* feat(share): add public share links via GitHub Gists
Adds a Share button to the chat panel header that publishes the
current session as a secret GitHub Gist on the user's account. The
share link routes through gistpreview.github.io to render share.html
as a styled kandev-themed page; snapshot.json + README.md ride along
for machine-readable + fallback views.
Snapshots run a redaction pass (absolute paths, common credential
shapes, .env contents, shell env vars) and strip <kandev-system>
prompt blocks via the shared sysprompt helper. Message filtering
whitelists user prose + tool_* types so internal noise never leaks
into shares.
Spec at docs/specs/public-share-links/spec.md.
* fix(share): coerce pendingClarification to bool for ClarificationSection
A late refactor extracted ClarificationSection with `pendingClarification: boolean`
on its props, but the call site passes the actual Message|null value. CI's tsc
build catches the mismatch; runtime was unaffected because the section only uses
the value in a truthiness check. Coerce with Boolean() at the call site to keep
the prop's intent narrow.
* fix(share): address PR #995 review feedback
Bugs (flagged by Greptile, Cubic, Claude):
- abs-path redaction split on `root + "/"` so sibling dirs like
/home/user/proj2 don't get corrupted when workspaceRoot is
/home/user/proj; regression test added.
- GHClient.DeleteGist now wraps "HTTP 404" stderr as *GitHubAPIError so
share.IsAlreadyGone matches consistently with the PAT client; this
restores the "gist already revoked upstream" soft-success path for
CLI-authenticated users.
- SnapshotSizeBytes is now measured from the same MarshalIndent payload
the gist backend uploads, so the size shown in the UI matches the
actual gist bytes (was ~20-40% underreported).
Polish (flagged by CodeRabbit, Cubic):
- Tool name + tool output now HTML-escaped; tool call args and output
render via <pre><code> instead of triple-backtick fences so payloads
containing ``` can't break out of the code block.
- share-list.tsx: clear `busy` via finally and break error layout out
of the justify-between row.
- use-shares.ts: convert UseSharesResult to interface (repo TS rules);
reset error+isLoading when ids are missing.
- use-shares.test.tsx: wait on the actual async outcome (shares
populated, listShares called) instead of isLoading flipping false.
- provider.go: replace `repo.Close` cleanup with an explicit noop
closure so the comment ("Cleanup is a no-op") and the impl match.
- spec.md: aligned the shareability rule across What/Failure/Scenarios/
Out-of-scope — any non-prehistory state can be shared; added language
tags to fenced code blocks for markdownlint.
Skipped intentionally (reply on each thread):
- HTML-escape every metadata badge: the values are controlled enum
strings from kandev's own schema, not user input. Defense-in-depth
noise. Tool name (which can come from agent runtime) IS escaped.
- useShares request-token guard against stale responses: the dialog
owns the hook lifecycle and opens against a single (taskId,
sessionId) — the race the comment describes can't occur here.
* fix(share): redact paths for every worktree, not just the first
A single TaskSession can carry multiple worktrees when the task spans
multiple repos. The redactor previously only saw the first worktree
path, so absolute paths from secondary worktrees leaked into public
shares instead of being rewritten to repo-relative form.
NewRedactor is now variadic and applies the abs-path rule against every
root it receives, sorted longest-first so a root that is a prefix of
another (e.g. /workspace vs /workspace/repo1) gets matched by the more
specific one first. workspaceRootFor in builder.go becomes
workspaceRootsFor and emits every WorktreePath plus the WorkspacePath
fallback.
Flagged by Cubic on PR #995.
* fix(share): address PR #995 review round 3
Bugs (flagged by Cubic + Claude on gh_client.go:714):
- DeleteGist now calls the existing isNotFoundErr helper instead of an
inline strings.Contains(err, "HTTP 404"). The helper also covers
"404 Not Found" and "status: 404", which some gh versions emit, so
the share service's "gist already revoked upstream" soft-success
path now fires consistently across CLI auth users.
Bugs (flagged by Claude on builder.go:220):
- toolCallBlock used to read metadata["tool_name"] and metadata["args"]
but production never writes those keys; the typed payload lives under
metadata["normalized"] (a *streams.NormalizedPayload) and the tool
kind is on Message.Type. Switched to deriving ToolName from
strings.TrimPrefix(m.Type, "tool_") (gives "read", "execute", "edit",
…) and reading args from metadata["normalized"], with backward-compat
fallback to the older "tool_name"/"args" keys for the debug-fixture
replay path. New TestBuildSnapshot_ToolCall_ProductionMetadataShape
exercises the real session shape.
Bugs (flagged by CodeRabbit on redaction.go:110):
- Bare workspace-root occurrences (pwd output, end-of-sentence, quoted
paths) used to leak because the matcher only handled root + "/".
Switched to a precompiled regex per root —
regexp.QuoteMeta(root) + `(?:/|\b)` — so both `root/...` and standalone
`root` get stripped, while sibling dirs like /workspace2 still don't
match. Test TestRedactor_AbsPath_StandaloneRootStripped covers the
bare-root variants.
Doc (flagged by Claude on spec.md):
- Two malformed fence closers (`go and `text on closing fences) corrected
to plain ``` so the file is markdownlint-clean.
* fix(share): narrow abs-path boundary to avoid sibling-prefix corruption
Round 4 review (Claude + CodeRabbit) caught a regression in the regex
introduced by 436cb66: `(?:/|\b)` used `\b` (word boundary) to catch
bare-root cases, but `\b` fires on any word↔non-word transition. So a
sibling path like `/workspace.bak`, `/workspace-old`, or
`/workspace-other` would false-match the `/workspace` root (since
e→. and e→- are word boundaries), stripping it and leaving a
corrupted `.bak` / `-old/file.ts` in the share.
Replaced the matcher with `(/|\s|$)`: path separator, whitespace, or
end-of-string only — never punctuation or quote chars. The boundary
is captured and re-emitted via ReplaceAllStringFunc (or dropped, in
the `/` case) so prose loses only the root, not the surrounding space.
Added TestRedactor_AbsPath_SiblingsWithPunctuationNotCorrupted covering
`.bak`, `-old`, `_alt` siblings; the digit-sibling test still passes.
The quoted-bare-root case (`"/workspace"`) no longer matches and was
dropped from the standalone-root test — contrived enough that a small
leak there is preferable to corrupting realistic sibling paths.
* refactor(share): move share button into chat status bar
Replaces the dedicated PanelHeaderBarSplit above the message list with
a slot in ChatStatusBar (the existing row holding TodoIndicator and
PRStatusChip just above the chat input). The button anchors right via
ml-auto; when the "proceed to next step" affordance is present it
takes the right anchor instead and the share button stays inline.
Also tightens the iconOnly variant in ShareButton to a ghost style
that matches the muted-foreground chips in the status bar, and exports
shareableSessionStateClient from share-button.tsx so the gating logic
lives next to the entry point it controls.
Net -23 lines in task-chat-panel.tsx: the extra header bar and the
local gate helper are gone.
* docs(share): broaden tooltip + add share/ to CLAUDE.md tree
The share tooltip read "Share this completed task with a public link",
but shareableSessionStateClient allows RUNNING / IDLE / WAITING_FOR_INPUT
/ FAILED / CANCELLED too. A user hovering mid-run saw misleading copy;
reworded to "Share this task conversation as a public link" which
matches the actual gate.
Also added the missing internal/task/share/ entry to the task/ subtree
in CLAUDE.md so the new package shows up in the repo map — flagged
across multiple rounds.
* docs: add internal/task/share/ to repo map
Carryover from the share PR — AGENTS.md (which CLAUDE.md symlinks to)
was missing the share/ subpackage under task/. Adding it now so the
repo map matches what's checked in.
---------
Co-authored-by: Kandev Agent <agent@kandev.dev>
* feat(orchestrator): persist N-entry FIFO message queue
Replaces the single-slot, latest-wins, in-memory queue with a SQLite-backed
FIFO queue (default cap 10 per session, configurable via
KANDEV_QUEUE_MAX_PER_SESSION). Drain stays one-per-turn so each entry gets its
own user message and `on_turn_*` workflow cycle. Inter-task `message_task_kandev`
returns a structured `queue_full` error with sender attribution when the
target session is saturated, so the calling LLM can decide whether to retry.
Surfaces a new WS action `message.queue.remove` for single-entry deletion;
`message.queue.cancel` now means clear-all. Edit/remove are keyed on the
entry's UUID, returning `entry_not_found` when atomic-take won the drain race.
* feat(web): N-entry queue state, API, and optimistic hook
Replaces the single-slot QueueStatus with an ordered `entries[]` per session
plus capacity meta (`count`, `max`). API exposes typed `QueueFullError` and
`QueueEntryNotFoundError`; the hook surfaces optimistic edit/remove with
drain-race rollback (refetch on `entry_not_found`).
* feat(web): inline ghost-message queue UI with sender chips
Replaces the floating QueuedMessageIndicator overlay with a stack of ghost
messages rendered above the chat input. Each ghost shows a sender chip
(user / agent) and per-entry edit/remove. A small "n/max queued · clear all"
pill summarizes capacity. Inter-task entries are visually distinct and edit
is gated to the original sender.
* fix(queue): address review — auth guard, TOCTOU, error UX
- Add `messagequeue.QueuedByUser` / `QueuedByAgent` constants and centralize
the `queue_full` error code so the WS and MCP handlers stay in sync.
- WS `message.queue.{add,append,update}` default empty user_id to QueuedByUser
so the UpdateContent ownership guard always runs against a non-empty owner;
inter-task entries (`queued_by="agent"`) can no longer be overwritten by
user-initiated edits even when `user_id` is omitted from the payload.
- memoryRepository: hold the mutex through the whole append-or-insert path
(prior split released the lock between tail-check and Insert, allowing two
same-sender callers to race past coalescing). TransferSession now shifts
source positions past the destination's max so transferred entries always
drain after pre-existing destination entries, matching SQLite semantics.
- Frontend: surface edit/clear/remove failures via `sonner` toasts (drain race
→ "message already sent"); exit edit mode on failure instead of leaving the
textarea silently stuck. Hide the remove button on read-only (agent) ghosts.
- `useQueue.clearAll` no longer derives `max` from pre-clear entry count;
resets to `metaMax ?? 0` so the next status_changed event lands authoritative
capacity.
- `senderKindOf` drops the unreachable metadata fallback; route on `queued_by`
alone since `dispatchTaskMessage` always sets it.
- Test loggers now write to stderr per CLAUDE.md "stderr for logs" guideline.
* test(web): align queue e2e selectors with new ghost-message UI
The ghost-message rewrite renamed button titles ("Cancel queued message"
→ "Remove queued message", "Edit message" → "Edit queued message") and
introduced a `data-testid="queued-ghost-list"` container. The
`message-queue.spec.ts` Playwright tests still pointed at the old titles.
* fix(queue): address second review pass — auth, drain race, tests
- WS handlers (`add` / `append` / `update`) now reject `user_id="agent"`
outright, closing the gap a hostile WS client could exploit by
impersonating the inter-task sender identity to overwrite agent
entries (greptile P1).
- WS `update` / `remove` now require `session_id` so the post-mutation
`MessageQueueStatusChanged` broadcast can never be silently skipped.
- `sqliteRepository.TakeHead` checks `RowsAffected()` after DELETE;
when two concurrent take callers race past the busy-timeout the
loser sees 0 affected rows and now returns `nil, nil` instead of
the already-drained message (would otherwise dispatch the same
prompt twice).
- `Service.GetStatus` returns `Entries: []` instead of `nil` on the
error path so the wire shape stays consistent (frontend already
guards but strict consumers wouldn't tolerate the type mismatch).
- New tests: `queue-api.test.ts` covers `rethrowQueueError` mappings
(queue_full / entry_not_found / generic / Error pass-through);
`session-slice.upsert.test.ts` covers `setQueueEntries`,
`removeQueueEntry`, and `clearQueueStatus`. Backend gains an
agent-impersonation handler test.
- `rethrowQueueError` now surfaces `WSError.message` for non-queue
errors instead of stringifying the whole object.
- Cleanup: replace unused `a, c := ...` discards in service_test
with `_, _ = ...`.
* fix(queue): scope update/remove by session_id (cross-session deletion guard)
The MCP `queue_full` error payload deliberately discloses sibling-task
entry IDs (so the calling agent's LLM can reason about the wedge state).
Without a session_id filter on `DeleteByID` / `UpdateContent`, a hostile
agent that received that payload could call `wsRemoveEntry` /
`wsUpdateMessage` against another task's entries and delete or rewrite
them — only the per-entry UUID was being checked.
- `Repository.UpdateContent` and `Repository.DeleteByID` now require
`sessionID` and add `AND session_id = ?` to the SQL guard. The memory
repo mirrors the scope (lookup keyed on session before the entry scan).
- `Service.UpdateMessage` / `Service.RemoveEntry` take `sessionID`; the
WS handlers thread `req.SessionID` through (already validated as
non-empty in the previous fixup).
- TypeScript: `session_id` is now required on `updateQueuedMessage` /
`removeQueuedEntry` params so callers fail at compile time, matching
the backend contract.
- Tests cover the cross-session attack: `RemoveEntry`/`UpdateMessage`
with an attacker-supplied session_id but a real entry id from another
session must return `ErrEntryNotFound` and leave the victim entry
untouched. SQLite repo gets parallel coverage.
* fix(queue): propagate transfer error + tighten WSError narrowing
- `Service.TransferSession` now returns an error so callers can fail
closed instead of silently leaving entries orphaned on the previous
session. The two existing call sites pick the right policy:
`switchSessionForStep` (reusing an existing session) propagates the
error so the workflow switch aborts; the new-session path logs and
continues, since the new session is already committed and rolling
back here would leave a half-switched workflow.
- `asWSError` now also narrows on `message` / `details`, so WS payloads
without a `code` field are still mapped to a meaningful `Error.message`
via `rethrowQueueError` (instead of "[object Object]"). Native
`Error` instances explicitly skip the WS-error branch so they pass
through untouched, preserving the existing pass-through guarantee.
- `queue-api.test.ts`: `expect.assertions(N)` on the catch-only tests
so a missing throw fails the test instead of silently passing.
* fix(queue): check TransferSession error in service tests
The TransferSession signature change in 83abe6d7 broke errcheck on two
test sites that hadn't been updated. Wrap them in require.NoError.
* feat(backend): multi-question support for ask_user_question_kandev
Reshape the MCP tool from a single prompt+options pair to a questions[]
array (1-4 entries, single-choice each, custom text fallback). The
clarification HTTP handler gates the response until every question has an
answer, fans out one chat message per question (sharing pending_id, with
question_index/question_total in metadata), and returns a JSON map keyed
by question id back to the agent. The orchestrator resume prompt now
formats multi-line Q1/A1, Q2/A2 summaries when the bundle has more than
one question. Mock-agent gets a clarification-multi scenario and
ask_user_question_test.go covers the new schema + map response. No
back-compat shim — local-first tool, single-PR breaking change.
* feat(web): refresh clarification overlay with multi-question + modern UI
Each per-question chat message gets its own card with a "Question N of M"
progress chip, a ✓ Answered badge once committed, and shadcn-aligned
option buttons with kbd hints. A new useClarificationGroup hook coordinates
the per-question commits (decision A) and posts every answer in a single
batch once the user finishes the bundle; X rejects the whole group. The
single-question case keeps the 1-9 / Esc keyboard shortcuts. Playwright
e2e adds 6 multi-question scenarios (rendering, full happy path, partial
state, mixed option/custom, skip-rejects-bundle, sibling-state) on top of
the existing single-question regression suite.
* fix(web): clarification carousel UX with stepper navigation + race fix
Replaces the vertically stacked card layout with a horizontal stepper:
numbered circle dots up top show step state (active/answered/pending),
the single active question card sits in the middle, and Back / Next /
Submit buttons sit at the bottom. Picking an option auto-advances to the
next step (single-question case auto-submits, no explicit Submit button).
The stepper steps are clickable so users can revise prior answers.
Keyboard shortcuts: 1-9 for option pick, ArrowLeft/Right for navigation,
Esc for skip.
Also fixes a race that surfaced as a 400 BadRequest on the
/api/v1/clarification/:id/respond endpoint when the user clicked an
option before all N messages of a multi-question bundle had streamed in:
findPendingClarificationGroup now gates on metadata.question_total and
returns an empty array until the full bundle has arrived.
Playwright suite expanded from 6 to 12 multi-question scenarios:
stepper rendering, auto-advance, Back, stepper jump, Submit gating,
happy path, mixed custom-text, revise-via-jump, skip-from-mid-step,
number-key shortcuts, Esc skip, first-step Back disabled.
* fix(clarification): address CodeRabbit/Greptile/Claude review comments
Backend:
- Single source of truth for question validation: rename/export
NormalizeAndValidateQuestions and call it from both the HTTP and the
WebSocket-side MCP handlers, so duplicate question ids and bad option
counts can't slip through either path.
- httpRespond now validates that every answer references a real question
id from the bundle and that no id is answered twice (regression tests
added).
- questionsFromMessages reorders by metadata.question_index so the
reconstructed bundle preserves the original question order.
- adapters.CreateClarificationRequestMessages rolls back already-created
messages when a later write in the bundle fails.
- extractQuestionAnswers always returns a JSON map (even on rejection,
with an envelope { rejected, reject_reason }) and simplifyAnswer
preserves both selected_option and custom_text when both are present.
- MCP tool description: rewrite the response example to match the actual
output shape; add minItems / maxItems to the questions and options
arrays in the JSON Schema.
- Test helper: drop the local contains() shim, use strings.Contains.
Frontend:
- useMemo([props]) was a no-op (props is always a fresh ref); split into
[message, messages] so it actually memoizes.
- Drop the dead "expired" SubmitState branch — the backend never returns
410 from this endpoint.
- Remove the asymmetric option-list dimming: the agent now receives both
selected_option and custom_text together, so a half-grey state was
misleading.
- New tests: pending-clarification.test.ts covers
findPendingClarificationGroup including the question_total gate, and
use-clarification-group.test.ts covers the hook's record/submit/skip
flows and the override-merge path used by the single-question case.
* fix(web): treat clarification 409 as success so duplicate submit doesn't stick the overlay
* fix(clarification): unblock agent on persistence failure + nil-guard rejected branch
- Add Store.CancelRequest(pendingID) and call it from both the WebSocket
MCP handler and the HTTP create-request handler when message creation
fails. Without it, the agent's WaitForResponse parked for the full
2-hour MCP timeout while no clarification cards ever reached the user.
Both handlers now return an explicit error response on failure.
- applyAnswersToMessages now guards h.repo == nil in the rejected branch
(parity with sibling expectedQuestionIDs).
- extractQuestionAnswers stubs for unanswered questions in a rejected
bundle now carry { answered: false, rejected: true } so agents reading
per-question payloads see the rejection without inspecting the
envelope.
- Hook test: added 409-as-ok regression coverage for both submit and
skip paths so the previous fix can't drift.
- Store test: covers CancelRequest unblocking a parked WaitForResponse.
* fix: drop unused e2e helper + use channel sync in CancelRequest test
* refactor(web): drop dead 'message' prop from clarification overlay
Both call sites pass 'messages' (the bundle group), so the legacy
single-message prop was unused. Tightening the prop type also lets the
useMemo depend on a single value.
* fix(web): ignore modifier-keyed events in clarification carousel shortcuts
Bare 1..9 / ArrowLeft / ArrowRight / Esc still drive the carousel, but
Cmd/Ctrl+1..9 (browser tab switch) and Alt+ArrowLeft (browser back) now
pass through as the user expects.
* feat(backend): add session-scoped message search endpoint
Adds SearchMessages across repository/service/handler layers and a new
WebSocket action "message.search". Returns lightweight hits with a
60-char snippet around the first match. Session-scoped only: results
never leak across sessions. Uses dialect-aware LIKE/ILIKE so it works
on SQLite today and Postgres later.
* feat(web): add shared panel search bar and Ctrl+F hook
Introduces the reusable PanelSearchBar overlay and usePanelSearch hook
used by the per-panel search features. The hook listens on the window
and fires only when focus is within the panel root, preventDefaults the
browser find-in-page, and wires Esc to close. PanelRoot becomes a
forwardRef so panel components can expose their root element to the
hook. Adds SHORTCUTS.FIND_IN_PANEL plus shared highlight/flash styles.
* feat(web): add Ctrl+F search to terminal panels
Mounts the xterm addon-search on both passthrough and shell terminals,
exposing next/prev plus case-sensitivity and regex toggles. Invalid
regex surfaces an inline error without firing the search. Because xterm
consumes keystrokes before React, the Ctrl/Cmd+F intercept is added to
attachCustomKeyEventHandler so the shortcut opens the bar from inside
the terminal. Results counter comes from onDidChangeResults; decorations
clear on close.
* feat(web): add Ctrl+F search to plan panel
Adds a custom TipTap extension that builds a ProseMirror DecorationSet
over the document and exposes setPlanSearchQuery / planSearchNext /
planSearchPrev / clearPlanSearch commands. The current match gets an
extra class and scrolls into view. Match count is read from the plugin
state via transaction subscription, so edits to the document update the
counter automatically. Find-only — no replace.
* feat(web): add Ctrl+F search to session chat panel
Wires the new backend message.search action into a debounced client
hook and renders a hits dropdown under the search bar, with the query
highlighted in each snippet. Clicking a hit scrolls the message into
view and flashes it. When the target message is older than what is
currently loaded, the hook loops through loadOlder pages until the
message renders, making search work across lazy-loaded history without
needing to scroll-and-wait manually.
* fix(web): lazy-load xterm addon-search to avoid SSR crash
addon-search is published as a UMD bundle that references `self` at
module load. Statically importing it breaks the Next.js static
prerender of /. Deferring the import to an effect keeps the module out
of the SSR graph without changing runtime behaviour.
* fix(web): propagate plan editor via state so plan-search hook re-runs
The plan-search hook subscribed to editor transactions inside a
useEffect that depended on a ref object. Because ref.current changes do
not trigger effect re-runs, if the Editor instance was not yet ready
when the hook first mounted, the transaction handler never attached and
the match counter stayed at 0 / 0. Tracking the editor via state (in
addition to the ref) causes the deps to change when the Editor becomes
available, so the handler attaches and the counter updates.
* feat(web): allow overriding message list renderer via ?renderer= query
Exposes the "native" / "virtuoso" choice at runtime so E2E coverage can
exercise both paths without a redeploy or code change. Defaults to the
existing "native" strategy when no override is present.
* test(web): add Playwright coverage for panel-scoped Ctrl+F search
Adds an e2e/tests/search/ suite with five specs:
- panel-search-shared: cross-panel contract (Ctrl+F opens, autofocus,
Esc closes, close button, state reset on reopen) parametrised over
session, plan, and terminal.
- terminal-search: xterm addon-search integration — key intercept
without leaking "f" to the PTY, case + regex toggles (incl. invalid
regex error), Esc does not reach the shell, backspace adjusts count.
- plan-search: TipTap plugin — highlight count, exactly one current
match, counter, Enter / Shift+Enter wrapping, close clears
decorations, no case/regex toggles in plan bar.
- session-search: WebSocket-backed hits list, "No matches" state,
click-to-flash, cross-session isolation, close, and virtuoso renderer
coverage via the new ?renderer= override.
- panel-focus-routing: bar mounts inside the focused panel only.
Three tests capture GIFs for the PR description via prCapture when
CAPTURE_PR_ASSETS is set (one per panel). Helpers live in
e2e/helpers/panel-search.ts and e2e/helpers/seed-session-messages.ts;
SessionPage gains search-bar locators.
* fix(web): prevent plan-search update loop on spurious editor transactions
TipTap fires `transaction` on every cursor move/click. The match-info
handler allocated a fresh `{current, total}` object each time, so React
always saw new identity, re-rendered, and could cascade with downstream
effects (e.g. `rehydrateCommentMarks`) into a "Maximum update depth
exceeded" loop. Use a functional setState with equality bail-out so the
handler only schedules a render when the counter actually changes.
* fix(web): retry terminal search once xterm addon-search finishes loading
The SearchAddon is dynamic-imported and was stored in a ref. The
query→findNext effect called `addonRef.current?.findNext(...)` — a
silent no-op if the addon wasn't loaded yet — and its deps didn't
include addon readiness, so a query typed before the import resolved
was dropped forever. Track the addon via state so the search effect
re-runs when it becomes available (and re-binds cleanly if the terminal
is ever re-created). Also deduplicate match-info updates via functional
setState + equality check.
* fix(web): route Ctrl+F to the last-clicked panel when focus is ambient
The chat panel's message list has no focus-capturing child, so clicking
on a message left `document.activeElement === <body>` and the existing
`isFocusWithin` gate silently fell through to the browser's native find.
Fix with two complementary mechanisms:
- `usePanelSearch`: track the most-recently-clicked panel via a
window-level pointerdown capture listener. When focus is ambient
(body/html), that panel claims Ctrl+F. Window-level capture dodges
any descendant `stopPropagation` that would skip a container-level
listener.
- Chat panel root: tabIndex=-1 + onMouseDown that focuses the root when
the click target isn't already focusable. Normal path matches how
plan/terminal already work (TipTap/xterm capture focus on click).
* feat(web): show agent profile name and logo in session search hits
The hit list showed uppercase "AGENT"/"USER" with vivid blue/green
colors that clashed with the app's muted metadata aesthetic. Replace
with the actual author identity:
- Agent hits render the profile name (falling back to "Agent") with
the profile's <AgentLogo> inline — same mechanism session tabs use.
- User hits render "You" with a subtle `text-primary/80` tint that
echoes the user bubble's `bg-primary/30`, keeping a light color
differentiator without reintroducing loud blue/green.
Added `useSessionAgentIdentity` (plus a `useSessionAgentProfile` helper)
that reads from the store via two primitive selectors — the id, then
the profile — and composes the final object inside `useMemo`. A naive
single selector returning `{label, name}` triggers the "getSnapshot
should be cached" infinite loop in `useSyncExternalStore`.
* style: apply prettier formatting to search files
* fix(web): inline resolveStrategy in useMemo to satisfy react-hooks/use-memo
* fix(backend): escape LIKE metacharacters in message search
User query of `%` previously built LIKE pattern `%%` which matched every
message in the session; `file_name` also matched `filename`, `file name`,
etc. Escape the three LIKE metacharacters (`\`, `%`, `_`) before wrapping
the pattern, and declare `ESCAPE '\'` on the clause. Add regression tests
covering wildcard literals and the literal-percent happy path.
* fix(backend): slice search snippet in rune space to keep UTF-8 valid
`buildSnippet` indexed and sliced on byte offsets, so content containing
multi-byte characters (emoji, CJK, accented letters) could be cut
mid-rune — `encoding/json` then replaced the invalid bytes with `\ufffd`
in the websocket response. Switch to rune slicing and per-rune
case-folding via `unicode.ToLower`; this also avoids the
`strings.ToLower` byte-length-change pitfall (e.g. Turkish İ, German ß).
Add a handler test covering emoji, CJK, and large multi-byte content.
* fix(web): cancel backfill loop when user clicks a newer session hit
`setActiveHit` awaits up to 40 `loadOlder()` iterations when the target
message isn't in the DOM. If the user clicked a second hit while those
awaits were in flight, the in-progress loop kept running and eventually
scrolled to the original target, overriding the user's intent. Gate the
loop on a generation counter so only the most recent call's backfill
drives the scroll.
* fix(web): gate plan-search scroll side-effect on dispatch
TipTap invokes commands with `dispatch === undefined` during `can()`
checks to test applicability (e.g. before enabling a menu item). The
next/prev commands called `scrollToMatch` outside the `if (dispatch)`
guard, so a dry-run invocation would scroll the editor to a match the
user didn't navigate to. Move the scroll inside the dispatch block.
* fix(web): address review nits on search UX
- formatTime now checks Number.isNaN(d.getTime()) before formatting;
Invalid Date doesn't throw, so the catch branch was unreachable and
hits with a malformed created_at rendered the literal "Invalid Date".
- prefers-reduced-motion now silences the .search-flash animation
alongside the other custom keyframe animations.
- planScript escapes `title` symmetrically with `content` so callers
passing arbitrary strings don't produce malformed JSON.
* fix(web): fold description into opts to keep seedTask under param limit
`seedTask` had 6 parameters, violating the repo's 5-parameter cap and
failing CI lint. Fold `description` into the existing `opts` object and
update all callsites.
* test(web): add unit tests for useSessionSearch + remove redundant ternary
Cover the non-trivial hook logic called out in review: debounced fetch
coalescing, request-ID cancellation for superseded queries, null
sessionId short-circuit, and close() resets. Also simplify the
`matches.length ? 0 : 0` ternary flagged as a dead branch — both sides
were already 0.
* fix(web): exclude PanelRoot (tabindex=-1) from the focus short-circuit
`target.closest("[tabindex]")` was matching PanelRoot itself (which has
`tabIndex={-1}` so it can receive programmatic focus), causing the guard
to early-return and the `panelRef.focus()` call to never fire — defeating
the Ctrl+F focus-routing that this handler was added for.
Narrow the selector to `[tabindex]:not([tabindex='-1'])` so it only
short-circuits on genuinely interactive elements.
* fix(web): read isReadOnlyMode from live ref in shell-terminal key handler
`attachCustomKeyEventHandler` is called once and has no dispose hook, so
the inner handler keeps its initial closure forever. Other values
(sessionId, send, onFindInPanel) are read via `stateRef` so they stay
current, but `isReadOnlyMode` was captured by closure — if the prop ever
toggled mid-session, the Cmd+Arrow→Home/End path would continue firing
against a read-only terminal. Include it in the state ref for
consistency.
* fix(web): flatten doubly-nested seedTask opts in virtuoso spec; drop unused MODIFIER
The regex-based transform in 7152b936 over-nested one call in
session-search.spec.ts — `opts.description` received an object rather
than a string, so the virtuoso renderer test never received a proper
task description. Flatten it.
Also remove MODIFIER imports/declarations in terminal-search.spec.ts and
panel-search-shared.spec.ts that were left dangling behind `void` suppressors.
* fix(web): abort session-search backfill loop on close/unmount
Move reset logic into `close()` so the lifecycle is explicit (and avoids
the set-state-in-effect lint path the previous reset effect triggered).
Bump the `activeHitGenRef` generation counter on both `close()` and
unmount so an in-flight `setActiveHit` backfill can no longer issue
`loadOlder()` calls for a hidden UI.
Also extract `useDebouncedSearch`, `useSetActiveHit`, and a
`focusMessageElement` helper to keep `useSessionSearch` under the
100-line function cap.
* perf(web): skip PanelSearchBar debounce for session search
Session search already debounces the WS call inside `useDebouncedSearch`
(180ms), but the shared bar adds its own 150ms debounce on top — giving
a 330ms keystroke→request latency instead of the intended ~180ms.
Pass `debounceMs={0}` so the bar forwards queries immediately and the
hook-level debounce is the single source of timing truth.
* feat: add context window usage display to task session
- Add TokenUsageDisplay component showing context window usage as a horizontal progress bar
- Add useSessionContextWindow hook with proper Zustand reactivity (subscribe to primitives)
- Add ContextWindowState to store with setContextWindow action
- Backend: propagate context_window data through orchestrator, watcher, and lifecycle events
- WebSocket handler extracts context_window from session.state_changed metadata
- Display positioned on left side of action bar, updates in real-time
* refactor: remove agent stream event WebSocket broadcasting
Agent stream events (message_chunk, tool_call, etc.) are now handled
internally by the lifecycle manager rather than being broadcast to
the frontend. Messages are accumulated and flushed as comments at
step boundaries.
Carlos Florêncio