[fork] downstream changes - #2
Closed
patroza wants to merge 52 commits into
Closed
Conversation
omegent-appBotforce-pushed
the
fork/changes
branch
from
July 24, 2026 20:32
a872da5 to
a671fa2Comparepatrozaforce-pushed
the
fork/tim
branch
2 times, most recently
from
July 25, 2026 05:23
c6b394a to
83d75d0Comparepatrozaforce-pushed
the
fork/changes
branch
2 times, most recently
from
July 25, 2026 08:30
2a67c78 to
27a72d4Comparepatrozaforce-pushed
the
fork/changes
branch
4 times, most recently
from
July 26, 2026 08:56
79176b5 to
d526bc4Compare#44) Source: pingdotgg#4506 Source SHA: f7eaa00 Imported unchanged as one candidate provenance commit.
…tgg#4558) Imported from https://github.com/pingdotgg/t3code/pull/4558\n\nAdapted to retain our provider restart-recovery constants while replacing the local default-title check with the shared policy.
…gdotgg#4379) (#312) Imported from pingdotgg#4379 at a27510d060645809ae1472bba4dbb248dc624e25. Open file previews revalidate on mount and subscribe to debounced native filesystem watches so external edits (editors, git, agents) show without a manual refresh. Co-authored-by: omegent-app[bot] <306514130+omegent-app[bot]@users.noreply.github.com> Co-authored-by: Enrico Polanski <16064771+enricopolanski@users.noreply.github.com> Co-authored-by: Patrick Roza <42661+patroza@users.noreply.github.com>
… (#328) Imported from pingdotgg#5344 at source SHA 783fd02 (commits b623dc2 + 783fd02 squashed into one provenance commit). Imported behavior: - `reduceThreadStreamItems`, a pure reducer that folds a batch of thread stream items into one state and one persistable snapshot. - `Stream.groupedWithin(64, 16ms)` on the live subscription so a burst of thread events publishes the `SubscriptionRef` once instead of per event, and web/mobile stop rebuilding large thread views per streamed event. - `eventBatchSize` on `EnvironmentThreadStateOptions`, plus the upstream regression tests for ordered single-publication bursts and for persisting a settled snapshot when a batch ends with a non-persistable turn start. Local adaptations: - Kept our `httpSnapshotLoadAttempted` guard around the HTTP snapshot fallback; the call now goes through `applyItems([...])`. - Restored `setDeleted` (removed upstream) for the terminal `thread-deleted` subscription failure, which never reaches the item stream and so cannot go through the batch reducer. Cache removal is shared with the reducer path via `removeCachedThread`. Excluded: - `tasks/todo.md`, the author's scratch checklist. Follow-up (fork/changes, not this layer): our `reload-required` branch and `reloadFromServer` are built on the deleted `setThread`, so rebasing fork/changes onto this layer must re-express them against the reducer (split the batch at the reload point, then re-enter `applyItems` with the remainder). Co-authored-by: T3 Code PR Stack <41898282+github-actions[bot]@users.noreply.github.com>
Fork platform layer in one commit: agent/stack docs, pr-stack and overlay ownership, restack/compose/fork-stack scripts, Fork CI and compose workflows, and root workspace/lockfile tooling.
Shared schemas for orchestration, RPC, settings, and related protocol surfaces used by server and clients.
Runtime helpers, product-family helpers, and oxlint plugin rules shared across the monorepo.
Client runtime state, connection diagnostics, and list helpers shared by web and mobile.
Supporting packages that diverge from upstream for the fork product.
Server product layer: work-item bridges, queue/recovery, VCS drivers, single-source RPC authorization scopes, worktree bootstrap, and related tests. Includes thread.context.compact contracts wiring on the server.
Web UI product deltas including lists, composer, usage gauges, file/content search, and thread-loading send guards.
Mobile shell product, list modes, composer queue API, and scroll behavior after rolling back experimental feed layout work.
Desktop packaging scripts/config and backend ownership behavior for the fork desktop shell.
Make the product default for draft threads match the preferred worktree workflow: New worktree with Start from origin enabled. Mobile falls back to the shared server defaults when config is still loading. Co-authored-by: omegent-app[bot] <306514130+omegent-app[bot]@users.noreply.github.com> Co-authored-by: Patrick Roza <42661+patroza@users.noreply.github.com>
* feat(server): reap idle dev stacks from a level-triggered sweep Dev stacks that repositories start outside T3's process tree have been surviving for hours after the work that needed them finished. In the case that prompted this, ten stacks stayed up for thirteen hours across two already-merged PRs, holding roughly 15 GB of RSS. Teardown was never broken; nothing invoked it. Both existing triggers are edge-triggered. runOnWorktreeRemove needs a worktree removal, and merging a PR does not remove one. runOnPrMerged fires from VcsStatusBroadcaster only when it observes a not-merged -> merged transition for a worktree it happens to be polling, so an agent that finishes its work and a human who merges an hour later produce no observer, no edge, and no teardown. Add a sweep that looks at state rather than events, so a missed hook costs one idle window instead of an unbounded number of hours. Both project scripts stay: they are the early reap and are strictly faster when they do fire. This is the floor under them, not a replacement. The registry format (`dev-stack/1`, contracts/devStack.ts) carries facts only — pids, ports, and where each process runs — because the sweep reads every project's stacks and cannot hold per-repo knowledge. Policy is declared per checkout under `devStacks` in t3.json: idle window, consumer patterns, and the ordered entry roles. That split matters beyond tidiness: an idle window is a property of a repository rather than of one running instance, so changing it now applies to stacks that are already up, on the next sweep. Decisions worth knowing: - Opt-in. A repository without `devStacks` is left alone rather than swept under a guessed policy. - `entryRoles` is ordered and we watch the first role present. A frontend holds keep-alive connections to its own API, so also watching the API port would read a stack as busy for as long as the frontend is up, and nothing would ever look idle. An api-only stack falls through to the API by the same rule. - An absent `entryRoles` watches every port. That over-detects activity, which is the right way to be wrong: a stack lives too long instead of dying mid-run. - A stack is never reaped on the sweep that first sees it. A cold boot plus a build can outlast the window, so the clock starts from an observation. - The producer's start/stop lock is honoured, so a stack that is halfway up is skipped rather than killed. - Ownership is confirmed against /proc/<pid>/cwd before signalling, so a recycled PID is never mistaken for a stack's own process. The sweep runs as one layer-scoped fiber in the server that already runs, rather than a separate unit: the server is what starts the agents that create these stacks, so "server down" means "no new stacks", and a per-stack systemd TTL is the better answer for a hung server than a second daemon would be. Failures are logged and swallowed. Each sweep reports its duration and counts, and logs at info when it acted or ran long, so whether it burdens the main loop stays a query rather than a guess. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * chore(contracts): keep the dev-stack contract free of downstream references The producer of dev-stack/1 is a private repository; naming it here, and using its worktree paths and tenant names as test fixtures, leaks downstream detail into a contract that is meant to be repo-agnostic and is read by anyone adopting the format. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(server): green dev stack reaper PR --------- Co-authored-by: Stack Test <stack-test@example.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com> Co-authored-by: omegent-app[bot] <306514130+omegent-app[bot]@users.noreply.github.com>
* fix(vcs): treat bare repositories as valid worktree sources A repository with `core.bare=true` was reported as "not a repository" at detection, because `detectRepository` gated on `rev-parse --is-inside-work-tree` — which conflates "is a Git repository" with "has a checkout". Every Git route then failed with "Failed to resolve the VCS driver for this Git command", including `fetchRemote` and `createWorktree`, so no thread could start against that repository. Starting a thread does not need the source repository to have a checkout: the thread gets its own worktree, and `git worktree add`, `git fetch`, and ref plumbing all work against a bare repository. Only operations that touch a checkout need one. - Detect bare repositories instead of rejecting them, and carry `bare` on the repository identity. Detection now uses one combined `rev-parse --is-bare-repository --is-inside-work-tree`, which separates "no repository" / "bare" / "checkout" in a single call, so negative detection stays at one git invocation. - Let the bare-safe routes through: createWorktree, fetchRemote, resolveRemoteTrackingCommit, removeWorktree, renameBranch, and createRef when it is not also switching. - Checkout-dependent routes (switchRef, pullCurrentBranch, stacked actions, PR thread preparation) now fail with the actual reason — that the repository has no working tree — instead of a routing error. - Status polling reports a bare repository as having no workspace rather than erroring on every poll. * style: format files drifted on fork/changes Whitespace only, no behavior change. These three files were already unformatted on fork/changes and are untouched by this branch's fix, but the agent ship gate runs a whole-repo `vp check`, so publishing is blocked until they are formatted. * fix(desktop): add the missing removeCommandLineSwitch test stub `ElectronApp` gained `removeCommandLineSwitch`, but the DesktopUpdates test mock was not updated, so `@t3tools/desktop` failed typecheck on fork/changes before this branch. Untouched by this branch's fix, but the agent ship gate runs the workspace typecheck, so publishing is blocked until it compiles. * fix(vcs): supply the required worktree path in the bare-repo test input * refactor(server): use Effect FileSystem and Schema for the runtime descriptor serverRuntimeStartup wrote and cleaned up the runtime descriptor with `node:fs/promises`, `node:path`, and hand-rolled `JSON.stringify`/`JSON.parse`, which the Effect diagnostics flag (nodeBuiltinImport, preferSchemaOverJson) and which failed the workspace typecheck. - Write and remove the descriptor through Effect `FileSystem` (`mode: 0o600` is preserved) and join its path with Effect `Path`. - Encode and decode with the existing `ServerRuntimeDescriptor` schema from `@t3tools/shared/serverRuntime` rather than untyped JSON. That schema was already the one the desktop client decodes with, so the file now round-trips through a single definition instead of two hand-written shapes. - Keep prior failure semantics: a descriptor that cannot be written is a defect (as it was under `Effect.promise`), and cleanup stays best-effort so a missing, unreadable, or malformed file cannot fail shutdown. - Pin the on-disk format with a test that decodes the written contents the same way `DesktopExistingBackend` does, since that shape is a cross-process contract. * test(server): decode the descriptor via the schema JSON codec --------- Co-authored-by: T3 Code PR Stack <41898282+github-actions[bot]@users.noreply.github.com>
… of limit (#321) * fix(usage): stop extra-usage pools from marking a provider out of limit The ai-usage daemon reports Claude's extra-usage credit buffer as a `monthly` window. That buffer is only drawn from once the 5-hour and weekly plan windows are exhausted, so a drained (or switched-off) pool says nothing about whether Claude is usable right now — yet `usageMarkerForItem` treated any window at 100% as a hard block, painting the provider dot red and the tooltip "limit reached" while 5h sat at 18%. Markers, the outlook ring, and `worstUsagePercent` now only consider enforced windows: daemon-flagged `informational` pools, extra-usage / credit / on-demand buffers, and windows labelled `(off)` are context only. The tooltip renders those rows with a neutral bar instead of a red one. Money-denominated plan windows (opencode-go monthly $, cursor billing cycle) keep enforcing. Also picks up the daemon's `informational` and `limit` window fields in the contract. * test(usage): assert the tooltip stays neutral for a drained extras pool --------- Co-authored-by: T3 Code PR Stack <41898282+github-actions[bot]@users.noreply.github.com>
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 freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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.
Managed downstream layer
fork/candidatesfork/changesfork/integrationThis is the permanent downstream product layer and the normal base for feature PRs. It contains our shared/fork-owned changes on top of the Tim and upstream-candidate provenance layers.
The following implementations are deliberately excluded and remain in managed draft overlays:
Private operations, credentials, machine topology, host configuration, and deployment implementation remain outside this public repository.
This PR intentionally remains open and draft. Do not merge it;
fork/integrationis the deployable aggregate.