Back Home

GitHub Repo

Open WebUI 0.13.3 Supports Scheduled Coding Agents Bound to a Terminal and Working Directory

Open WebUI released version 0.13.3 on November 8, allowing automated schedules to use a coding agent in a specified terminal. The update adds pre-run checks for environment and agent availability, though the official documentation still reflects an earlier version.

NASA · Public domain · Image source
zh-Hant

Open WebUI released version 0.13.3 on November 8, adding coding agents to automated schedules. In the automation editor, users can select a terminal and working directory, then choose an agent available in that environment. When a schedule starts, it uses the saved settings and sends the result to a chat. Standard models can still send results to channels, while agent schedules are limited to chat destinations. Release announcement

The update lets scheduled prompts run in an explicitly configured environment. According to the official architecture, Open WebUI manages conversations, tools, and permissions, while Open Terminal provides an environment for files, command-line access, and process execution. It can run in a Docker container or directly on the host. For tasks such as periodically checking code or running tests, specifying the environment and working directory is necessary to reproduce results. This describes the feature’s engineering use; it is not evidence of measured results. Open Terminal documentation

The code shows that the new version calls a validation function when schedules are created, edited, and run. It returns an error if an agent has no terminal selected, the connection cannot be resolved, the specified agent is not in the available list, or the destination is set to a channel. A new agent discovery endpoint also takes the automation ID and working directory so the list is retrieved for the schedule’s environment. Implementation diff

Deployers should still check that the documentation is up to date: the current automation guide says the editor has no terminal settings and must inherit them from model settings, which conflicts with the new release announcement. After upgrading, engineers should verify the interface, then run a manual check of the directory, agent sign-in status, and output location before enabling a recurring schedule. Available sources confirm that the feature has been released, but provide no reliability data for long-running schedules. Pre-run checks also cannot guarantee that subsequent commands or tests will succeed. Automation guide

Sources

  1. Open WebUI v0.13.0 release
  2. Open Terminal
  3. Updating Open WebUI
  4. Open WebUI v0.13.3 release
  5. 排程代理與終端驗證實作
  6. Automations