### Background
We don't have an ordering guarantee on the node execution.
Let's say we have a node that has to execute different data A, B, and C.
The current implementation limits the execution to 1 execution at a time, but there is no guarantee that A, B, and C will be executed in order.
The initial implementation did not have any restrictions, so it used to be A, B, and C executed in parallel
In the current implementation with the per-node constraint, it's A, B, C are executed serially but with no guarantee of ordering.
The scope of this PR is to guarantee that order.
### Changes 🏗️
Guaranteeing the execution per node ordering by avoiding any re-enqueue mechanism. If there are two executions run in the same node, the first one will be executed and the other will block. The blocking mechanism is indeed sub-optimal, the performance improvement can be done later (a follow-up issue will be added).
* feat(builder): checkbox for tos on login page
* feat(builder): submit agent page
DOES NOT WORK
* feat(builder): basic upload (not working)
* feat(builder): submit page more working but still not
* fix(builder): working categories, not dynamic
* feat(builder, server): enable submissions (auth error)
* fix(lint): linting
* feat(builder): submit page terms of service
* fix(builder): update lockfile
* lint(builder): lint marketplace files
- Add static link/connection support on the frontend and display them as dashed lines
- Remove queueing for static connections - there'll always be only one bead waiting at the end
- Make beads slightly larger and further from the end arrow
* Add Advanced Chatbot with History using Discord
* Update Discord Chatbot with History_v145.json
update is_active to false and is_template to true
---------
Co-authored-by: Bently <tomnoon9@gmail.com>
The execution graph is supposed to be typed, but there are cases where generic types like Any were used, and there are cases, where incompatible data passed into the wrong type.
If such a thing happens on runtime, we should do the best-effort conversion instead of breaking the run. E.g.: try to json-stringify the object to str input, or try to parse number in the string to int input, etc.
* Make `userId` required on DB entities `AgentGraph`, `AgentGraphExecution`, and `AgentGraphExecutionSchedule`
* Add SQLite and Postgres migrations to make `userId` required and set `userId` to `3e53486c-cf57-477e-ba2a-cb02dc828e1a` on existing entries without `userId`
* Amend `create_graph` endpoint and `.data.graph`, `.data.execution` methods to handle required `user_id`
* Add `.data.user.DEFAULT_USER_ID` constant to replace hardcoded literals
* refactor(builder): Migrate `FlowEditor` to use ReactFlow's state management system
We have been keeping two copies of node and edge data: one inside ReactFlow and one outside.
It works, but it's accidental and implicit and there is no reason to be using shadow copies rather than a single data source.
- Replace `useNodesState` and `useEdgesState` with `useReactFlow` hook
- Use `addNodes`, `addEdges`, and `deleteElements` where appropriate instead of `setNodes`/`setEdges` to allow use of event hooks
- Consolidate all edge -> node state sync logic into `onEdgesChange` event handler
This replaces `updateNodesOnEdgeChange`, part of `onConnect`, and `onEdgesDelete`.
- Move node deletion logic from `CustomNode` to `FlowEditor:onNodesChange`
* fix(builder): Refactor and fix copy-paste mechanism
- Rename variables for readability
- Use an ID map to correctly set the source and target IDs for the pasted edges
- Move `monitor/page.tsx` to `page.tsx`
- Remove redirect from `/` to `/build`
- Set temporary redirect from `/monitor` to `/` to prevent breaking UX (may be removed after a grace period, e.g. 2024-09-01)
* adding auth to store
* Add ability to submit agents and review them before being added to the market
* Added auth decorator
* Added auth to market api client
* fix(builder): Fix drag-select behavior on `NodeKeyValueInput`
* fix(github): Added in fallback variables for postgres testing (#7715)
Co-authored-by: Leslie Cruz <lelcruz@users.noreply.github.com>
Co-authored-by: Nicholas Tindle <nicholas.tindle@agpt.co>
* feat(builder): basic tally feedback form (#7725)
* removed database changes
* moved auth to libs project
* fixed formatting
* cleaned up auth
* Added tests and database migration
* delete decorator
* feat(builder): Add new icons and integrate with ControlPanel
Introduce SVG icons and replace ControlPanel dependencies with new icons. The ControlPanel component now uses the new icon components for improved consistency.
* feat(builder): add additional icon and update icons in NavBar
Introduced the new icon with relevant documentation and examples. Replaced existing icon imports with the newly defined icons in the NavBar component for consistency.
* feat(builder): Add icon for megaphone and replace usage
Introduced the IconMegaphone for reuse across the application. Updated TallyPopup to utilize IconMegaphone instead of the previous Megaphone icon from lucide-react.
* fix(builder): Running prettier to format changed files.
Adjusted various files for consistent code formatting. Ensured proper spacing and alignment of imports, JSX tags, and object properties to enhance readability and maintain coding standards. No functional changes made.
- Amend `AutoGPTServerAPI.sendWebSocketMessage(..)` to automatically (re)connect the websocket if disconnected
- Amend `AutoGPTServerAPI.connectWebSocket()` to prevent race conditions
* adding auth to store
* Added auth decorator
* Added auth to market api client
* removed database changes
* moved auth to libs project
* fixed formatting
* Switched to using fastapi dependencies
* Return a user object
* removed logging of the token
* Added tests
