cleared a number input:
1. `handleChange` returned `undefined` (because `v === "" ? undefined :
Number(v)`)
2. In `useAnyOfField`, `isEnabled = formData !== null && formData !==
undefined` became `false`
3. The input field disappeared
### Fix
Changed `useAnyOfField.tsx` line 67:
```diff
- const isEnabled = formData !== null && formData !== undefined;
+ const isEnabled = formData !== null;
```
This way:
- When toggle is OFF → `formData` is `null` → `isEnabled` is `false`
(input hidden) ✓
- When toggle is ON but input is cleared → `formData` is `undefined` →
`isEnabled` is `true` (input visible) ✓
## Test plan
- [x] Verified INTEGER inputs now render correctly with `type="number"`
- [x] Verified clearing a number input keeps the field visible
- [x] Verified toggling the nullable switch still works correctly
Fixes#11594🤖 AI-assisted development disclaimer: This PR was developed with
assistance from Claude Code.
---------
Signed-off-by: majiayu000 <1835304752@qq.com>
Co-authored-by: Abhimanyu Yadav <122007096+Abhi1992002@users.noreply.github.com>
Add auto-selection of creator based on provider in AddModelModal for better usability. Update EditModelModal to use a select dropdown for credential provider, add helper text, and set credential_type as a hidden default input.
Improves backend handling of LLM model cost updates by separating scalar and relation field updates, ensuring costs are deleted and recreated as needed. Optional cost fields are now only included if present, and metadata is handled as a Prisma Json type. On the frontend, removes the 'Enabled' checkbox from the EditModelModal component.
## Changes 🏗️
Move the `<NodeInput />` component next to the old builder code where it
is used.
## Checklist 📋
### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Run app locally and click around, E2E is fine
(#11693)
Issues fixed by this PR
- https://github.com/Significant-Gravitas/AutoGPT/issues/11688
- https://github.com/Significant-Gravitas/AutoGPT/issues/11687
### **Changes 🏗️**
Added keyboard delete functionality to the ReactFlow editor by enabling
the `deleteKeyCode` prop with both "Backspace" and "Delete" keys. This
allows users to delete selected nodes and edges using standard keyboard
shortcuts, improving the editing experience.
**Modified:**
- `Flow.tsx`: Added `deleteKeyCode={["Backspace", "Delete"]}` prop to
the ReactFlow component to enable deletion of selected elements via
keyboard
### **Checklist 📋**
#### **For code changes:**
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Select a node in the flow editor and press Delete key to confirm
it deletes
- [x] Select a node in the flow editor and press Backspace key to
confirm it deletes
- [x] Verify deletion works for multiple selected elements
Introduces the LlmModelCreator entity to distinguish model creators (e.g., OpenAI, Meta) from providers, with full CRUD API endpoints, database migration, and Prisma schema updates. Backend and frontend are updated to support associating models with creators, including admin UI for managing creators and selecting them when creating or editing models. Existing models are backfilled with known creators via migration.
## Changes 🏗️
<img width="800" height="964" alt="Screenshot 2026-01-05 at 15 26 21"
src="https://github.com/user-attachments/assets/f8c7fc47-894a-4db2-b2f1-62b4d70e8453"
/>
- Adjust the new builder to use the Design System components
- Re-structure imports to match formatting rules
- Small improvement on `use-get-flag`
- Move file which is the main hook
## Checklist 📋
### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Run locally and check the new buttons look good
(#11658)
## Summary
Implements an auto-save draft recovery system that persists unsaved flow
builder state across browser sessions, tab closures, and refreshes. When
users return to a flow with unsaved changes, they can choose to restore
or discard the draft via an intuitive recovery popup.
https://github.com/user-attachments/assets/0f77173b-7834-48d2-b7aa-73c6cd2eaff6
## Changes 🏗️
### Core Features
- **Draft Recovery Popup** (`DraftRecoveryPopup.tsx`)
- Displays amber-themed notification with unsaved changes metadata
- Shows node count, edge count, and relative time since last save
- Provides restore and discard actions with tooltips
- Auto-dismisses on click outside or ESC key
- **Auto-Save System** (`useDraftManager.ts`)
- Automatically saves draft state every 15 seconds
- Saves on browser tab close/refresh via `beforeunload`
- Tracks nodes, edges, graph schemas, node counter, and flow version
- Smart dirty checking - only saves when actual changes detected
- Cleans up expired drafts (24-hour TTL)
- **IndexedDB Persistence** (`db.ts`, `draft-service.ts`)
- Uses Dexie library for reliable client-side storage
- Handles both existing flows (by flowID) and new flows (via temp
session IDs)
- Compares draft state with current state to determine if recovery
needed
- Automatically clears drafts after successful save
### Integration Changes
- **Flow Editor** (`Flow.tsx`)
- Integrated `DraftRecoveryPopup` component
- Passes `isInitialLoadComplete` state for proper timing
- **useFlow Hook** (`useFlow.ts`)
- Added `isInitialLoadComplete` state to track when flow is ready
- Ensures draft check happens after initial graph load
- Resets state on flow/version changes
- **useCopyPaste Hook** (`useCopyPaste.ts`)
- Refactored to manage keyboard event listeners internally
- Simplified integration by removing external event handler setup
- **useSaveGraph Hook** (`useSaveGraph.ts`)
- Clears draft after successful save (both create and update)
- Removes temp flow ID from session storage on first save
### Dependencies
- Added `dexie@4.2.1` - Modern IndexedDB wrapper for reliable
client-side storage
## Technical Details
**Auto-Save Flow:**
1. User makes changes to nodes/edges
2. Change triggers 15-second debounced save
3. Draft saved to IndexedDB with timestamp
4. On save, current state compared with last saved state
5. Only saves if meaningful changes detected
**Recovery Flow:**
1. User loads flow/refreshes page
2. After initial load completes, check for existing draft
3. Compare draft with current state
4. If different and non-empty, show recovery popup
5. User chooses to restore or discard
6. Draft cleared after either action
**Session Management:**
- Existing flows: Use actual flowID for draft key
### Test Plan 🧪
- [x] Create a new flow with 3+ blocks and connections, wait 15+
seconds, then refresh the page - verify recovery popup appears with
correct counts and restoring works
- [x] Create a flow with blocks, refresh, then click "Discard" button on
recovery popup - verify popup disappears and draft is deleted
- [x] Add blocks to a flow, save successfully - verify draft is cleared
from IndexedDB (check DevTools > Application > IndexedDB)
- [x] Make changes to an existing flow, refresh page - verify recovery
popup shows and restoring preserves all changes correctly
- [x] Verify empty flows (0 nodes) don't trigger recovery popup or save
drafts
(#11669)
Fixes duplicate "Publish to Marketplace" buttons in the builder by
adding a `showTrigger` prop to control modal trigger visibility.
<img width="296" height="99" alt="Screenshot 2025-12-23 at 8 18 58 AM"
src="https://github.com/user-attachments/assets/d5dbfba8-e854-4c0c-a6b7-da47133ec815"
/>
### Changes 🏗️
**BuilderActionButton.tsx**
- Removed borders on hover and active states for a cleaner visual
appearance
- Added `hover:border-none` and `active:border-none` to maintain
consistent styling during interactions
**PublishToMarketplace.tsx**
- Pass `showTrigger={false}` to `PublishAgentModal` to hide the default
trigger button
- This prevents duplicate buttons when a custom trigger is already
rendered
**PublishAgentModal.tsx**
- Added `showTrigger` prop (defaults to `true`) to conditionally render
the modal trigger
- Allows parent components to control whether the built-in trigger
button should be displayed
- Maintains backward compatibility with existing usage
### Checklist 📋
#### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Verify only one "Publish to Marketplace" button appears in the
builder
- [x] Confirm button hover/active states display correctly without
border artifacts
- [x] Verify modal can still be triggered programmatically without the
trigger button
## Changes 🏗️
- Make `<Navbar />` a client component so its rendering is more
predictable
- Remove the `useMemo()` for the chat link to prevent the flash...
- Make sure chat is added to the navbar links only after checking the
flag is enabled
- Improve logout with `useTransition`
- Simplify feature flags setup
## Checklist 📋
### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Run locally and test the above
<!-- CURSOR_SUMMARY -->
---
> [!NOTE]
> Ensures the `Chat` nav item is hidden when the feature flag is off
across desktop and mobile nav.
>
> - Inline-filters `loggedInLinks` to skip `Chat` when `Flag.CHAT` is
false for both `NavbarLink` rendering and `MobileNavBar` menu items
> - Removes `useMemo`/`linksWithChat` helper; maps directly over
`loggedInLinks` and filters nulls in mobile, keeping icon mapping intact
> - Cleans up unused `useMemo` import
>
> <sup>Written by [Cursor
Bugbot](https://cursor.com/dashboard?tab=bugbot) for commit
79c42d87b4adb05155be684e1d0576073e872680. This will update automatically
on new commits. Configure
[here](https://cursor.com/dashboard?tab=bugbot).</sup>
<!-- /CURSOR_SUMMARY -->
## Changes 🏗️
Use the Design System `<Dialog />` on the old builder, which supports
long content scrolling ( the current one does not, causing issues in
graphs with many run inputs )...
## Checklist 📋
### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Run locally and test the above
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Added Enhanced Rendering toggle for improved output handling and
display (controlled via feature flag)
* **Improvements**
* Refined dialog layouts and user interactions
* Enhanced copy-to-clipboard functionality with toast notifications upon
copying
<sub>✏️ Tip: You can customize this high-level summary in your review
settings.</sub>
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Changes 🏗️
Sometimes, on Dev, when navigating between pages, the Favico colour
would revert from Green 🟢 (Dev) to Purple 🟣(Default). That's because the
`/marketplace` page had custom code overriding it that I didn't notice
earlier...
I also made it use the Next.js metadata API, so it handles the favicon
correctly across navigations.
## Checklist 📋
### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Run locally and test the above
## Summary
This PR implements a comprehensive marketplace update notification
system that allows users to discover and update to newer agent versions,
along with enhanced publishing workflows and UI improvements.
<img width="1500" height="533" alt="image"
src="https://github.com/user-attachments/assets/ee331838-d712-4718-b231-1f9ec21bcd8e"
/>
<img width="600" height="610" alt="image"
src="https://github.com/user-attachments/assets/b881a7b8-91a5-460d-a159-f64765b339f1"
/>
<img width="1500" height="416" alt="image"
src="https://github.com/user-attachments/assets/a2d61904-2673-4e44-bcc5-c47d36af7a38"
/>
<img width="1500" height="1015" alt="image"
src="https://github.com/user-attachments/assets/2dd978c7-20cc-4230-977e-9c62157b9f23"
/>
## Core Features
### 🔔 Marketplace Update Notifications
- **Update detection**: Automatically detects when marketplace has newer
agent versions than user's local copy
- **Creator notifications**: Shows banners for creators with unpublished
changes ready to publish
- **Non-creator support**: Enables regular users to discover and update
to newer marketplace versions
- **Version comparison**: Intelligent logic comparing `graph_version` vs
marketplace listing versions
### 📋 Enhanced Publishing Workflow
- **Builder integration**: Added "Publish to Marketplace" button
directly in the builder actions
- **Unified banner system**: Consistent `MarketplaceBanners` component
across library and marketplace pages
- **Streamlined UX**: Fixed layout issues, improved button placement and
styling
- **Modal improvements**: Fixed thumbnail loading race conditions and
infinite loop bugs
### 📚 Version History & Changelog
- **Inline version history**: Added version changelog directly to
marketplace agent pages
- **Version comparison**: Clear display of available versions with
current version highlighting
- **Update mechanism**: Direct updates using `graph_version` parameter
for accuracy
## Technical Implementation
### Backend Changes
- **Database schema**: Added `agentGraphVersions` and `agentGraphId`
fields to `StoreAgent` model
- **API enhancement**: Updated store endpoints to expose graph version
data for version comparison
- **Data migration**: Fixed agent version field naming from `version` to
`agentGraphVersions`
- **Model updates**: Enhanced `LibraryAgentUpdateRequest` with
`graph_version` field
### Frontend Architecture
- **`useMarketplaceUpdate` hook**: Centralized marketplace update
detection and creator identification
- **`MarketplaceBanners` component**: Unified banner system with proper
vertical layout and styling
- **`AgentVersionChangelog` component**: Version history display for
marketplace pages
- **`PublishToMarketplace` component**: Builder integration with modal
workflow
### Key Bug Fixes
- **Thumbnail loading**: Fixed race condition where images wouldn't load
on first modal open
- **Infinite loops**: Used refs to prevent circular dependencies in
`useThumbnailImages` hook
- **Layout issues**: Fixed banner placement, removed duplicate
breadcrumbs, corrected vertical layout
- **Field naming**: Fixed `agent_version` vs `version` field
inconsistencies across APIs
## Files Changed
### Backend
- `autogpt_platform/backend/backend/server/v2/store/` - Enhanced store
API with graph version data
- `autogpt_platform/backend/backend/server/v2/library/` - Updated
library API models
- `autogpt_platform/backend/migrations/` - Database migrations for
version fields
- `autogpt_platform/backend/schema.prisma` - Schema updates for graph
versions
### Frontend
- `src/app/(platform)/components/MarketplaceBanners/` - New unified
banner component
- `src/app/(platform)/library/agents/[id]/components/` - Enhanced
library views with banners
- `src/app/(platform)/build/components/BuilderActions/` - Added
marketplace publish button
- `src/app/(platform)/marketplace/components/AgentInfo/` - Added inline
version history
- `src/components/contextual/PublishAgentModal/` - Fixed thumbnail
loading and modal workflow
## User Experience Impact
- **Better discovery**: Users automatically notified of newer agent
versions
- **Streamlined publishing**: Direct publish access from builder
interface
- **Reduced friction**: Fixed UI bugs, improved loading states,
consistent design
- **Enhanced transparency**: Inline version history on marketplace pages
- **Creator workflow**: Better notifications for creators with
unpublished changes
## Testing
- ✅ Update banners appear correctly when marketplace has newer versions
- ✅ Creator banners show for users with unpublished changes
- ✅ Version comparison logic works with graph_version vs marketplace
versions
- ✅ Publish button in builder opens modal correctly with pre-populated
data
- ✅ Thumbnail images load properly on first modal open without infinite
loops
- ✅ Database migrations completed successfully with version field fixes
- ✅ All existing tests updated and passing with new schema changes
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Lluis Agusti <hi@llu.lu>
Co-authored-by: Ubbe <hi@ubbe.dev>
Co-authored-by: Reinier van der Leer <pwuts@agpt.co>
<!-- Clearly explain the need for these changes: -->
The delete button on flow editor edges is always visible, which creates
visual clutter. This change makes the button only appear on hover,
improving the UI while keeping it accessible.
### Changes 🏗️
- Added hover state management using `useState` to track when the edge
delete button is hovered
- Applied opacity transition to the delete button (fades in on hover,
fades out when not hovered)
- Added `onMouseEnter` and `onMouseLeave` handlers to the button to
control hover state
- Used `cn` utility for conditional className management
- Button remains interactive even when `opacity-0` (still clickable for
better UX)
### Checklist 📋
#### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Hover over an edge in the flow editor and verify the delete button
fades in smoothly
- [x] Move mouse away from edge and verify the delete button fades out
smoothly
- [x] Click the delete button while hovered to verify it still removes
the edge connection
- [x] Test with multiple edges to ensure hover state is independent per
edge
- #11603
### Changes 🏗️
Frontend:
- Make `okData` infer the response data type instead of casting
- Generalize infinite query utilities from `SidebarRunsList/helpers.ts`
- Move to `@/app/api/helpers` and use wherever possible
- Simplify/replace boilerplate checks and conditions with `okData` in
many places
- Add `useUserTimezone` hook to replace all the boilerplate timezone
queries
Backend:
- Fix response type annotation of `GET
/api/store/graph/{store_listing_version_id}` endpoint
- Fix documentation and error behavior of `GET
/api/review/execution/{graph_exec_id}` endpoint
### Checklist 📋
#### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- CI passes
- [x] Clicking around the app manually -> no obvious issues
- [x] Test Onboarding step 5 (run)
- [x] Library runs list loads normally
We'll soon be needing a more feature-complete external API. To make way
for this, I'm moving some files around so:
- We can more easily create new versions of our external API
- The file structure of our internal API is more homogeneous
These changes are quite opinionated, but IMO in any case they're better
than the chaotic structure we have now.
### Changes 🏗️
- Move `backend/server` -> `backend/api`
- Move `backend/server/routers` + `backend/server/v2` ->
`backend/api/features`
- Change absolute sibling imports to relative imports
- Move `backend/server/v2/AutoMod` -> `backend/executor/automod`
- Combine `backend/server/routers/analytics_*test.py` ->
`backend/api/features/analytics_test.py`
- Sort OpenAPI spec file
### Checklist 📋
#### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- CI tests
- [x] Clicking around in the app -> no obvious breakage
## Summary
• Redesigned Human-in-the-Loop review interface with yellow warning
scheme
• Implemented separate approved_data/rejected_data output pins for
human_in_the_loop block
• Added real-time execution status tracking to legacy flow for review
detection
• Fixed button loading states and improved UI consistency across flows
• Standardized Tailwind CSS usage removing custom values
<img width="1500" alt="image"
src="https://github.com/user-attachments/assets/4ca6dd98-f3c4-41c0-a06b-92b3bca22490"
/>
<img width="1500" alt="image"
src="https://github.com/user-attachments/assets/0afae211-09f0-465e-b477-c3949f13c876"
/>
<img width="1500" alt="image"
src="https://github.com/user-attachments/assets/05d9d1ed-cd40-4c73-92b8-0dab21713ca9"
/>
## Changes Made
### Backend Changes
- Modified `human_in_the_loop.py` block to output separate
`approved_data` and `rejected_data` pins instead of single reviewed_data
with status
- Updated block output schema to support better data flow in graph
builder
### Frontend UI Changes
- Redesigned PendingReviewsList with yellow warning color scheme
(replacing orange)
- Fixed button loading states to show spinner only on clicked button
- Improved FloatingReviewsPanel layout removing redundant headers
- Added real-time status tracking to legacy flow using useFlowRealtime
hook
- Fixed AgentActivityDropdown text overflow and layout issues
- Enhanced Safe Mode toggle positioning and toast timing
- Standardized all custom Tailwind values to use standard classes
### Design System Updates
- Added yellow design tokens (25, 150, 600) for warning states
- Unified REVIEW status handling across all components
- Improved component composition patterns
## Test Plan
- [x] Verify HITL blocks create separate output pins for
approved/rejected data
- [x] Test review flow works in both new and legacy flow builders
- [x] Confirm button loading states work correctly (only clicked button
shows spinner)
- [x] Validate AgentActivityDropdown properly displays review status
- [x] Check Safe Mode toggle positioning matches old flow
- [x] Ensure real-time status updates work in legacy flow
- [x] Verify yellow warning colors are consistent throughout
🤖 Generated with [Claude Code](https://claude.ai/code)
---------
Co-authored-by: Lluis Agusti <hi@llu.lu>
We want to provide Single Sign-On for multiple AutoGPT apps that use the
Platform as their backend.
### Changes 🏗️
Backend:
- DB + logic + API for OAuth flow (w/ tests)
- DB schema additions for OAuth apps, codes, and tokens
- Token creation/validation/management logic
- OAuth flow endpoints (app info, authorize, token exchange, introspect,
revoke)
- E2E OAuth API integration tests
- Other OAuth-related endpoints (upload app logo, list owned apps,
external `/me` endpoint)
- App logo asset management
- Adjust external API middleware to support auth with access token
- Expired token clean-up job
- Add `OAUTH_TOKEN_CLEANUP_INTERVAL_HOURS` setting (optional)
- `poetry run oauth-tool`: dev tool to test the OAuth flows and register
new OAuth apps
- `poetry run export-api-schema`: dev tool to quickly export the OpenAPI
schema (much quicker than spinning up the backend)
Frontend:
- Frontend UI for app authorization (`/auth/authorize`)
- Re-redirect after login/signup
- Frontend flow to batch-auth integrations on request of the client app
(`/auth/integrations/setup-wizard`)
- Debug `CredentialInputs` component
- Add `/profile/oauth-apps` management page
- Add `isOurProblem` flag to `ErrorCard` to hide action buttons when the
error isn't our fault
- Add `showTitle` flag to `CredentialsInput` to hide built-in title for
layout reasons
DX:
- Add [API
guide](https://github.com/Significant-Gravitas/AutoGPT/blob/pwuts/sso/docs/content/platform/integrating/api-guide.md)
and [OAuth
guide](https://github.com/Significant-Gravitas/AutoGPT/blob/pwuts/sso/docs/content/platform/integrating/oauth-guide.md)
### Checklist 📋
#### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Manually verify test coverage of OAuth API tests
- Test `/auth/authorize` using `poetry run oauth-tool test-server`
- [x] Works
- [x] Looks okay
- Test `/auth/integrations/setup-wizard` using `poetry run oauth-tool
test-server`
- [x] Works
- [x] Looks okay
- Test `/profile/oauth-apps` page
- [x] All owned OAuth apps show up
- [x] Enabling/disabling apps works
- [ ] ~~Uploading logos works~~ can only test this once deployed to dev
#### For configuration changes:
- [x] `.env.default` is updated or already compatible with my changes
- [x] `docker-compose.yml` is updated or already compatible with my
changes
- [x] I have included a list of my configuration changes in the PR
description (under **Changes**)
Introduced a new LlmModelMigration model to track migrations when disabling LLM models, allowing for revert capability. Updated the toggle model API to create migration records with optional reason and custom pricing. Added endpoints for listing and reverting migrations, along with corresponding frontend actions and UI components to manage migrations effectively. Enhanced the admin dashboard to display active migrations, improving overall usability and tracking of model changes.
Updated the toggle LLM model API to include an optional migration feature, allowing workflows to be migrated to a specified replacement model when disabling a model. Refactored related request and response models to accommodate this change. Improved error handling and logging for better debugging. Updated frontend actions and components to support the new migration parameter.
## Summary
- Fixed auto-frame timing in new builder - now calls `fitView` after
nodes are rendered instead of on mount
- Replaced manual viewport calculation in legacy builder with React
Flow's `fitView` for consistency
- Both builders now properly center and frame all blocks when opening an
agent
## Test plan
- [x] Open an existing agent with multiple blocks in the new builder -
verify all blocks are visible and centered
- [x] Open an existing agent in the legacy builder - verify all blocks
are visible and centered
- [x] Verify the manual "Frame" button still works correctly
## Changes 🏗️
Adds the following improvements:
### Prevent credential row overflowing on mobile 📱
**Before**
<img width="300" height="469" alt="Screenshot 2025-12-15 at 16 42 05"
src="https://github.com/user-attachments/assets/0d27394c-cec9-45a4-be82-804827343212"
/>
**After**
<img width="300" height="446" alt="Screenshot 2025-12-15 at 16 44 22"
src="https://github.com/user-attachments/assets/0f19e220-500d-4488-955e-612d38704727"
/>
_Just hide the ****** on mobile..._
### Make touch targets bigger on 📱 on the mobile menu
**Before**
<img width="300" height="607" alt="Screenshot 2025-12-15 at 16 58 28"
src="https://github.com/user-attachments/assets/762b7d4e-5269-41a4-88d2-ea745c50324e"
/>
Touch targets were quite small on mobile, especially for people with big
fingers...
**After**
<img width="300" height="589" alt="Screenshot 2025-12-15 at 16 54 02"
src="https://github.com/user-attachments/assets/beede0c4-5439-47a9-8bec-143b44306c6b"
/>
### New `<OverflowText />` component
<img width="600" height="551" alt="Screenshot 2025-12-15 at 16 48 20"
src="https://github.com/user-attachments/assets/a969223f-dd6a-497a-857e-18483aea28d7"
/>
A component that will render text like `<Text />`, but automatically
displays `...` and the full text content on a tooltip if it detects
there is no space for the full text length. Pretty useful for the type
of dashboard we are building, where sometimes titles or user-generated
content can be quite long, making the UI look whack.
### Google Drive Picker
Only allow the removal of files if it is not in read-only mode.
## Checklist 📋
### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Checkout branch locally
- [x] Test the above
- Updated Next.js configuration to set body size limits for server
actions and API routes.
- Enhanced error handling in the API client to provide user-friendly
messages for file size errors.
- Added user-friendly error messages for 413 Payload Too Large responses
in API error parsing.
These changes ensure that file uploads are consistent with backend
limits and improve user experience during uploads.
### Checklist 📋
#### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
<!-- Put your test plan here: -->
- [x] Upload a file bigger than 10MB and it works
- [X] Upload a file bigger than 256MB and you see a official error
stating the max file size is 256MB
<!-- Clearly explain the need for these changes: -->
Agent blocks require different handling compared to standard blocks,
particularly for:
- Handle ID generation (using direct keys instead of generated IDs)
- Form data storage structure (nested under `inputs` key)
- Field ID parsing (filtering out schema path prefixes)
This PR implements special handling for `BlockUIType.AGENT` throughout
the form rendering and output handling components to ensure agents work
correctly in the flow editor.
### Changes 🏗️
<!-- Concisely describe all of the changes made in this pull request:
-->
- **CustomNode.tsx**: Pass `uiType` prop to `OutputHandler` component
- **FormCreator.tsx**:
- Store agent form data in `hardcodedValues.inputs` instead of directly
in `hardcodedValues`
- Extract initial values from `hardcodedValues.inputs` for agent blocks
- **OutputHandler.tsx**:
- Accept `uiType` prop
- Use direct key as handle ID for agents instead of
`generateHandleId(key)`
- **useMarketplaceAgentsContent.ts**:
- Fetch full agent details using `getV2GetLibraryAgent` before adding to
builder
- Ensures agent schemas are properly populated (fixes issue where
marketplace endpoint returns empty schemas)
- **AnyOfField.tsx**: Generate handle IDs for agents by filtering out
"root" and "properties" from schema path
- **FieldTemplate.tsx**: Apply same handle ID generation logic for agent
fields
### Checklist 📋
#### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Add an agent block from marketplace and verify it renders
correctly
- [x] Connect inputs/outputs to/from an agent block and verify
connections work
- [x] Fill in form fields for an agent block and verify data persists
correctly
- [x] Verify agent blocks work in both new and existing flows
- [x] Test that non-agent blocks still work as before (regression test)
## Changes 🏗️https://github.com/user-attachments/assets/7e49ed5b-c818-4aa3-b5d6-4fa86fada7ee
When the content of Summary + Outputs + Inputs is long enough, it will
show in this new `<ScrollableTabs />` component, which auto-scrolls the
content as you click on a tab.
## Checklist 📋
### For code changes
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Run the app locally
- [x] Check the new page with scrollable tabs
## Changes 🏗️
The main goal of this PR is to improve how we display inputs used for a
given task.
Agent inputs can be of many types (text, long text, date, select, file,
etc.). Until now, we have tried to display them as text, which has not
always worked. Given we already have `<RunAgentInputs />`, which uses
form elements to display the inputs ( _prefilled with data_ ), most of
the time it will look better and less buggy than text.
### Before
<img width="800" height="614" alt="Screenshot 2025-12-14 at 17 45 44"
src="https://github.com/user-attachments/assets/3d851adf-9638-46c1-adfa-b5e68dc78bb0"
/>
### After
<img width="800" height="708" alt="Screenshot 2025-12-14 at 17 45 21"
src="https://github.com/user-attachments/assets/367f32b4-2c30-4368-8d63-4cad06e32437"
/>
### Other improvements
- 🗑️ Removed `<EditInputsModal />`
- it is not used given the API does not support editing inputs for a
schedule yt
- Made `<InformationTooltip />` icon size customisable
### Checklist 📋
#### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Run the app locally
- [x] Check the new view tasks use the form elements instead of text to
display inputs
## Changes 🏗️
We have the setup for light/dark mode support ( Tailwind + `next-themes`
), but not the capacity yet from contributions to make the app dark-mode
ready. First, we need to make it look good in light mode 😆
This disables `dark:` mode classes on the code, to prevent the app
looking oopsie when the user is seeing it with a browser with dark mode
preference:
### Before these changes
<img width="800" height="739" alt="Screenshot 2025-12-14 at 17 09 25"
src="https://github.com/user-attachments/assets/76333e03-930a-40b6-b91e-47ee01bf2c00"
/>
### After
<img width="800" height="722" alt="Screenshot 2025-12-14 at 16 55 46"
src="https://github.com/user-attachments/assets/34d85359-c68f-474c-8c66-2bebf28f923e"
/>
## Checklist 📋
### For code changes
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Run the app on a browser with dark mode preference
- [x] It still looks in light mode without broken styles
Replaces AddModelForm and AddProviderForm components with AddModelModal and AddProviderModal, converting the add model/provider flows to use modal dialogs instead of inline forms. Updates LlmRegistryDashboard to use the new modal components and removes dropdown/form selection logic for a cleaner UI.
Moved LLM registry backend code into a dedicated llm_registry module with submodules for model types, notifications, schema utilities, and registry logic. Updated all backend imports to use the new structure. On the frontend, redesigned the admin LLM registry page with a dashboard layout, modularized data fetching, and improved forms for adding/editing providers and models. Updated UI components for better usability and maintainability.
- Resolves#11599
### Changes 🏗️
- Manually update item counts when initiating a task from `EmptyTasks`
view
- Other improvements made while debugging
### Checklist 📋
#### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] `NewAgentLibraryView` transitions to full layout when a first task
is created
- [x] `NewAgentLibraryView` transitions to full layout when a first
trigger is set up
Preserve user searches in the new builder and cache search results for
more efficiency.
Search is saved, so the user can see their previous searches.
### Changes 🏗️
- Add `BuilderSearch` column&migration to save user search (with all
filters)
- Builder `db.py` now caches all search results using `@cached` and
returns paginated results, so following pages are returned much quicker
- Score and sort results
- Update models&routes
- Update frontend, so it works properly with modified endpoints
- Frontend: store `serachId` and use it for subsequent searches, so we
don't save partial searches (e.g. "b", "bl", ..., "block"). Search id is
reset when user clears the search field.
- Add clickable chips to the Suggestions builder tab
- Add `HorizontalScroll` component (chips use it)
### Checklist 📋
#### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Search works and is cached
- [x] Search sorts results
- [x] Searches are preserved properly
---------
Co-authored-by: Reinier van der Leer <pwuts@agpt.co>
Single dollar signs ($10, $variable) are commonly used in content and
were being incorrectly interpreted as inline LaTeX math delimiters. This
change disables that behavior while keeping double dollar sign ($$...$$)
math blocks working.
## Changes 🏗️
• Configure remarkMath plugin with singleDollarTextMath: false in
MarkdownRenderer.tsx
• Double dollar sign display math ($$...$$) continues to work as
expected
• Single dollar signs are no longer interpreted as inline math
delimiters
## Checklist 📋
For code changes:
-[x] I have clearly listed my changes in the PR description
-[x] I have made a test plan
-[x] I have tested my changes according to the test plan:
-[x] Verify content with dollar amounts (e.g., “$100”) renders as plain
text
-[x] Verify double dollar sign math blocks ($$x^2$$) still render as
LaTeX
-[x] Verify other markdown features (code blocks, tables, links) still
work correctly
Co-authored-by: Claude <noreply@anthropic.com>
- Resolves#11586
- Follow-up to #11580
### Changes 🏗️
- Fix logic to include manual triggers as a possibility
- Fix input render logic to use trigger setup schema if applicable
- Fix rendering payload input for externally triggered runs
- Amend `RunAgentModal` to load preset inputs+credentials if selected
- Amend `SelectedTemplateView` to use modified input for run (if
applicable)
- Hide non-applicable buttons in `SelectedRunView` for externally
triggered runs
- Implement auto-navigation to `SelectedTriggerView` on trigger setup
### Checklist 📋
#### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Can set up manual triggers
- [x] Navigates to trigger view after setup
- [x] Can set up automatic triggers
- [x] Can create templates from runs
- [x] Can run templates
- [x] Can run templates with modified input
I want to be able to automate some actions on social media or our
sevrver in response to actions from discord
<!-- Clearly explain the need for these changes: -->
### Changes 🏗️
Add trigger blocks for common GitHub events to enable OSS automation:
- GithubReleaseTriggerBlock: Trigger on release events (published, etc.)
- GithubStarTriggerBlock: Trigger on star events for milestone
celebrations
- GithubIssuesTriggerBlock: Trigger on issue events for
triage/notifications
- GithubDiscussionTriggerBlock: Trigger on discussion events for Q&A
sync
<!-- Concisely describe all of the changes made in this pull request:
-->
### Checklist 📋
#### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
<!-- Put your test plan here: -->
- [x] Test Stars
- [x] Test Discussions
- [x] Test Issues
- [x] Test Release
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
<!-- Clearly explain the need for these changes: -->
In the File Widget, the upload button was incorrectly behaving like a
submit button. When users clicked it, the rjsf library immediately
triggered form validation and displayed validation errors, even though
the user was only trying to upload a file.
This happened because HTML buttons inside a form default to
`type="submit"`, which triggers form submission on click. By explicitly
setting `type="button"` on all file-related buttons, we prevent them
from submitting the form while still allowing them to trigger the file
input dialog.
### Changes 🏗️
<!-- Concisely describe all of the changes made in this pull request:
-->
- Added `type="button"` attribute to the clear button in the compact
variant
- Added `type="button"` attribute to the upload button in the compact
variant
- Added `type="button"` attribute to the "Browse File" button in the
default variant
This ensures that clicking any of these buttons only triggers the
intended file selection/upload action without causing unwanted form
validation or submission.
### Checklist 📋
#### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Tested clicking the upload button in a form with File Widget -
form should not submit or show validation errors
- [x] Tested clicking the clear button - should clear the file without
triggering form validation
- [x] Tested clicking the "Browse File" button - should open file dialog
without triggering form validation
- [x] Verified file upload functionality still works correctly after
selecting a file
<!-- Clearly explain the need for these changes: -->
When opening a graph from the Library, the `flowVersion` query parameter
was not being set in the URL. This caused issues when the graph data
didn't contain an internal `graphVersion`, resulting in the builder
failing and graphs breaking when running.
The `useGetV1GetSpecificGraph` hook relies on the `flowVersion` query
parameter to fetch the correct graph version. Without it being properly
set in the URL, the graph loading logic fails when the version
information is missing from the graph data itself.
### Changes 🏗️
- Added `setQueryStates` to the `useQueryStates` hook return value in
`useFlow.ts`
- Added logic to sync `flowVersion` to the URL query parameters when a
graph is loaded
- When `graph.version` is available, it now updates the `flowVersion`
query parameter in the URL (defaults to `1` if version is undefined)
This ensures the URL stays in sync with the loaded graph's version,
preventing builder failures and execution issues.
### Checklist 📋
#### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Open a graph from the Library that doesn't have a version in the
URL
- [x] Verify that `flowVersion` is automatically added to the URL query
parameters
- [x] Verify that the graph loads correctly and can be executed
- [x] Verify that graphs with existing `flowVersion` in URL continue to
work correctly
- [x] Verify that graphs opened from Library with version information
sync correctly
## Changes 🏗️
### Adjust layout and styles on mobile 📱
<img width="448" height="843" alt="Screenshot 2025-12-09 at 22 53 14"
src="https://github.com/user-attachments/assets/159bdf4f-e6b2-42f5-8fdf-25f8a62c62d1"
/>
### Make the sidebar cards have contextual actions
<img width="486" height="243" alt="Screenshot 2025-12-09 at 22 53 27"
src="https://github.com/user-attachments/assets/2f530168-3217-47c4-b08d-feccbb9e9152"
/>
Depending on the card type, different type of actions are shown...
### Make buttons in "About agent" card do something
<img width="344" height="346" alt="Screenshot 2025-12-09 at 22 54 01"
src="https://github.com/user-attachments/assets/47181f80-1f68-4ef1-aecc-bbadc7cc9c44"
/>
### Other
- Hide `Schedule` button for agents with trigger run type
- Adjust secondary button background colour...
- Make drawer content scrollable on mobile
## Checklist 📋
#### For code changes:
- [x] I have clearly listed my changes in the PR description
- [x] I have made a test plan
- [x] I have tested my changes according to the test plan:
- [x] Run locally and test the above
Co-authored-by: Abhimanyu Yadav <122007096+Abhi1992002@users.noreply.github.com>
