Reconcile #5873's "timestamps in copied events" (merged to main) with
this branch's full-body copy rewrite: traceEventCopyText now prefixes
each line with the RFC 3339 timestamp when created_at is valid, on top of
the complete (untruncated) body, events separated by a blank line. The
two #5873 copy tests are updated to the full-body format; the orphaned
formatEventForClipboard helper is removed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Review follow-ups (no behavior change):
- Delete tool_calls_one/tool_calls_other from all four locales — the
metadata chip that used them was removed earlier in this branch.
- Correct the AttributionBadge inline-variant JSDoc: it renders the bare
name (no "on behalf of" wrapper), matching the implementation and test.
- en events_one grammar: "1 event" not "1 events".
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
handleCopyAll built its text from traceEventSummary — the one-line,
whitespace-collapsed, 200-char row preview — so every copied tool output
and agent reply was truncated with "...". Add traceEventCopyText, which
returns the complete body (full tool input JSON / result output / prose),
redact it like the detail view, and join events with blank lines.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The row put the agent, the trigger word, and the person at near-equal
weight with two same-size avatars, so it read as a jumble of names with
"comment-triggered" floating between them.
- The agent is the sole primary identity (avatar + medium weight).
- Trigger + person collapse into one muted secondary unit set apart from
the agent, ordered "<person> · <how>".
- Drop the person's avatar here (new AttributionBadge `hideAvatar`) so
two same-size faces don't read as two agents; the name alone, being
muted, can't be confused with the agent.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- The accountable human returns to the identity row next to the trigger
mechanism (status · agent · trigger · person), where it's visible
rather than buried in the ⓘ popover.
- Toolbar facts now carry explicit labels — "Created <time> · Took
<duration> · N events" — instead of bare values the reader had to
infer.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Three follow-ups from review:
- Remove the "Preserve filters" toggle. Persisting the filter selection
is the only sensible behavior (a facet a run lacks already no-ops), so
it should not be a user-facing switch. Filters now always persist;
drop preserveFilters + the session-vs-persisted branching from the
store and dialog.
- Inline attribution showed "on behalf of <name>", which read backwards
and doubled up under the ⓘ "Triggered by" label. Render just the
avatar + name; the source stays in the tooltip.
- Surface the run time (started ?? created) on the toolbar left next to
duration and event count — "when did this happen" is read-before-you-
read context. Full-precision timestamps remain in the ⓘ popover.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Toolbar controls were hand-rolled <button>s, so they never got the
project's built-in aria-expanded active state — opening a menu left the
trigger looking idle. Route every control through the shared Button:
- Density / filter / ⓘ / close → Button variant="ghost" via the trigger
render prop, so an open menu shows the muted active background for free.
- Active filter → Button variant="brand" (the design-system ON state),
replacing the ad-hoc blue classes.
- Sort → ghost Button on the same chassis (was a segmented tab strip).
- Copy → ghost Button.
Wording:
- Sort reads "Oldest first" / "Newest first" (was the ambiguous
"chronological / time order").
- Expand-mode "Smart" → "Focus" — names the result, not a hollow adjective.
Start/created/completed timestamps remain in the ⓘ Run-details popover.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- Tool-result summary took the first non-empty line, so pretty-printed
JSON previewed as a lone "[" or "{". Collapse whitespace instead so it
reads "[ { "id": ... } ]".
- Sort was a segmented tab strip, reading as a different control family.
Make it a single two-state toggle button on the shared toolbar chassis
(shows current direction, flips on click) — every toolbar control now
shares one chassis with a type-appropriate affordance (toggle / menu /
action), not one forced shape.
- Attribution left the identity row. The accountable human is audit
metadata, not read-time context, and "on behalf of" misread the
direction; the trigger *mechanism* stays on the identity row, the
*person* moves to the ⓘ popover as "Triggered by".
- Drop the tool-call count from the toolbar: redundant with the event
count and informs no reading decision.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Following PR #5747's design lesson: split header metadata by whether a
viewer needs it BEFORE reading. The flat facts line put diagnostics
(provider, runtime hostname, mode, workdir, timestamps) on the
always-visible surface with equal weight.
- Identity row (tier 1): status · agent · trigger source · attribution.
Trigger source ("Initial run" / "From a comment" / "Retry" / ...)
answers "why does this run exist" — restored from #5747, was dropped.
- Status badge now covers the full state machine (queued / dispatched /
cancelled / waiting), not just running/completed/failed.
- ⓘ Run-details popover (tier 2): runtime, provider, mode, workdir
(copyable), created/started/completed — off the default surface.
- Toolbar left carries duration + event/tool counts (a read-before-you-
read summary), so the control row balances instead of stranding the
controls at the right with empty left space.
Persisted preferences (sort, filter selection, preserve-filters, expand
density) are untouched — same controls, same behavior.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The header mixed five different natures in one visual container: status
pill, attribution badge, nine equal-weight metadata chips (including
truncated agent-description prose and a hostname), and the list controls
— everything a rounded bordered capsule, so information and controls
were indistinguishable and wrapped into a ragged 2-3 lines.
Restructure into three fixed-purpose rows:
- Identity: status pill anchors the left edge (the fact every viewer
opens the dialog for), then the agent through its existing avatar
component (hover card for details; fixes the empty-name case via
agentInfo fallback) and a new borderless `inline` AttributionBadge
variant. Close stays at the right.
- Facts: one dot-separated plain-text line — provider, runtime mode
(hostname in hover title), duration, counts, timestamp; the workdir
path collapses into a copy icon with the path as tooltip. The
agent-description chip is gone (it lives in the avatar hover card).
- List toolbar: expand mode / sort / filter / copy move to their own
row attached to the list they operate.
Only controls keep borders; only status keeps color; entities render
through their identity components; facts are typography.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The transcript rendered all five event kinds as identical truncated
one-liners: agent replies and errors were clipped to half a sentence and
only readable through an 11px bordered scroll box, indistinguishable in
weight from dozens of tool rows.
- trace-event-presenter.ts: pure presentation rules — kind, verbatim
tool labels, newline-free one-line summaries, shell-wrapper stripping
for command summaries, and per-kind default expansion.
- Smart reading hierarchy: agent text renders in place through
RichContent (compact density + transcript-prose log scale: headings
demoted to body size, 12.5px/11px two-step type ramp) and errors read
unboxed; thinking and tool rows stay folded to one line.
- Expand mode menu replaces the expand-visible toggle: a persisted
three-way preference (smart / expand all / collapse all) with per-item
descriptions; row-level toggles override it until the mode changes.
transcript-view-store migrates the legacy defaultExpanded boolean.
- Tool params/output expand into a quiet borderless surface; long
content fades behind "Show all" instead of a nested scrollbar.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Virtuoso items opt out of native scroll anchoring, so in newest-first
sort every live flush prepended rows and shifted the reading position.
Compensate with a counting-down firstItemIndex (Virtuoso's prepend
contract); sort/filter/task changes remount the list via an epoch key so
the value never increases within one instance. scrollToIndex keeps
data-relative indices — verified against scrollToIndexSystem, which
never reads firstItemIndex.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Long runs produce hundreds to thousands of task-message events; the
transcript dialog mapped every event straight into the DOM, freezing the
browser when the log grew. Render the event list through react-virtuoso
so only the visible window mounts. Timeline segment navigation switches
from node refs + scrollIntoView to index-based scrollToIndex, and sort
toggling scrolls the virtualized list instead of the container div.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Feishu 话题 sessions are isolated per topic (#5061), but the group
recent_context prefetch read the whole chat: a @-mention inside topic B
could pull topic A's messages into B's prompt and persist them into B's
turn (#5835). Fix the leak at the fetch:
- ListChatMessages sends container_id_type=thread&container_id=<thread_id>
when a topic id is present; the thread container rejects end_time, so the
window is anchored to the trigger time client-side instead.
- Parse thread_id off the REST item and fail-close on it: a topic fetch
keeps only exact thread_id matches, dropping any missing/mismatched item
so a sibling topic can never leak even if the API returns one.
- A topic fetch failure degrades to the readable note and NEVER falls back
to a chat-wide fetch (that would re-open the leak); a very busy topic may
get empty context, which is the safe trade.
- Exclude the Bot's own interactive-card replies from the window (they
flatten to a zero-signal [interactive card] placeholder).
- Non-topic group @-mentions keep the chat-level, end_time-anchored path
unchanged.
Closes#5835
Co-authored-by: Bohan-J <bohan@devv.ai>
Co-authored-by: multica-agent <github@multica.ai>
Renumber the six VCS migrations off the 213/214/215 prefix collisions to a unique contiguous tail 216-221 (tables before indexes), and make 214_chat_session_project idempotent (ADD COLUMN IF NOT EXISTS) so the #5868 rename-induced re-run no-ops instead of crashing. Fixes the red migration lint on main and unblocks the dev deploy.
* feat(issues): show agent-working indicator in Table view
Reuse the self-contained IssueAgentActivityIndicator (already in List and
Board views) in the Table view's Issue column, rendered right after the
identifier and before the title — matching the List view placement so the
'agent working' cue sits in the same spot across all three views. The
indicator returns null when no agent is active, so inactive rows are
unchanged.
MUL-5258
Co-authored-by: multica-agent <github@multica.ai>
* test(issues): cover Table agent-working badge insertion
The previous stub returned null with no assertion, so deleting the badge
insertion in table-view.tsx would still pass every test here. Render a
marker carrying the issue id and assert the Table cell mounts it for the
row's issue, positioned between the identifier and the title (identifier →
activity → title, matching List/Board). Verified the new test fails when the
insertion is removed.
MUL-5258
Co-authored-by: multica-agent <github@multica.ai>
---------
Co-authored-by: Bohan-J <bohan@devv.ai>
Co-authored-by: multica-agent <github@multica.ai>
Adds self-hosted Git provider support (Forgejo, Gitea, GitLab) alongside GitHub:
per-workspace token connection, a provider-dispatched webhook, PR/MR and CI
mirroring, and the shared issue auto-link / auto-close machinery. Off until
MULTICA_VCS_SECRET_KEY is set, so existing deployments are unaffected.
Co-authored-by: Bohan <bohan@devv.ai>
Chat project context (MUL-5150, #5765) is rendered into the run brief by the
daemon, so daemons older than the release that ships it silently drop the
project description while still honoring the server-extracted repos. Mirror
the handoff-note soft gate: resolve the active agent's runtime cli_version
from the warm runtime cache and, when it is a stale release build, surface a
warning next to the composer chip and inside the project submenu. Selection
is never blocked — dev-describe builds and unknown runtimes stay silent.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
PR #5765 (MUL-5150) branched before #5841 merged, so both landed a 213
migration and Backend CI's TestMigrationNumericPrefixesStayUniqueAfterLegacySet
failed on main. Bump #5765's pair to the next free prefixes, preserving the
column-before-index order:
213_chat_session_project -> 214_chat_session_project
214_chat_session_project_index -> 215_chat_session_project_index
MUL-5251
Co-authored-by: Bohan-J <bohan@devv.ai>
Co-authored-by: multica-agent <github@multica.ai>
* feat(chat): add project context
Co-authored-by: multica-agent <github@multica.ai>
* fix(chat): resolve MUL-5150 review blockers
- Renumber project-context migrations to unique prefixes after current main:
206_chat_session_project -> 212 (column), 207_chat_session_project_index ->
213 (concurrent index). 206/207 collided with 206_agent_disabled_runtime_skills
and main's 207-211 client_usage_daily set.
- Add the 4 missing chat input.project_context keys to ja/ko locales so the
locale parity test passes (en/zh-Hans already had them).
- Lock the project-context control while a send is in flight (isSubmitting),
not just while the agent is running. A brand-new chat creates its session
lazily during send bound to the project at click time; switching project
mid-send would create the session against the stale project and clear the
editor as if the send landed on the new selection. Add a regression test.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: multica-agent <github@multica.ai>
* fix(chat): complete project context handling
* fix(chat): pin fresh chat to open session's agent on project switch
Switching an existing session to a different project opens a fresh chat but
only cleared the active session, dropping selection back to the stored
`selectedAgentId`. When that preference was stale (open session belongs to
agent B while the persisted pick is still agent A), the lazily-created session
and its first send bound to the wrong agent (agent A).
Extract the project-switch decision into a shared `planProjectContextChange`
pure helper in use-chat-controller.ts and route both chat surfaces (the chat
tab controller and the floating ChatWindow) through it, so the fresh chat is
pinned to the open session's agent and the rule cannot drift between the two
copies. Add a dual-entry regression test (pure-fn guard + controller
integration) covering the stale selectedAgentId case.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: multica-agent <github@multica.ai>
* chore(ci): re-trigger required checks on latest head
The prior push updated the branch ref but GitHub did not emit a pull_request
synchronize for it (PR head-sync lag), so CI/Mobile Verify never ran on the
commit carrying the stale-agent project-switch fix. Empty commit to force a
fresh synchronize on a head that includes it.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: multica-agent <github@multica.ai>
* fix(chat): renumber project migrations to 213/214 after main added 212
Current main added 212_agent_service_tier; the PR's 212/213 chat migrations
collided with it on the merge ref, failing TestMigrationNumericPrefixesStay
UniqueAfterLegacySet. Merge current main and move the chat column migration to
213 and the concurrent index migration to 214 (column before index preserved).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: multica-agent <github@multica.ai>
* fix(chat): lock ProjectPicker clear control during send (keyboard path)
The send-pending lock only put pointer-events-none on the wrapper, which
blocks the mouse but leaves ProjectPicker's inline clear button in the tab
order — a keyboard user could Tab to "Remove from project" and press Enter
mid-send, detaching the project after the lazily-created session already went
out with the old one (reopens the mid-send retarget path via keyboard).
Add an explicit `disabled` capability to the shared ProjectPicker that locks
the trigger, the menu (forced closed), and the inline clear button (disabled +
out of the tab order). Defaults to false, so issue/create/autopilot callers
keep their hover/keyboard clear. ChatInput passes disabled while the project
selection is locked.
Tests: real-ProjectPicker regression (keyboard activation of the clear control
is inert when disabled; still works when enabled) + ChatInput wiring assertion
that the picker is disabled mid-send.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: multica-agent <github@multica.ai>
---------
Co-authored-by: Lambda <lambda@multica.ai>
Co-authored-by: multica-agent <github@multica.ai>
Co-authored-by: Walt <walt@multica.ai>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: Naiyuan Qing <145280634+NevilleQingNY@users.noreply.github.com>
Co-authored-by: NevilleQingNY <nevilleqing@gmail.com>
* fix(web): resolve upstream URLs at runtime
* fix(web): keep unconfigured upstreams same-origin
* fix(web): restore dev-only localhost fallbacks for API and docs upstreams
`next dev` on a developer machine now falls back to the conventional
http://localhost:8080 backend (honoring BACKEND_PORT) and
http://localhost:4000 docs origin when nothing is configured, so a bare
`pnpm dev:web` keeps proxying out of the box. Builds and the runtime
proxy keep the strict resolvers, so prebuilt images still leave unset
upstreams unproxied.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Naiyuan Qing <145280634+NevilleQingNY@users.noreply.github.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
MUL-5240
cursor-agent's stream-json never populates total_cost_usd or a per-step
cost — the result event's usage object carries token counts only. Both
fields were speculatively copied from Claude Code's schema in the original
Cursor runtime PR (#1057) and were never read. Verified against the real
CLI (2026.07.20, stream-json and json) plus Cursor's CLI docs.
Remove the two dead fields and document that Cursor spend stays estimated
from the static rate table (no authoritative per-turn cost to carry, unlike
Grok). No behavior change.
Co-authored-by: Bohan-J <bohan@devv.ai>
Co-authored-by: multica-agent <github@multica.ai>
Cursor stream-json emits reasoning and tool calls as top-level `thinking` /
`tool_call` events; the parser only looked for them inside assistant messages,
so transcripts showed a single step. Match the top-level events, taking the
tool name from the nested `<name>ToolCall` key and normalizing the packed
call_id. Subtypes are matched explicitly — only `started` opens a tool, only
`completed` closes it, only `delta` carries reasoning — so an unknown/missing
subtype is ignored rather than synthesizing a fake result (which would decrement
the daemon in-flight tool count early and misfire the watchdog) or polluting
reasoning. Covered by a recorded-stream fixture test, an unknown-subtype
regression test, and an opt-in real-CLI smoke test.
kimi-code 0.29 dropped the top-level `models.availableModels` /
`currentModelId` block from its ACP `session/new` response and moved the
same catalog into a `configOptions` entry with `id`/`category` of
"model". parseACPSessionNewModels only understood the old shape, so
discovery silently returned an empty catalog and the model picker showed
"no available models" for an online, correctly-detected kimi runtime.
Parse `configOptions` as a fallback: the `models` block still wins when
present, so no existing ACP provider changes behaviour. `options[].value`
becomes the model id, `options[].name` the label, and `currentValue`
marks the default. Non-model options (thinking level) are deliberately
skipped — they are a separate product surface and would offer values
`session/set_model` cannot honour.
Also log a debug line with the top-level response keys (keys only, never
values) when session/new succeeds but advertises no catalog, so the next
round of upstream schema drift is visible in daemon.log instead of
looking like a PATH or install failure.
Co-authored-by: Bohan-J <bohan@devv.ai>
Co-authored-by: multica-agent <github@multica.ai>
* fix(agent): attribute Grok usage from the turn's own model id
A resumed Grok session with no configured model recorded its entire spend
under the model id "unknown", which matches no pricing row — so the task
reported $0 cost instead of its real spend.
grok.go only learned the model from the session handshake, and ACP's
`session/load` carries no model id (only `session/new` does). When neither
the agent nor MULTICA_GROK_MODEL pins a model, `daemon.go` legitimately
passes an empty model, leaving nothing to attribute the usage to.
Every Grok turn stamps `result._meta.modelId` with what it actually billed
against. Parse it in the shared ACP result parser and use it as the fallback
in grok.go. Other ACP backends are untouched — they keep whatever the
handshake gave them.
Co-authored-by: J <agent@multica.ai>
Co-authored-by: multica-agent <github@multica.ai>
* fix(metrics): price the Grok catalog in server-side cost metrics
server/internal/metrics/pricing.go carried no Grok rows at all, so
RecordLLMUsage took the unpriced branch for every Grok turn: llm_cost_usd
reported zero Grok spend while the tokens accumulated in
llm_unpriced_tokens. Internal cost monitoring simply could not see Grok.
Add the six SKUs xAI publishes rates for, mirroring the frontend table in
packages/views/runtimes/utils.ts. Aliases are anchored exact matches like
the gpt-5.6 rows, so `grok-composer-*` (in the catalog, absent from the
price sheet) stays unmapped instead of inheriting a guessed rate.
Short-context tier on purpose: xAI bills a request at 2x once its prompt
reaches 200K tokens, but a usage record aggregates every model call in a
turn and cannot say which tier an individual request hit.
A regression test re-derives the cost of a real grok 0.2.106 turn from the
table and checks it against the costUsdTicks xAI returned for that turn.
Co-authored-by: J <agent@multica.ai>
Co-authored-by: multica-agent <github@multica.ai>
* docs(changelog): scope the Grok cost claim to what was actually fixed
The v0.4.9 entry promised "accurate cost" in all four languages, but the
fix corrected catalog pricing and cached-input double-counting — it did not
implement xAI's 2x long-context tier, so a turn whose requests reach 200K
prompt tokens still under-reports by up to 50%. Say what was fixed instead.
Also correct two stale claims in the pricing comment: the daemon tags usage
rows with the runtime provider `grok`, not `xai` (the bare `grok-*` keys are
what make them resolve), and record why thresholding the long-context tier
on an aggregated row would be worse than not pricing it at all.
Co-authored-by: J <agent@multica.ai>
Co-authored-by: multica-agent <github@multica.ai>
* feat(usage): carry the provider's own cost through to the usage record
Cost has always been derived client-side as tokens x a static rate, which
cannot express request-level pricing rules. xAI bills a Grok request at 2x
once its prompt reaches 200K tokens, and a task_usage row aggregates every
model call in a turn — so the stored token counts genuinely cannot say which
tier any individual request hit. Thresholding on the aggregate would be worse
than the status quo: it turns a bounded 50% under-estimate into an unbounded
over-estimate for turns made of many short requests.
Grok already reports what it charged, per turn, in `_meta.usage.costUsdTicks`.
Parse it, carry it through agent -> daemon -> API, and store it on task_usage
as a nullable BIGINT of 1e-10 USD ticks (integer, so sub-cent turns stay exact
end to end). NULL means the provider reported no cost — every pre-existing row
and every provider that doesn't return one. No backfill: there is no
authoritative figure to recover for those, and inventing one is the guess this
removes.
A single hourly bucket can mix rows that carry a cost with rows that don't, so
task_usage_hourly gains both halves: `cost_usd_ticks` sums the authoritative
side, and `uncosted_*_tokens` carry exactly the tokens that still need a
rate-table estimate. Consumers report authoritative + estimate(uncosted),
which degrades to today's behaviour when nothing in the bucket is
authoritative. The existing token columns keep covering every row, so token
displays are untouched. The new columns are additive with defaults, so the
unique key, the dirty-queue shape, and migration 102's triggers are unaffected.
Co-authored-by: J <agent@multica.ai>
Co-authored-by: multica-agent <github@multica.ai>
* feat(usage): prefer the provider's own cost over the rate table
With the authoritative figure now stored, both cost consumers use it: the
usage dashboard (estimateCost / estimateCostBreakdown) and the server-side
llm_cost_usd metric. Each reports `authoritative + estimate(uncosted tokens)`,
so a row or bucket that mixes priced and unpriced sources stays whole.
The static rate tables remain, but for Grok they are now a fallback — they
still price usage recorded by a daemon too old to report cost, and every
provider that reports none. Custom pricing overrides likewise apply only to
the estimated half: they are a user's guess at a rate, and the authoritative
half is not a guess. A model with no rate-table row but a provider-reported
cost now also drops out of the "unmapped models" banner, since asking the user
to supply a rate for it would invite overriding a real bill.
llm_cost_usd is labelled by token_type and the provider reports one number per
turn, so the charge is distributed across the buckets in the rate table's own
proportions. Only the total is authoritative; the split stays an estimate,
which is why this scales the existing buckets rather than inventing a label.
estimateCostBreakdown does the same, keeping the stacked chart summing to the
headline figure instead of silently under-drawing every Grok row.
Co-authored-by: J <agent@multica.ai>
Co-authored-by: multica-agent <github@multica.ai>
* docs(changelog): say Grok cost now follows xAI's actual charge
The earlier wording scoped the claim down to catalog pricing and cached input
because the long-context tier was still unhandled. It is handled now — the
cost comes from what xAI charged for the turn — so the entry can say so.
Co-authored-by: J <agent@multica.ai>
Co-authored-by: multica-agent <github@multica.ai>
* fix(usage): keep the provider's cost when the model has no rate row
Both cost consumers bailed out before reading the authoritative figure when
the rate table had no row for the model. A `grok-composer-*` turn — in the
Grok Build catalog, absent from xAI's price sheet — was therefore reported as
$0 spend even though xAI told us exactly what it charged.
Worse on the client: estimateCost returned the real cost while
estimateCostBreakdown returned zeros, so the headline and the stacked chart
disagreed on precisely the rows whose cost is exact — and the unmapped-models
banner was (correctly) hidden, so nothing explained the discrepancy.
Handle the charge before the rate lookup in both places. Without rates there
is nothing to split a total by, so it lands whole in the `input` bucket, the
same fallback distributeAuthoritativeCost already uses when it has no shape to
scale. Tokens with no rate keep going to llm_unpriced_tokens: "unpriced"
describes the rate table, not the money.
Co-authored-by: J <agent@multica.ai>
Co-authored-by: multica-agent <github@multica.ai>
* perf(usage): drop the historical rewrite from the cost-split migration
Migration 213 rewrote every existing task_usage_hourly row to seed the
uncosted counters. That is a full-table UPDATE inside a schema migration —
lock time, WAL and bloat all scaling with table size — for rows this issue
explicitly does not care about.
Deleting the UPDATE alone would have zeroed historical cost: with
`NOT NULL DEFAULT 0`, an untouched row asserts "nothing here needs
estimating", so every pre-split bucket would report $0 until the rollup
happened to touch it. Make the uncosted columns nullable with no default
instead. NULL means "never recomputed since the split existed", readers
COALESCE it to the row's own token total ("estimate all of it"), and the
pre-split behaviour is preserved exactly — with nothing to seed, so no
rewrite. A bare ADD COLUMN is metadata-only, so this is now fast DDL.
Rows heal into the split naturally as the rollup recomputes their buckets.
Verified on a fresh database: a legacy-shaped row reads back as its full
tokens to estimate, and a group mixing legacy and post-split buckets sums to
the authoritative cost plus both rows' estimable tokens.
Co-authored-by: J <agent@multica.ai>
Co-authored-by: multica-agent <github@multica.ai>
---------
Co-authored-by: Bohan-J <bohan@devv.ai>
Co-authored-by: J <agent@multica.ai>
Co-authored-by: multica-agent <github@multica.ai>
* fix(github): surface CI status on PR cards (MUL-5180)
The CI mirroring pipeline (MUL-2228, MUL-2392) has never received a single
event in production. The GitHub App setup docs only ever asked operators to
grant `pull_requests: read` and subscribe to `pull_request`, so GitHub never
delivered `check_suite` — `handleCheckSuiteEvent` sat dead behind a
subscription nobody was told to enable. Every linked PR reports
checks_passed/failed/pending = 0 and the sidebar row falls through to
"Checks haven't reported yet" forever.
Docs (the root cause), all four locales:
- add `Checks: Read-only` permission + `Check suite` event to the App setup
table
- drop the stale "CI check states are not modeled" claim, which predates
MUL-2228 and is what let the setup table stay incomplete
- add a "PR rows show no CI status" troubleshooting entry with the public
`/apps/<slug>` probe to confirm what an App is actually subscribed to, and
a warning that existing installations must accept the new permission
before any `check_suite` is delivered
UI:
- give the actionable status kinds (checks failed/pending/passed, conflicts,
ready) their own icon + color. CI outcome previously rendered as plain
muted 11px text, visually identical to the diff stats beside it — a failing
build read the same as "+437 −6 · 6 files". Terminal and unknown kinds stay
muted; the row's state icon already carries that meaning.
Co-authored-by: multica-agent <github@multica.ai>
* fix(github): unbreak docs build, stop overclaiming CI completeness (MUL-5180)
Both must-fixes from review.
1. docs production build failed. `<your App>` in prose was parsed as a JSX
tag, so `pnpm --filter @multica/docs build` died with `Expected a closing
tag for <your>`. Dropped the angle brackets. Repo CI never caught this
because no workflow runs the docs production build — only Vercel does,
which is why the PR's GitHub checks were green while the deployment
errored.
2. `Checks: Read-only` cannot support the pending status the docs promised.
GitHub's webhook contract delivers `check_suite.requested` /
`.rerequested` only to Apps holding Checks *write*; read-level access
receives `completed` only. Verified against GitHub's published docs.
Direction chosen: keep read-only, degrade honestly to final-results-only.
Checks *write* is a repo-write capability (create/update check runs), not
a wider read — escalating every installation to it just to render an
in-flight spinner is not a trade to make on the operator's behalf, and it
contradicts the integration's read-only posture.
The concrete bug this leaves is premature green: with two reporting apps,
the first to complete makes total=1/passed=1 and the row claimed "All
checks passed" while the second was still running and might fail. Copy is
now "Checks passed" in all four locales — it reports what reported and
never asserts completeness. `derivePullRequestStatusKind` documents why.
Docs gain a "what CI status can and cannot tell you" section (all four
locales) with the read-vs-write delivery table, both consequences stated
plainly, and the opt-in path for teams that do want in-flight status: set
Checks to Read and write on their own App and the existing pending code
lights up with no code change. The pending promise is removed from the
read-only setup path.
Co-authored-by: multica-agent <github@multica.ai>
* fix(github): ignore non-completed check_suite actions (MUL-5180)
Review was right: the `Read and write` opt-in the previous commit documented
does not produce reliable pending, and following it would break the card.
`check_suite.requested` / `.rerequested` are not observations that some CI
provider started. GitHub sends them only to Apps holding Checks write, and
per the CI-checks App docs they mean "GitHub has created a check suite for
YOUR app on this commit; now add your check runs to it".
Multica observes other apps' results and never creates check runs. Recording
such a suite parks a `queued` row nothing can ever complete, and since
`checks_pending` outranks `checks_passed` in derivePullRequestStatusKind, one
stuck row freezes every PR on that installation at "checks running" and hides
the real pass/fail result. Any self-hoster who already grants Checks write
hits this on every push, so the gate is on the action, not the permission.
- handleCheckSuiteEvent drops every action except `completed`, with the
reasoning and the "don't resurrect requested as a running signal" warning
recorded at the gate.
- TestWebhook_CheckSuite_QueuedCountsAsPending encoded the wrong delivery
semantics (two external apps sending `requested`, which GitHub never does).
Replaced by TestWebhook_CheckSuite_NonCompletedActionsIgnored, which pins
the drop and checks a later `completed` suite still lands.
- The two out-of-order stash tests used `requested` payloads to exercise
paths that are really about completed suites; both now use `completed` and
assert the same guarantees.
- Docs (four locales): the write opt-in is gone. In-flight CI is documented
as unsupported at any permission level, with the actual reason and the note
that real running status needs polling or a check_run model instead.
Co-authored-by: multica-agent <github@multica.ai>
* fix(github): make legacy non-completed check suites inert (MUL-5180)
Review was right again: the previous commit gated the webhook entry point but
left the pre-upgrade state — and the people it was meant to protect (self-
hosters who already granted Checks write) are exactly the ones holding it.
Two leftovers, both now closed:
1. Rows already in github_pull_request_check_suite. The old handler stored
GitHub's `requested` suites as `queued`; nothing will ever complete them.
ListPullRequestsByIssue still counted them, so `checks_pending` kept
outranking `checks_passed` and the PR stayed pinned to "checks running"
for as long as its head SHA stood. The aggregation now selects only
`completed` suites.
Filtering beats deleting here: recovery is automatic on deploy, needs no
migration over a table that can be large, and holds for any writer that
misses a gate — not just for today's legacy rows. DISTINCT ON runs after
the filter, so an app whose newest suite is a stuck `queued` still reports
its most recent completed verdict instead of disappearing.
2. Rows already in github_pending_check_suite. replayPendingCheckSuitesForPR
is a second write path into the live table that never passes through
handleCheckSuiteEvent, so the next `pull_request` event would re-inject a
permanently-queued suite after the fix shipped. It now skips non-completed
rows; the drain is DELETE ... RETURNING, so skipping discards them.
Both are covered by regression tests that seed the legacy row directly — the
fixed handler can no longer produce one — and both were confirmed to fail
with their respective fix reverted. The stash test additionally asserts its
fixture landed under the repo address the drain keys on; the first draft used
the wrong owner and passed vacuously.
Also corrects the aggregateChecksConclusion doc comment, which still
described "pending" as a not-yet-completed suite. It is now reachable only
for a completed suite carrying a null conclusion, and is explicitly not a
"CI is running" signal.
Co-authored-by: multica-agent <github@multica.ai>
* test(github): assert the legacy stash row is consumed by the drain (MUL-5180)
Review nits.
The stash test proved its fixture existed before the webhook but never that
the drain consumed it, so a future change to firePullRequestWebhookWithHead's
repo address would make the assertions pass for the wrong reason again — the
same way the first draft of this test did. Asserting the stash is empty
afterwards closes that gap from the other side.
Also fixes two comment typos: `an "CI is running"` -> `a`, and drops the
"merged-but-open PR" state, which cannot exist.
Co-authored-by: multica-agent <github@multica.ai>
---------
Co-authored-by: Bohan-J <bohan@devv.ai>
Co-authored-by: multica-agent <github@multica.ai>
* fix(runtime): forbid blocking on external CI in the brief (MUL-5223)
The external-work boundary added in #5803 did not stop agents from
waiting on GitHub Actions. Two holes: the section's only concrete
"how to wait" example was a blocking foreground call, which is exactly
the shape of `gh pr checks --watch`; and the "unless acceptance
criteria require it" escape was satisfied by the repo's own merge
requirement that CI be green.
Name the banned tool shapes, allow a single non-blocking status
snapshot, deny branch protection as an acceptance criterion, and give
the replacement hand-off phrasing (local test result + PR link).
Co-authored-by: multica-agent <github@multica.ai>
* fix(runtime): scope the CI-wait ban so the explicit exception stays executable (MUL-5223)
Review feedback on #5840:
- The ban read as absolute ("Blocking on external CI is never part of
your deliverable") while the next bullet allowed waiting when the
task explicitly asks for the CI result, leaving no way to satisfy
both. The ban is now scoped to "unless the explicit exception below
applies", and the exception names the one executable shape: a single
foreground blocking watch inside the same turn.
- `gh pr merge --auto` enables auto-merge and returns; it is not a
wait. Only waiting for it to land is banned.
Both hard-pin tests now also pin the exception so it cannot be dropped
or re-absolutised.
Co-authored-by: multica-agent <github@multica.ai>
* polish(runtime): group Background Task Safety into run-owned and external-CI clusters (MUL-5223)
Co-authored-by: multica-agent <github@multica.ai>
* polish(runtime): cut redundant phrasing from the external-CI cluster (MUL-5223)
The cluster said "report and finish" three different ways and carried
two rhetorical tails. Fold the delivery template into the post-push
playbook bullet, tighten the merge-gate denial, and drop filler.
5 bullets -> 4, -36 words, every behavioral fact and test pin intact.
Co-authored-by: multica-agent <github@multica.ai>
---------
Co-authored-by: Bohan-J <bohan@devv.ai>
Co-authored-by: multica-agent <github@multica.ai>
Grok Build reports cachedReadTokens inside inputTokens (totalTokens ==
input + output on a real 0.2.106 turn, and that turn's costUsdTicks
matches xAI's rates only when the cached prefix is billed once). The
shared ACP parser persisted both counters raw, so the usage dashboard
charged the cached prefix at the full input rate *and* the cache-read
rate — ~4x the real spend on a cache-heavy turn.
Re-bucket cached reads out of input when totalTokens proves the overlap,
the same normalization codex.go already applies. Backends that report
mutually-exclusive buckets or omit totalTokens are untouched.
Co-authored-by: J <agent@multica.ai>
Co-authored-by: multica-agent <github@multica.ai>
Windows Security flags multica.exe inside app.asar.unpacked as
Trojan:Script/Wacatac.B!ml and quarantines it. The !ml suffix is a
machine-learning heuristic verdict, not a signature match: our Windows
builds are not Authenticode-signed (release.yml sets
CSC_IDENTITY_AUTO_DISCOVERY=false and no certificate is configured), and
an unsigned, low-prevalence Go binary that spawns background processes
and opens sockets is the classic false-positive profile.
Add a Desktop docs section (en/zh/ja/ko) covering how to verify the
binary against checksums.txt, restore it from quarantine, exclude both
the install directory and %APPDATA%\\Multica (Desktop re-downloads the
CLI there, so excluding only the install dir loops), and report the
false positive to Microsoft.
MUL-5216
Co-authored-by: Bohan-J <bohan@devv.ai>
Co-authored-by: multica-agent <github@multica.ai>
* feat(usage): price the published xAI Grok catalog
Grok usage resolved to no pricing row, so every Grok token contributed
$0.00 to cost totals and the models sat permanently in the unmapped
diagnostic. Add the six SKUs xAI publishes rates for.
Short-context rates on purpose: xAI bills a request at 2x once its
prompt reaches 200K tokens, but aggregated usage rows carry no
per-request prompt sizes — the same trade-off the Anthropic `[1m]`
context tag already takes. `grok-composer-*` ships in the Grok Build
catalog but is absent from the price sheet, so it stays unmapped rather
than inheriting a guessed rate.
Source: https://docs.x.ai/developers/pricing
Co-authored-by: multica-agent <github@multica.ai>
* fix(usage): keep custom pricing editable after it is saved
The custom-pricing dialog had exactly one entry point: the button inside
the unmapped-models banner. That banner only rendered while something in
the window was unpriced, and `resolvePricing` falls back to the custom
store — so saving a rate for the last unmapped model made it resolve,
the banner disappeared, and the saved rate could no longer be edited or
removed. Clearing localStorage was the only way out, which also wiped
every other override.
Keep the bar mounted whenever there is something to manage: warning
state while models are unpriced, a quiet row with an edit button once
they all resolve but overrides exist. Hidden entirely otherwise.
The existing store mock hard-coded `getCustomPricing` to undefined, so
no test could ever observe a saved override; make it stateful and add
the regression case.
Co-authored-by: multica-agent <github@multica.ai>
---------
Co-authored-by: Bohan-J <bohan@devv.ai>
Co-authored-by: multica-agent <github@multica.ai>
* fix(desktop): open in-app links in a tab instead of the browser (MUL-5208)
A link written as an absolute URL on this deployment's own origin
(`https://<app-host>/acme/issues/1` — what "copy link" produces, and what
agents paste into chat) fell through openLink's external branch to
window.open, which Electron routes to shell.openExternal. Clicking an issue
link in Desktop chat therefore opened a browser window instead of a tab.
openLink now resolves such a URL back to its in-app path and takes the same
route a relative path does. Backend-served prefixes (/api/, /_next/) stay
external so attachment downloads keep working.
Two supporting fixes the change depends on:
- The `multica:navigate` event had no listener on web, so in-app paths in
content were dead links there; normalizing app URLs would have extended
that to every pasted app URL. The web platform layer now answers the event
with a router push.
- The desktop handler opened every path inside the active workspace's tab
group. A cross-workspace link now goes through switchWorkspace, matching
what the navigation adapter already does for pushes.
Co-authored-by: multica-agent <github@multica.ai>
* fix(desktop): scope in-app link conversion to workspace pages, answer it in issue windows
Review follow-ups on MUL-5208.
1. The non-page exclusion list only covered /api/ and /_next/, so a
same-origin /uploads/* link — local-storage attachments, served by the
backend and proxied by web — was routed as an app page, opening a dead tab
instead of the file. Replaced the deny-list with the app's own routing
model: an absolute URL converts to an in-app path only when its first
segment is a slug a workspace could own, which the existing reserved-slug
list (shared with the backend) already answers. /api, /uploads, /_next,
/favicon.ico and the pre-workspace routes all stay external without a
second list to keep in sync.
2. A dedicated issue window derives the same app origin from its adapter but
had no multica:navigate listener, so a same-origin link there became a
silent no-op (it used to reach the browser). The window now answers the
event: another issue opens in place — matching what its adapter push and
mention chips already do — and any other app page, which this single-route
window cannot host, goes to the browser.
Co-authored-by: multica-agent <github@multica.ai>
---------
Co-authored-by: J <agent@multica.ai>
Co-authored-by: multica-agent <github@multica.ai>
* fix(resize): keep the drag cursor consistent for table and chat resize
Table column resize and the chat floating-window resize locked the active
cursor via document.body.style.cursor, which loses the CSS cascade to any
descendant that declares its own cursor (clickable rows, inputs, buttons).
So the resize cursor showed on hover but flipped back to pointer/text/default
once the drag pointer swept over those elements — inconsistent with the
sidebar rail and react-resizable-panels, which force it globally.
Switch both to the same robust contract used by the sidebar/panels: set a
data attribute on <html> during the drag and let a global
`html[...], html[...] * { cursor: ...; user-select: none } !important` rule in
base.css win over every descendant. Chat keys the attribute by direction
(left/top/corner) so col/row/nw-resize stay correct.
MUL-5166
Co-authored-by: multica-agent <github@multica.ai>
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* fix(chat): clean up chat resize cursor lock on every drag-end path
The chat window resize started the drag with setPointerCapture but only
removed data-chat-resizing on pointerup. A pointer-capture drag can end via
pointercancel, lostpointercapture, or window blur without ever firing
pointerup; since the attribute now drives a global
`html[...] * { cursor; user-select } !important` rule, a missed cleanup would
strand it and freeze the whole page in the resize cursor with text selection
disabled.
Mirror the sidebar rail: one idempotent finishDrag() wired to pointerup,
pointercancel, lostpointercapture and blur, releasing pointer capture and
removing every listener plus the attribute from a single path. Add a
regression test covering each termination path.
MUL-5166
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: multica-agent <github@multica.ai>
* fix(ui): release table resize cursor lock on window blur too
Table column resize wires stopResize to pointerup/pointercancel but not to
window blur. With the resize cursor now held by a global
`html[data-table-resizing] * { cursor; user-select } !important` rule, an
alt-tab mid-drag (button released while the window is blurred, so no pointerup
arrives) would strand the attribute and freeze the page. stopResize is
idempotent, so also wire it to blur — matching the sidebar rail and chat.
MUL-5166
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: multica-agent <github@multica.ai>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: multica-agent <github@multica.ai>
The chat composer was capped at a fixed max-h-40 (160px), so a long draft
was previewed through a five-line porthole no matter how large the chat
window was. Cap the composer at 50% of the surface it sits on instead,
with a 384px absolute ceiling, and let the card shrink into that cap via
a flex-column wrapper.
Verified in Chromium across all four surfaces that mount ChatInput
(floating window at min and expanded size, chat tab, agent builder): a
long draft caps at min(50% of the surface, 384px) and scrolls internally,
a short draft still hugs its content, and the message list keeps space.
Co-authored-by: Bohan-J <bohan@devv.ai>
Co-authored-by: multica-agent <github@multica.ai>
* docs(self-host): add Upgrading section to the self-host quickstart (MUL-5190)
The quickstart had no upgrade section — the only mention of upgrades was a
side note that migrations run automatically. Community users had to be walked
through it by hand every time, and the two circulating recipes (`make selfhost`
vs bare `docker compose pull && up -d`) were never documented as equivalent.
Adds an `Upgrading` section between Step 7 and the Kubernetes section covering:
- both command forms, and that `make selfhost` is a wrapper around them
- what `git pull` actually updates (the compose file, not the image version)
- the `MULTICA_IMAGE_TAG` pinning trap — pinned `.env` silently blocks upgrades
- `.env` is only generated when missing, never overwritten
- Postgres backup before upgrading
- migrations run on backend startup, with the log command to watch them
- verify with `/readyz`, not `/health` (liveness passes on failed migrations)
- cross-reference to the existing Kubernetes upgrade path instead of duplicating
Synced to the zh / ja / ko translations.
Co-authored-by: multica-agent <github@multica.ai>
* docs(self-host): fix upgrade guidance for k8s pull policy and pg_dump (MUL-5190)
Review follow-ups on the new Upgrading section.
1. The Kubernetes upgrade path was wrong. Both the new section and the
pre-existing Kubernetes section presented `kubectl rollout restart` as a way
to pull new images, but the chart ships `pullPolicy: IfNotPresent`
(deploy/helm/multica/values.yaml), so a node that already cached the tag
reuses the old image and the restart silently changes nothing — the same
failure mode as a pinned MULTICA_IMAGE_TAG. `helm upgrade` with an updated
`images.*.tag` is now the primary path; the floating-tag route is documented
only with the `pullPolicy: Always` prerequisite it requires.
2. The Postgres backup masked `pg_dump` failures. `pg_dump … | gzip > out.gz`
takes the pipeline's exit status from gzip, so a failed dump exits 0 and
leaves a valid, empty 20-byte archive. Now redirects to a file first so
pg_dump's own status gates compression via `&&`, with a callout explaining
why.
3. Dropped the `Makefile:79-107` / `Makefile:80-95` line references. The
selfhost target actually spans 79-133 and does more than pull + up (creates
.env when missing, waits on /health, prints a status summary). Reworded to
"same result on an existing install" and referenced the target by name so
the claim cannot drift with line numbers.
Applied across all four language versions.
Co-authored-by: multica-agent <github@multica.ai>
---------
Co-authored-by: Bohan-J <bohan@devv.ai>
Co-authored-by: multica-agent <github@multica.ai>
* fix(views): raise hover delay on the agent activity badge
The per-issue agent activity badge sits on the right edge of dense
scrolling lists (inbox rows, issue rows, board cards) and appears on
every issue an agent currently touches. Base UI's 600ms default opened
the 288px activity card on pointer travel rather than on intent.
Raise the open delay to 900ms and drop the close delay to 150ms. The
card body is read-only, so there is no hover bridge to protect.
MUL-5189
Co-authored-by: multica-agent <github@multica.ai>
* fix(inbox): drop the agent activity hover card from inbox rows
The badge already shows who is running and whether they are working or
queued. On a triage surface the card's only incremental fact is elapsed
time, which never changes the one decision an inbox row supports — do I
open this? The row also carries the ActorAvatar hover card on the left,
so a second popup was mostly noise.
Add an opt-out `hoverCard` prop to IssueAgentActivityIndicator and pass
false from the inbox row. Issue lists and board cards keep the card:
monitoring work in flight is what those views are for, and they keep the
900ms dwell delay from the previous commit.
MUL-5189
Co-authored-by: multica-agent <github@multica.ai>
---------
Co-authored-by: Bohan-J <bohan@devv.ai>
Co-authored-by: multica-agent <github@multica.ai>
