feat(web): start and stop managed VM instances through hooks - #1
Closed
IAmJSD wants to merge 9 commits into
Closed
Conversation
Co-authored-by: t3-code[bot] <269035359+t3-code[bot]@users.noreply.github.com>
Co-authored-by: t3-code[bot] <269035359+t3-code[bot]@users.noreply.github.com>
The settings panels read and write server settings through the primary
environment, which only resolves via a `PrimaryConnectionTarget`. The
hosted app never has one — every device is paired remotely — so the
panels degraded silently there:
- `primaryServerProvidersAtom` returned no providers, leaving the text
generation model picker with nothing to offer ("No models found"), so
the model backing thread titles, commit messages, change request
content and branch names could not be changed at all.
- `useUpdatePrimarySettings` resolved to a null environment id, and
`useUpdateSettingsTarget` no-ops on null, so every write from these
panels was dropped without an error.
- Reads fell back to `DEFAULT_SERVER_SETTINGS`, presenting the schema
default as if it were the user's configuration.
Resolve the settings target to the primary device when there is one and
the first connected device otherwise. Sessions with a primary device are
unaffected. Rename the hook pair to `useGlobalSettings` /
`useUpdateGlobalSettings`, since the module header documents the
primary-only scoping as deliberate and it no longer holds.
Left alone: the primary-scoped atoms behind the primary-device update
notification, sidebar and command palette, and the diagnostics panel,
whose RPCs target the primary environment.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Two consumers were left on the primary environment after the settings panels moved to the resolved settings environment, so on a hosted session with no primary device they disagreed with the panels that write them. - New-thread defaults (`defaultThreadEnvMode`, `newWorktreesStartFromOrigin`) read `primaryServerSettingsAtom` in `useNewThreadHandler` and `ChatView`, so the General controls saved and re-displayed while new drafts kept the schema defaults. Both now read `settingsServerSettingsAtom` — the environment those controls write. - `SourceControlSettingsPanel` still resolved its discovery target and gated `SourceControlWritingSettingsSection` on `usePrimaryEnvironment`, so the writer-model and fetch-interval controls never mounted there. It now uses `useSettingsEnvironmentId`, matching the section's own hooks. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
VM management solutions had no way to boot or shut down the machine hosting a T3 server from inside T3 Code, so users had to open the management app before every remote session. The server's settings gain nullable startHookUrl/stopHookUrl, delivered to clients inside the cached server config. On Settings -> Connections, Connect now POSTs the start hook first, renders its component form when the endpoint asks for input, polls until the instance reports ready with a 204, then runs the normal connect flow. Connected environments with a stop hook show a Stop button backed by a new server.runStopHook RPC; the server DELETEs the endpoint and clears the setting on a 404 so clients drop the control. Done by Claude Fable 5 on Claude Code. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
Bugbot is not enabled for your account, so this pull request was not reviewed. Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs. |
github-actions
Bot
added
vouch:trusted
Member
Author
|
Closing — reopening this against the upstream repo pingdotgg/t3code. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
VM management solutions have no way to boot or shut down the machine hosting a T3 server from inside T3 Code — users must open the management app before and after every remote session.
Solution
Instance lifecycle hooks, configured by the management solution in the server's settings (
startHookUrl/stopHookUrl, nullable, null by default). Clients receive them insideServerConfig.settings, which is already cached per environment, so the start hook URL stays known while the instance is off.Start hook (client-side, on Settings → Connections where the Connect button lives): clicking Connect first POSTs the hook URL with no content.
200with{ "poll_url": ..., "retry_secs": ... }→ poll until a204says the instance is up, then run the normal connect flow.400with a component form → a dialog renders the components (informational text, selects, regex-validated text inputs), then POSTs the resolved values back as a JSON array; re-prompts are handled.Stop hook (server-side): connected environments whose settings carry a stop hook show a Stop button backed by the new
server.runStopHookRPC (orchestration:operate). The server DELETEs the endpoint:204reports the instance as stopping;404clears the setting, which streams to clients and removes the button. The saved environment keeps its pairing and cached config, so the next Connect starts it again.Protocol reference for management-solution authors:
docs/internals/instance-hooks.md. User docs indocs/user/remote-access.md.Notes
Testing
vp test run apps/server/src/instanceHooks.test.ts apps/web/src/components/settings/startHook.test.ts— 14 tests covering the DELETE flow (204/404/500/unconfigured) and the client protocol (poll state, form parsing, immediate 204, JSON array submission, polling loop, regex validation).Done by Claude Fable 5 on Claude Code.
🤖 Generated with Claude Code