software by Stafford Williams https://staffordwilliams.com/ Wed, 08 Jul 2026 02:35:43 GMT Wed, 08 Jul 2026 02:35:43 GMT Astro Two backlog features led the day. The bigger one was sub-tasks: backlog items can now carry sub-tasks that get attached and checked for completion (`addSubtask`, `attachSubtasks`, `checkSubtasksComplete`), with a `/subtask` command to add them and the review prompt aware of them so an item won't sail through with unfinished sub-tasks. The second was a `backlog move-phase` command for reordering phases on an item — `movePhase` plus a `reorderPhaseRows` helper to keep the phase ordering consistent after the shuffle, which meant carving the phase-update commands out into their own `registerUpdatePhaseCommand` module. On the review side I added Review-and-post entries to the review menus, so I can kick off a review that also posts its comments in one action rather than two, backed by a small `reviewModeArgs` mapping. I also disambiguated the review completion gate in `buildReviewPrompt` — the phrasing around when an item is genuinely done versus needing more work was still ambiguous enough to trip a session up. The rest were fixes across the stack. `backlog add` now requires a git remote (via `ensureRemoteOrigin`) so items can't be created in a repo with nowhere to push. Sessions whose launch actually failed are no longer auto-dismissed, so I can see what went wrong. macOS notifications stopped hanging their caller — the notify call wasn't returning cleanly. And a Windows-proxy fix strips stray replay-query escape sequences out of the scrollback so resumed sessions don't render with junk in them. Tue, 07 Jul 2026 00:00:00 GMT https://staffordwilliams.com/devlog/assist/v0.353/ https://staffordwilliams.com/devlog/assist/v0.353/ assist claude-code The main build today was a proper `branch` command. I kept starting work on a stale local base by accident, so `assist branch` now cuts a new branch off the *fresh remote default* — it resolves the default branch, fetches it, and builds a sanitised branch name from a slug (`resolveDefaultBranch`, `buildBranchName`, `validateSlug`). To make it the path of least resistance I added a matching `/branch` slash command and a `cli-hook` deny that intercepts raw `git checkout -b` / `git branch` and points at `assist branch` instead, the same pattern I used earlier for `git commit`. Branch config also shows up in `--help` now so the naming rules are discoverable. On the web side I grouped session cards by repo. When I've got sessions running across several projects the flat list gets noisy, so `groupSessionsByRepo` buckets them and a `SessionGroups`/`SessionGroupSection` pair renders each repo under its own heading. There's a fair bit of test coverage behind the grouping since the ordering and single-repo edge cases matter for it not feeling jumpy. Rounding out the day: a new `/update-jira` command for posting a concise session summary onto a ticket, extending the comment policy so the redundant-comment guard also covers YAML files (`extractYamlComments`, `decideCommentGuard`), and a settings-sync fix so that syncing preserves keys that only exist in my user settings rather than clobbering them with the shared set. Mon, 06 Jul 2026 00:00:00 GMT https://staffordwilliams.com/devlog/assist/v0.350/ https://staffordwilliams.com/devlog/assist/v0.350/ assist claude-code git A single focused feature today: repo summary chips across the top of the backlog. I wanted an at-a-glance count of what's outstanding per repo without scrolling the whole list, so `loadRepoSummaries` aggregates item counts by repo, `getBacklogSummary` serves them, and a `RepoSummaryChips` component renders them above the `ItemList`, fed by a `useRepoSummaries` hook and `fetchRepoSummaries`. While I was in the request layer I pulled CSS serving out into its own `createCssHandler`, which trimmed a bunch of inline branching out of `handleRequest` and made room for the new endpoint without the handler getting messier than it already was. Sun, 05 Jul 2026 00:00:00 GMT https://staffordwilliams.com/devlog/assist/v0.346/ https://staffordwilliams.com/devlog/assist/v0.346/ assist claude-code react The headline was cost visibility per backlog phase. I started recording per-phase token usage and active time, so each phase of an item now carries how many tokens it burned and how long it was actively running. That reshaped a good chunk of the backlog data layer — `loadRelations` and `rowToItem` were pulling everything through one tangled query, so I split the relation lookups into a dedicated `relationQueries` module and added the token/time fields to the item `types`. On the web side it surfaces as a `PhaseUsageLine` under each `PhaseCard` and an `ItemUsageTotal` rolling it up for the whole item, with `formatTokens` and `formatActiveTime` doing the human-readable formatting. With that data flowing, I used it to seed the session timer. When a run reuses a session, `seedRunningMsFromUsage` and `loadPhaseActiveMs` pull the accumulated active time out of the usage records so the timer picks up where the phase left off rather than resetting to zero — the elapsed time now reflects real work done across resumes, not just the current sitting. Smaller but handy alongside it: a new `assist backlog set-status` command for setting an item's status directly, which I'd been wanting whenever an item ended up in the wrong state and I just wanted to nudge it without going through a phase transition. The rest of this release cycle was about making sessions behave the way I expect. The one I'd been circling longest: stopping a session from continuing past a point where it's asked me something and I'm not there to answer. I first wrote it as a hard rule in `CLAUDE.md` — never continue after "No response after 60s — continued without an answer" — but a prose warning is exactly the kind of thing a session can rationalise its way past. So I replaced it with the real fix: setting `CLAUDE_AFK_TIMEOUT_MS` to max int in `settings.json`, which just stops the AFK timeout from ever firing. If a session asks a question, it waits for me, full stop. Two more status fixes rounded it out. The review button on a session card was opening the wrong PR, so I pulled the target-PR resolution into a `reviewTargetPr` helper (with tests) and tightened `prStatus` and `createCachedGhJson` so the button links to the PR that actually belongs to the session's branch. And resuming a waiting session and typing a new prompt should flip the card to running immediately — it was staying waiting until some later signal caught up — so I added a narrow `watchPromptSubmit` watcher hooked into `writeToSession`: if a restored session is waiting and the keystrokes include an Enter, it flips straight to running, without second-guessing status anywhere else. Thu, 02 Jul 2026 00:00:00 GMT https://staffordwilliams.com/devlog/assist/v0.344/ https://staffordwilliams.com/devlog/assist/v0.344/ assist claude-code Continuing the run of PR review work, I made the review synthesis viewable straight from the web UI. A `ViewReviewButton` on the card opens a `ReviewSynthesisDialog` that pulls the synthesis for the current branch — `findSynthesisForBranch` locates it, `getReviewSynthesis` serves it, and a `useReviewSynthesis` hook feeds the dialog. So once a review session has run I can read its summary in place rather than digging for the file. I also consolidated the transcript commands. The old `transcript-format` and `transcript-summarise` slash commands collapsed into a single `transcripts` workflow that formats and summarises end to end, with the internal `processVttFile` reshuffled into clearer `convert`/`parseVtt` pieces and the allow lists and README updated to match. Small follow-up: transcript move output paths are now labelled so it's obvious where files landed. On the backlog side, new bugs automatically get a Fix phase added when created, which saves a manual step every time I file one. The rest was session state correctness, which keeps being the gift that gives. Idle sessions now restore as waiting rather than running, which touched a fair bit of the restore path — `restoreSession`, `restoreInteractiveSession`, `runningSession`, and the backlog `resumeNudge`. Session titles now survive a restore instead of being lost, draft-to-run reuse resets stale state before starting, and review session chips and status resolve correctly through `postAndMaybeSubmit`. Capped off with a tiny CSS fix so markdown lists actually indent in the web UI. Wed, 01 Jul 2026 00:00:00 GMT https://staffordwilliams.com/devlog/assist/v0.342/ https://staffordwilliams.com/devlog/assist/v0.342/ assist claude-code react The usage history view got proper pagination today. It was rendering everything at once, so I taught `listUsagePeaks` to page at the query level and added a `useUsageHistoryPage` hook plus a reusable `PageShell` to drive it, which let `ListPage` and `UsageHistoryView` shed a chunk of one-off layout code. Following on from yesterday's PR review work, I added an address-comments review mode — a `ReviewDropdownContent` that lets me kick off a session specifically to work through review comments — and swapped the open-PR button over to a proper GitHub icon. A good slice of the day was session card behaviour, which had accumulated small annoyances. Cards that were still starting up now show a spinner while their chips load, driven by a little `areChipsLoading` helper and a new `CardBody`. Clicking inside a dialog no longer selects the card underneath it — I wrapped the offending controls in a `StopClickPropagation` component — and navigating to `/usage` now correctly deselects whatever card was active via the `useActiveIdReconciler`. Terminal links open without an extra prompt now too. Two behavioural fixes stood out. Resumed sessions now open in a waiting state rather than pretending to be running, which meant threading the distinction through `SessionManager`, `resumeSession` and the message handlers (and a note in `CLAUDE.md` about it). And I moved the UI-rendered status logging into a `useReportRenderedStatus` hook so what the dashboard actually shows gets echoed to `daemon.log` — much easier to debug status drift when the log tells me what the UI believed, not just what the daemon sent. Tue, 30 Jun 2026 00:00:00 GMT https://staffordwilliams.com/devlog/assist/v0.338/ https://staffordwilliams.com/devlog/assist/v0.338/ assist claude-code react The headline today was pulling PR review into the sessions web view so I don't have to bounce out to GitHub to look at what a session produced. A new `ReviewButton`, `ReviewDropdown` and `ReviewPrList` surface the open PRs for a repo, backed by a `prList`/`prStatus` layer and a `createCachedGhJson` helper so the `gh` calls don't hammer the API. The daemon side grew to match, with PR metadata flowing through `lifecycleHandlers`, `toSessionInfo` and the session `types`. Then came the inevitable trail of follow-ups: an open-PR button right on the session cards, hiding the review dropdown entirely when there are no open PRs, and a fix to get the review card chip and PR meta rendering correctly. Alongside that I reworked the web layout. `SessionsView` became `AppLayout` with a persistent sidebar that survives navigation instead of being tied to a single view, and `AppRoutes`/`AppSidebar` were reshuffled to suit. The history list got simpler too — I scoped it to the selected repo and deleted the now-redundant `FilterDropdown`, `ProjectFilter` and `uniqueProjects`, which felt good to see go. And because the web bundle and the daemon can drift out of sync, I added a `VersionBadge` that shows the web version whenever it mismatches the running daemon, wiring the version through the WebSocket handshake. The rest was fixes, mostly around sessions and the backlog signalling. A stale socket now triggers a single session restore instead of looping, `AskUserQuestion` correctly shows a waiting status via the hooks settings, and PTY sessions keep their working directory through an `ASSIST_SESSION` env var. On the backlog side I stopped signal files from leaking into foreign repos when phase results are resolved, and tightened the `cli-hook` deny so it catches raw `git commit` calls buried inside a larger command rather than only at the front. Mon, 29 Jun 2026 00:00:00 GMT https://staffordwilliams.com/devlog/assist/v0.334/ https://staffordwilliams.com/devlog/assist/v0.334/ assist claude-code react Most of today refined yesterday's comment gate. The first rough edge was that the gate blocked *all* comment edits, including removing them — so I relaxed it to allow trimming, meaning I can still delete or shrink existing comments freely while adding new ones stays gated. The bigger change was hardening the escape hatch: the pin that authorises an allowed comment is now delivered out-of-band (`issuePin`) rather than passed inline, so it can't be forged in the same edit it's meant to permit. With the gate carrying the policy properly, I dropped the old block-comments verify step, which had become redundant. The other feature ties backlog items to Jira. A new `assist backlog associate-jira` command — and a `/associate-jira` slash command in front of it — links a ticket to a backlog item: it validates the key shape, fetches the issue to confirm it actually exists, and stores the key, with `--clear` to remove the association. It's the connective tissue I'd been missing between my own backlog and the tickets that drive some of the work. On the daemon side, I added a protocol-version handshake so the WSL and Windows daemons agree on a protocol version up front and a mismatch is caught cleanly at connect time rather than surfacing as confusing downstream failures. Three fixes rounded out the day: a `--once` session re-watches its done signal after a restore so it still auto-dismisses, the daemon no longer replays its whole log buffer to a reconnecting browser (which had been dumping a wall of stale lines on every reconnect), and the web UI now shows unmerged files in the git status. A couple more fixes landed later in the cycle, both about sessions behaving as I expect. A session would go idle after a tool ran even though the agent was still working, because I only flipped the status to running on `UserPromptSubmit` and `PreToolUse` — adding a `PostToolUse` hook in `ensureHooksSettings` closes that gap so the status follows the agent when a tool completes and it resumes. And the review phase was getting too eager about rewinding: the prompt in `buildReviewPrompt` told the agent to rewind whenever a criterion still failed, which meant it would sometimes throw away already-approved, committed work. I replaced that with explicit precedence rules — approval plus a successful commit means the work is done, diverged-but-approved implementations still complete by default, and rewinding is now a confirm-first exception that must ask me before running `assist backlog rewind`. Never silently. Fri, 26 Jun 2026 00:00:00 GMT https://staffordwilliams.com/devlog/assist/v0.329/ https://staffordwilliams.com/devlog/assist/v0.329/ assist claude-code jira The theme I'm most pleased with today is enforcing the self-documenting code policy at the hook level rather than hoping I remember it. A new comment edit gate is a PreToolUse hook (`decideCommentGuard`) that blocks `Edit`/`Write` from adding code comments, with an escape hatch: when a comment genuinely earns its place I can confirm it through `codeComment` and have it allowed. I also hardened the block-comments verify gate so the check can't be quietly sidestepped. Alongside it, the maintainability gate got a proper override: a special marker comment exempts a file that legitimately can't hit the index, and the `edit-hook` protects the marker itself — Edit/Write can't add, change, or remove one — so the exemption is a deliberate act rather than something a refactor can flip. The complexity output now surfaces which files are running under an override. The other substantial piece was observability. Daemon logging is now unified into a single `assist.log` stream instead of scattered outputs, daemon session operations are logged as they happen, and I dropped a redundant lifecycle logger that was just adding noise. To go with it, a new `assist daemon drain` command clears all sessions from the local daemon for a clean slate, and a Windows-proxy create refusal now gets logged so I can see why a session didn't start. The rest was hardening, much of it on the Windows path now that it carries real traffic. I capped the Windows connect probe at 2s and tightened the create timeout to 5s so a stuck host fails fast, surfaced a failed Windows create to the UI instead of leaving it silent, gated the Windows proxy off when assist is itself running on the Windows host, and stopped an unbounded `w-` growth in Windows session ids. A few general fixes too: a hard cap on session fan-out so a runaway can't spawn endlessly, the web server now binds before the daemon is ready so the page is reachable during startup, the session-card prompt is clamped to five lines, and the 7-day usage row gets a tinted background to set it apart. Thu, 25 Jun 2026 00:00:00 GMT https://staffordwilliams.com/devlog/assist/v0.326/ https://staffordwilliams.com/devlog/assist/v0.326/ assist claude-code windows The macOS webview window had a subtle but irritating problem: because it runs from a bare binary that tao promotes to a foreground app at startup, it had no main menu — and a regular macOS app with no main menu leaves the left side of the system menu bar dead, so the Apple menu and the app menu didn't respond to clicks. Worse, without an Edit menu the standard Cmd+X/C/V/A shortcuts didn't work inside the page either. The fix installs a minimal main menu when `NSApplication` finishes launching (the same `StartCause::Init` moment I already use to set the dock icon, since anything done earlier is discarded by the promotion). It's just two menus: an app menu with Quit, and an Edit menu wiring Cut/Copy/Paste/Select All to their standard selectors — which both revives the menu bar and restores clipboard shortcuts in the webview. This pulled in the `NSMenu`/`NSMenuItem` features from `objc2-app-kit`. Wed, 24 Jun 2026 00:00:00 GMT https://staffordwilliams.com/devlog/project-switch/v0.20.1/ https://staffordwilliams.com/devlog/project-switch/v0.20.1/ project-switch rust macos