Repair signalled "reinstall me" by deleting the bootstrap marker. That was
already destructive -- repair is reachable from a transient backend error on a
fully working install -- and it stranded users in first-run setup with no way
back short of hand-writing the marker file.
Carry the intent in an explicit flag instead. Repair forces the next resolve
through the installer and clears itself once the reinstall starts, so a forced
reinstall still works without destroying provenance about how the install was
created.
Closes#72166
Marker presence was the launch gate, but the marker is provenance about who
ran the install -- not proof the runtime works. A CLI-installed repo+venv, or
a healthy install whose marker a repair deleted, both read as "never
installed" and dropped the user into first-run bootstrap on every launch.
Split the two questions: classifyActiveRuntime() reports marker validity and
runtime usability separately, and the resolver launches whenever the runtime
is usable, logging when it proceeds without a marker. An unusable runtime
still falls through to bootstrap even with a valid marker, so an interrupted
install can't spawn a dead backend.
Drops isBootstrapComplete(), which had no callers left once the gate moved.
Co-authored-by: iveywest <iveywest@users.noreply.github.com>
Co-authored-by: lihengming <lihengming@users.noreply.github.com>
install.ps1 wrote the marker on Windows and the Rust installer now writes it,
but install.sh -- the path every Mac and Linux CLI install takes -- never did.
A machine set up with install.sh therefore looked uninstalled to the desktop
app, which re-ran first-run bootstrap on every launch.
Stamp the same schema-v1 payload install.ps1 writes, from both the staged
`complete` stage and monolithic main(). An unresolvable HEAD skips the marker
rather than writing one the desktop validator rejects: absent reads as a clean
"bootstrap needed", malformed reads as a confusing half-state.
The macOS launcher fast path gates on hermes_is_installed(), which needs
.hermes-bootstrap-complete next to a built desktop app. Nothing in the Rust
bootstrap pipeline ever wrote that marker -- only install.ps1 did -- so every
reopen of /Applications/Hermes.app re-ran setup instead of launching.
Publish the marker atomically (temp sibling + fsync + rename) because
hermes_is_installed() only checks existence: a torn direct write would arm
the fast path against a half-installed tree. A marker write failure emits
BootstrapEvent::Failed so the installer UI leaves the progress state.
Co-authored-by: giggling-ginger <giggling-ginger@users.noreply.github.com>
A pasted link went in as raw URL text, wrapping across the composer and
staying inert. It now becomes the same `@url:` reference the "+ → Add URL"
dialog inserts — parsed in place, so a link mid-sentence keeps its position
and the punctuation that ended the sentence stays outside the chip. Typing
one and pressing space commits it the same way.
Both rich-editor surfaces get it: the composer and the message-edit box,
whose paste went through `execCommand` and could not produce a chip at all.
`insertPlainTextAtCaret` dropped its text in verbatim, so a caller with
directives in hand had no way to land them as chips. It becomes
`insertComposerContentsAtCaret`, sharing the parse `renderComposerContents`
already uses, and gains a sibling `replaceBeforeCaret` for swapping a
just-typed token for a chip.
The composer labeled a `@url:` chip with `refLabel`, which takes the last
path segment — three PR links all read as their number. The transcript had
its own better labeler that stopped at the hostname, so the same reference
read differently before and after send.
One `refChipLabel` now serves both: host without `www.`, path riding along
for the chip's existing truncate to cut, and the full value on the chip's
title so a cut-off link is still readable on hover.
* feat(billing): add payment_method union to the billing-state wire type
* feat(billing): carry the payment-method union through the gateway
NAS now sends a typed `paymentMethod` union on /api/billing/state alongside
the legacy `card` field. The gateway parses payloads field-by-field, so the
new field was dropped on the floor before reaching TUI/Desktop.
Parse it into PaymentMethodInfo and re-emit it as snake_case `payment_method`,
matching the translation the rest of this payload already does. The payment
method id is deliberately not carried through — clients have no use for it.
No client rendering changes: `card` stays populated for cards, so every
existing consumer behaves exactly as before and the new field is inert until
a surface opts into reading it.
* fix(billing): send only the fields each payment-method kind declares
The serializer emitted every key for every kind, so a Link method went out
carrying brand, last4 and wallet set to null. That contradicts the shared
type, where each kind declares its own fields: a client testing `'brand' in
pm` would read every Link method as a card, and one trusting the declared
non-null `brand` could crash on it.
Send each kind's own fields, and forward an unrecognized kind by name alone
so a client that predates it can still say something honest. The shared type
gains the matching fallback arm its own comment already promised.
Tests now follow a payload from the server response through to the client
wire for each kind, rather than checking parsing and serializing separately —
which is why the old expectation locked in the wrong shape without noticing.
* fix(billing): keep the payment-method kind narrowable
Typing the fallback arm's kind as `string & {}` borrowed a trick that only
works on unions of plain strings. On a union of objects it makes the
discriminant non-literal, so TypeScript stops narrowing on every arm — even
`if (pm.kind === 'card')` no longer gives you `brand`. The first client to
use this would have hit a compile error and reached for a cast.
An unrecognised kind now arrives as `unknown`, carrying the real name
alongside it, so every arm has a literal discriminant. A type-level test
pins this: it fails to compile if the discriminant stops narrowing.
The parser settles which kind it is, the way the card parser already does,
so the record cannot hold fields that do not belong to its kind and the
serializer no longer re-checks. The type comment also stops claiming `card`
is a safe signal — it is null for Link, so `!card` does not mean "nothing on
file".
The sweeper review on #66435 flagged that the collector doesn't bind
session["profile_home"]. That binding is intentionally unnecessary: the
kanban board is shared across profiles by design — kanban_home() anchors
on get_default_hermes_root(), which resolves the process env and ignores
context-local profile overrides (see the kanban_db.py module docstring).
Add a regression test that claims a subscription while a foreign-profile
set_hermes_home_override() is active, proving delivery still works for
non-launch-profile Desktop sessions.
Covers the poller wiring above _collect_kanban_notifications, per the
hermes-sweeper review: status.update emission, agent-turn dispatch via
_run_prompt_submit when the session is idle, and the busy-session
pending buffer that flushes once the session goes idle.
kanban_create auto-subscribes TUI/desktop sessions with platform="tui" and
chat_id=HERMES_SESSION_KEY, and tools/kanban_tools.py documents that the
TUI notification poller (tui_gateway/server.py) reads kanban_notify_subs
and posts completion messages into the running session — but that reader
was never implemented. The poller only watched process_registry completion
events, and the gateway notifier skips "tui" rows because no such
messaging adapter exists. Result: subscriptions accumulate with
last_event_id=0 forever and no task event is ever delivered (18 subs,
29 terminal events, 0 deliveries in the report).
Implement the missing delivery path in the TUI notification poller:
- every 5s, claim unseen terminal events for this session's
platform="tui" subscriptions via claim_unseen_events_for_sub — the
same atomic cursor-claim the gateway notifier uses, so an event is
delivered exactly once even with a gateway polling the same board DB
- format events with the same wording as the gateway notifier
(done/blocked/gave up/crashed/timed out/status; archived and
unblocked are claimed but silent, so they can't wedge the cursor)
- emit a status.update for user visibility, then chain an agent turn
when the session is idle — mirroring process-completion handling;
claimed events buffer in the session until it goes idle since the
cursor (unlike the process queue) cannot re-queue
- unsubscribe only at a truly final task status (done/archived),
matching the gateway rule so respawned tasks keep notifying
- multi-board: iterate boards, polling each resolved DB path once
Fixes#59890
The default max_iterations/agent.max_turns budget was set when long
agentic runs were rare; complex tasks now routinely exceed 90 tool
calls. Raise the default to 500 across every surface that hardcodes
the fallback: AIAgent constructor, DEFAULT_CONFIG, CLI resolution
chain, gateway env bridge, cron scheduler, and TUI gateway. Explicit
user config values are unaffected (deep-merge preserves them; no
_config_version bump needed).
Docs (en + zh-Hans), CLI help text, tips, and pinned tests updated
to match.
Two remaining formatting variants that silently killed file delivery:
- A trailing sentence period (MEDIA:/x/data.csv.) failed the boundary
lookahead, so the tag neither extracted nor stripped. The period is now
accepted as a boundary only when followed by whitespace/EOL, keeping
multi-part extensions (.tar.gz) intact.
- A whole tag wrapped in inline code (`MEDIA:/x/data.csv`) was masked as
a prose example (#35695). Models routinely format paths as inline code,
eating real deliveries. Inline-code tags now deliver when the path
validates on disk; non-existent example paths stay masked and fenced
code blocks remain fully masked.
Adds a regression matrix covering both plus the salvaged contributor
fixes (emphasis wrap, glued tags, glued [[as_document]], dedupe,
unknown-extension and extensionless delivery).
MEDIA_DELIVERY_EXTS in gateway/platforms/base.py omitted .3gp, causing
MEDIA: tags with .3gp files to leak as plain text instead of being
extracted for native video delivery. _VIDEO_EXTS in
tools/send_message_tool.py and _MIGRATION_VIDEO_EXTS in the Feishu
adapter omitted .webm, causing .webm files to be classified as
documents instead of video on Telegram and other platforms.
Both extensions are already present in every gateway-side _VIDEO_EXTS
definition (run.py, kanban_watchers.py, weixin.py, base.py local).
Closes#71621, Closes#71603
Surface a user-visible delivery notice when non-streaming media dispatch
gets success=False after MEDIA tags were stripped, and validate forum
starter-message attachments the same way as direct channel sends (#66797).
Outbound MEDIA video/document tags on the Discord non-streaming path
were extracted (stripped from the visible text) but never delivered:
no attachment, no error, and no dispatch log line (#66797). The
open-handle discord.File plus singular file= form could race Discord's
multipart encoder after an earlier image batch on the same channel and
return a successful message carrying zero attachments.
Fix _send_file_attachment (used by send_video/send_document/
send_image_file) to:
- use a path-based discord.File via the plural files=[...] kwarg, the
same pattern as the working send_multiple_images batch path;
- pre-flight os.path.isfile and return a File not found result instead
of raising deep inside discord.File;
- fail loud when Discord accepts the message but attaches nothing, so
the dispatch loop surfaces a warning rather than a silent drop;
- add INFO dispatch logs (base non-image MEDIA fan-out, video send, and
file attachment) so a MEDIA:.mp4 cannot vanish without a trace.
Tests (tests/gateway/test_discord_send.py):
- path-based files=[...] kwarg used, singular file= unset;
- fail-loud when the returned message has no attachments;
- missing file fails fast without resolving the channel;
- forum-parent delivery routes through create_thread with files=[...];
- end-to-end image+video response routes the mp4 to send_video while
images still batch.
Fixes#66797
When the same file is referenced multiple times in one message (common
when the agent emits MEDIA tags both inline and in a summary footer),
the platform adapter uploads/sends the same file twice — visible as
duplicate attachments in Telegram, duplicate posts in Slack, etc.
Set-based dedup on the expanded path, preserving first-occurrence order.
Two MEDIA: path tags emitted back-to-back without a separator merged
into a single invalid path and were silently dropped. The same happened
for extension-less tags.
Root cause: both regexes in gateway/platforms/base.py used greedy
quantifiers in their path class, causing adjacent tags to be absorbed.
Fix: make both regexes non-greedy (add ? to quantifiers) and add
MEDIA: to the trailing lookahead boundary set so the next MEDIA:
keyword stops the current match cleanly.
Models routinely present a file to the user with the delivery tag wrapped in
Markdown emphasis — `**MEDIA:/path.pptx**`, `*MEDIA:/path*`, `_MEDIA:/path_`.
MEDIA_TAG_CLEANUP_RE only tolerated a single leading/trailing quote or backtick
(`[`"']?`), and its closing lookahead set excluded `*` and `_`, so an
emphasis-wrapped tag never matched. The file was then silently never delivered
and the literal `MEDIA:/path` text leaked into the chat instead — the user sees
a path, not the attachment.
Allow a short run of emphasis/quote markers (`[`"'*_]{0,3}`) on both sides of
the tag and add `*`/`_` to the closing lookahead. Code-block, inline-code and
blockquote contexts are still neutralised earlier by `_mask_protected_spans`
(#35695), so documentation/example tags remain non-deliverable; the
absolute-path anchor still rejects relative paths; `_` inside a filename is
unaffected.
Adds regression coverage in TestExtractMedia for bold/italic/underscore
wrapping, mid-prose bold, emphasis-wrapped .html, underscore-in-filename, and
emphasis-wrapped relative-path rejection.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
MEDIA_TAG_CLEANUP_RE failed to match when a directive like [[as_document]]
was concatenated directly to the file extension without whitespace
(e.g., MEDIA:/home/user/report.xlsx[[as_document]]). This caused the
file to be silently not delivered while the gateway reported success.
Root cause: the lookahead character class [\s`",;:)\}\]|$) did not
include [, so [[as_document]] immediately after the extension broke
the lookahead assertion.
Fix: add \[ to the lookahead class so directives can follow the path
without whitespace. This is safe because [ is already stripped elsewhere
in the same file via .replace("[[as_document]]", "").
Added regression tests covering:
- Directive glued to extension (issue case)
- Directive with whitespace (baseline)
- Tag at end of string ($ anchor)
The statusbar subscribed to `$focusedSessionState` — a projection of
`$sessionStates`, which is republished on every message delta — but reads
only three fields off it. Every token therefore re-ran useStatusbarItems
and rebuilt all ~9 item objects, and since StatusbarItemView was not
memoized, one changed item (the running timer) re-rendered the whole bar.
Adds `useStoreSelector` beside the existing `useSessionSlice`, the same
narrowing idea for a scalar instead of a keyed array: subscribe to the
store, but bail out unless the selected value changes. Applies it to the
three fields the statusbar actually reads, and memoizes StatusbarItemView.
Measured over five concurrent streaming tabs (`render-churn`):
total renders 78,385 -> 21,701 (-72%)
wasted renders 10,432 -> 6,112 (-41%)
TooltipContent 2,304 -> 143 (-94%)
StatusbarItemView 2,174 -> 0
The Windows-footgun check is red on current main, not just on this branch:
a2c42be93c added `encoding="utf-8"` to one `write_text` in this file and
missed the other two.
Line 190 is what the checker reports. Line 211 has the same defect but the
call is split across lines, so the single-line regex never flagged it —
fixing only the reported site would have left the same bug in the file and
re-armed it for the next reader.
Both now pass an explicit encoding, matching line 68.
PR infographics belong in the PR description, referenced from the
image-provider URL. The binary never enters git history.
This rule has been established twice and leaked twice. #48261 removed the
first batch. #54564 removed a second batch and added `infographic/` to
.gitignore — but .gitignore only stops an accidental `git add`. It does
nothing against `git add -f`, and nothing for a directory that does not
literally match the pattern. In the four weeks after that rule landed,
nine more PNGs were force-added, and an `infograficos/` directory
(#70552's loophole, never actually closed) slipped a tenth past the
pattern entirely.
Removes 11 tracked images (~14MB) with `git rm --cached`, so local copies
survive. Adds an infographic-check CI job that matches on the IMAGE rather
than on one directory spelling, so a localized or typo'd path cannot
sidestep it, and extends the .gitignore pattern list as the first line of
defence.
Verified the guard both ways against synthetic repos: it fires on
`git add -f` into `infographic/`, on the `infograficos/` spelling, and on
nested `docs/pr/infographics/*.jpg`; it does not fire on legitimate
product imagery under `docs/assets/` or `website/`, nor on non-image
files.
The identity-refresh TTL used 0.0 as the "never checked" sentinel and
compared it against time.monotonic(). That epoch is arbitrary and starts
near zero on a freshly-booted host, so on CI runners and containers
`monotonic() - 0.0` was itself below the TTL — "never checked" read as
"checked just now" and the first identity refresh was suppressed for the
first 5 minutes of uptime. The stale-handle recovery therefore did nothing
on exactly the machines most likely to be freshly booted.
Invisible on a long-lived dev box (uptime >> TTL); caught by CI.
- Sentinel is now None, meaning never checked and always stale.
- Both TTL comparison sites route through _bot_identity_is_fresh().
- Regression test pins the invariant under a faked 12s-uptime clock.
Verified by re-running the recheck tests against a simulated 3s-uptime
host: green with the fix, and restoring the 0.0 sentinel reproduces the
exact CI failure locally.
Renaming a Telegram bot's @username in BotFather silently stopped the
gateway from answering in groups.
PTB caches getMe() in Bot._bot_user and only rewrites it inside get_me(),
so after a rename the adapter kept comparing mentions against the OLD
handle. The exclusive-mention gate then saw the new @handle, failed to
match itself, and concluded the message was addressed to a different bot
— dropping it before the reply and wake-word fallbacks could run. Native
replies to the bot were discarded too. Polling mode recovered on the next
90s heartbeat; webhook mode never calls get_me() again, so it stayed dead
until restart.
Separately, the bot-handle pattern assumed every bot username ends in
"bot". Collectible (Fragment) usernames can be assigned to bots and drop
that suffix (@jarvis, @pic), so such a bot could not recognise its own
handle in the entity-less fallback and was suppressed by any message that
also named another bot.
- Route every mention comparison through _current_bot_username(), which
prefers the last observed handle over PTB's cache.
- Learn the live handle from inbound updates: Telegram stamps the current
username on our own messages and on reply_to_message. Guarded by user
id, so another account's handle is never adopted.
- Re-check identity out of band (TTL-bounded, one getMe per 5 min) when
the exclusive gate is about to drop a message — the exact stale-handle
symptom — so the mistake self-corrects instead of persisting.
- Refresh identity in webhook mode via a dedicated low-frequency loop,
cancelled on the same teardown fence as the heartbeat.
- Match our own handle by identity rather than shape. Foreign handles keep
the deliberate "...bot" narrowing so human @handles still never act as
routing hints (the intent behind ce4d857021).
Validation: 11 regression tests; sabotage runs confirm each behavioral
test fails with the fix reverted. 157 tests green across the Telegram
gating, reconnect, and topic-mode suites.
Connecting Discord asked five questions when the platform needs one. The card
listed a home channel ID you need Developer Mode to copy, an allow-all-users
security toggle, a reply-threading preference, and a home channel display name
— all with working defaults, none discoverable from the form.
Drops them from the setup surfaces entirely: the dashboard/Desktop channel
cards and the `hermes setup gateway` wizard. Discord is now bot token +
allowlist. Matrix drops from 11 fields to 7, Mattermost from 6 to 3.
Suffix-matched (`*_HOME_CHANNEL*`, `*_ALLOW_ALL_USERS`, `*_REPLY_TO_MODE`,
`*_REQUIRE_MENTION`, `*_AUTO_THREAD`, `*_FREE_RESPONSE_*`, `*_PROXY`) so plugin
platforms nobody enumerated get the same treatment. Allowlists deliberately
stay — the gateway denies everyone until one is set, so that IS the decision a
new user has to make. Required credentials are never hidden.
Nothing is removed from the product. The vars still work through
`hermes config set`, .env, and config.yaml, the gateway reads them unchanged,
and dropping them from the cards hands them back to the Keys page rather than
orphaning them (Keys hides only what a Channels card owns).
CI slices 2 and 7 caught three tests broken by always-defer:
- /tools (CLI show_tools + TUI gateway tools.show) now passes
skip_tool_search_assembly=True — it's a discovery/inspection surface,
so users verifying an MCP installed must see deferred tools, not a
collapsed bridge row. This also fixes
test_slash_worker_mcp_discovery (profile MCP tool visible in /tools).
- test_plugins.py::test_plugin_tools_in_definitions: 'visible' becomes
'reachable' — direct schema OR listed in the bridge description;
scope-exclusion assertions unchanged (not direct AND not listed).
- test_discord_tool.py dynamic-schema-rebuild test reads the
pre-assembly list (the rebuilt schema is what tool_describe serves).
Banner/status tool counts intentionally keep the post-assembly view —
they reflect what the model actually sees.
The include/exclude filter matched exact names only — glob-style entries
('*_radar_*') silently matched nothing, so a Cloudflare flat-mode config
meant to trim 3,320 tools to ~1,900 actually registered 3,319. Unmatched
patterns produced no warning.
- matches_name_filter(): exact membership first (O(1) for literal lists),
then fnmatch.fnmatchcase for entries containing * ? [ — same pattern
semantics as approvals.deny. Entries without metacharacters stay
strictly literal ('docs' never matches 'docs_search').
- _should_register() uses it for both include and exclude (symmetric)
- hermes mcp tools picker (mcp_config.py) pre-selection uses the same
matcher so the UI agrees with runtime registration
E2E against the live Cloudflare capture with the real exclude list:
3,320 -> 1,905 surviving (1,415 excluded); radar/DLP gone,
purge_cache/dns_records kept. 220/220 mcp_tool tests (4 new).
Teknium review changes on the tiered policy:
1. threshold_pct default 10 -> 5 (listing budget = min(5% of context,
listing_max_tokens)); unknown-context fallback 20K -> 10K.
2. Tier 2 no longer leaves the model blind: when even names-only doesn't
fit, the bridge description now carries a one-line-per-server summary
('cloudflare (3320 tools)') plus an instruction to search FIRST rather
than substitute a generic tool or claim the capability is missing —
the measured tier-2 failure mode (core-tool substitution) at zero
meaningful token cost (~50 tokens/server).
3. Listing degradation is now PER SERVER, largest first: one oversized
server (Cloudflare) collapses to its summary line while small
co-attached servers (Linear) keep their full per-tool listings
('mixed' form). Previously global: attaching Cloudflare next to
Linear silently cost Linear its listing. Greedy fit is deterministic
(size then label) so the rendered block stays byte-stable per catalog
— prompt-prefix cache safe.
E2E on real captures (defaults, 200K ctx): linear alone -> tier 1 full;
unreal alone -> tier 2 groups (5% budget) / tier 1 names at 1M;
cloudflare alone -> tier 2 groups; linear+cloudflare -> tier 1 MIXED
(linear fully listed, cloudflare summarized). 48/48 tests.
Cloudflare's flat API MCP ships 61 property keys that violate Anthropic's
^[a-zA-Z0-9_.-]{1,64}$ pattern (query-filter params like 'issue_class~neq'
and 'meta.<field>[<operator>]'). One bad key anywhere in the tools array
400s the ENTIRE request — measured live: Anthropic, Bedrock, Google Vertex,
and Azure all rejected an eager 3,320-tool Cloudflare request at validation,
before token limits even applied.
- schema_sanitizer: rename non-conforming property keys deterministically
(bad chars -> '_', 64-char truncation, collision dedup with numeric
suffixes), nested schemas included; required[] remapped alongside
- unrename_tool_args(): reverse map applied in coerce_tool_args at dispatch,
so the MCP server receives the original wire names; recurses into object
values and array items
- deterministic on both sides: the rename map is recomputed from the
registry's original schema at dispatch time, no state carried
E2E on the real capture: 61 -> 0 violations across 3,320 tools; round-trip
verified on get_accounts_intel_attacksurfacereport_issues (5 '~neq' keys).
Tier 0: no MCP/plugin tools -> everything eager (pass-through).
Tier 1: deferred tools whose catalog listing fits min(threshold_pct%
of context, listing_max_tokens) -> bridge + skills-style
listing, degrading to names-only over budget.
Tier 2: listing over budget even names-only (Cloudflare's flat API
surface: 3,320 tools, names alone ~32K tokens) -> bare bridge,
discovery through tool_search only.
The old activation threshold (defer only when schemas > threshold_pct
of context) let mid-size catalogs ride eager and pay full schema cost;
with servers like Cloudflare (~597K tokens of schema, would not even
fit a 200K window) the binary gate is the wrong shape. Activation is
now driven purely by deferrable-tool presence; threshold_pct is
repurposed as the listing budget's context-relative leg.
- AssemblyResult gains tier + listing_form for observability
- listing_max_tokens default 4000 -> 20000 (cap 60000) so an 830-tool
catalog keeps a names-only listing while Cloudflare-scale drops to
bare bridge
- E2E verified against real captures: Linear 24 tools -> tier 1 full,
Epic UE 5.8 830 tools -> tier 1 names-only, Cloudflare 3,320 tools
-> tier 2 (both 200K and 1M context)
Bridge vs listing only (Opus 4.8, 830 real UE schemas, 3 reps/cell).
Excluding one both-modes mock artifact: listing 24/24 vs bridge 20/24,
searches/task 0.2 vs 4.0. Bridge failures: core-tool substitution at
frontier tier (ran the host test suite via terminal instead of
discovering RunTests, 2/3 reps), up to 8 searches to prove a negative,
and search-vocabulary misses on paraphrase. Listing asserts absence in
zero searches and answers a 5-way capability survey in 1 API call.
Scenarios target real confusion clusters in Epic's UE 5.8 catalog
(StaticMesh vs SkeletalMesh set_material, three tag systems, CurveTable
vs DataTable rows, Niagara Component vs System variables, four capture
variants, zero-keyword phrasing). Mocks return realistic editor errors
on wrong-type calls; scoring is strict (clean solve = correct tool with
zero distractor calls; first-call accuracy tracked separately).
Key result: first-call selection is unreliable in EVERY mode — eager
with all 199K of schemas in context managed 2/10 — but clean solves stay
75-95% because agents probe (get_components, get_material_slots) before
committing. The probe loop works through the 3-tool bridge at 1/4 the
cost of eager ($1.60-1.69 vs $6.49/task, Opus 4.8). On Haiku the
listing beats bare bridge 18/20 vs 15/20 (core-tool substitution again).
Zero distractor invocations across all 50 Opus runs.
Replays the actual tool schemas captured from Epic's UE 5.8
ModelContextProtocol + AllToolsets plugins (830 tools / 52 toolsets) as
live registry tools with mocked editor responses, then benchmarks
eager vs bare-bridge vs bridge+listing at two scales (62-tool editor
subset, full 830) on Claude Opus 4.8 (1M ctx; eager at 830 does not fit
any 200K model — first call requests ~266K tokens).
Headline (full 830, mean per task, rescored): eager 8/8 at 810,578
input tokens ($4.05); bare bridge 16/16 at 160,844 ($0.80); listing
16/16 at 257,264 ($1.29). Frontier model erases the accuracy gap in
every mode; cost is the differentiator. At 62 tools eager wins on cost
— consistent with the auto-threshold design.
Also parameterizes livetest harness model + listing_max_tokens via
env/args (TS_UE_MODEL, TS_UE_SCALE, TS_UE_MODES, TS_UE_LISTING_MAX).
Deferred MCP/plugin tools become invisible once the tool_search bridge
activates — live benchmarking (48 runs, Claude Haiku 4.5) showed models
substituting visible core tools (terminal/web_search/browser) for deferred
capabilities or declaring them nonexistent instead of searching: 16/24
task success vs 24/24 with eager loading.
Skills never had this failure mode because every skill keeps a ~21-token
name+description listing line in the system prompt. This ports that exact
pattern to the tool bridge: when tool_search activates, a grouped manifest
of every deferred tool (name + first sentence of description, clipped to
60 chars, grouped per MCP server / toolset) is embedded in the tool_search
bridge description.
- tools/tool_search.py: build_catalog_listing() with deterministic
ordering (byte-stable across assemblies -> prompt prefix stays
cacheable); token-budget fallbacks full -> names-only -> legacy bare
count; bridge_tool_schemas(listing=...) embeds it and instructs the
model to skip tool_search when the exact name is visible (one fewer
round-trip per use)
- config: tools.tool_search.listing auto|on|off (default auto),
listing_max_tokens (default 4000, clamped 200..20000); legacy bool
shapes keep working
- tests: 8 new tests (config parsing/clamps, short-desc clipping,
deterministic rendering, budget fallbacks, bridge embedding, assembly
on/off paths); full file green (47 passed)
- docs: tool-search.md config table + rationale
- scripts/tool_search_livetest2.py: benchmark harness v2 with real
per-call token accounting (normalize_usage spy) and a third 'listing'
mode for A/B/C comparison
The sidebar strip and the Channels page could contradict each other on
the same page load — "Gateway running" next to "The gateway is not
running." /api/status and /api/messaging/platforms each open-coded their
own liveness ladder: status probed GATEWAY_HEALTH_URL and scoped its
PID/state reads to the requested profile, messaging did neither and used
the uncached raw PID probe.
Three deployments hit the split: a cross-container gateway (no local PID,
only the health probe can see it), a profile-scoped dashboard (messaging
borrowed a DIFFERENT profile's runtime state, reporting a false
"connected" that hides a real outage — #71211), and a launch-service
managed gateway with no PID file.
Adds resolve_gateway_liveness() in gateway/status.py as the single ladder
(cached PID -> HTTP health probe -> runtime-status PID with
expected_home) and routes both endpoints, /api/messaging/platforms/{id}/test,
and the kanban dispatcher-presence probe through it. Probe callables are
injectable so the existing monkeypatch seams keep working, and
GatewayLiveness.probe_error distinguishes "down" from "couldn't tell" so
the kanban warning keeps failing OPEN instead of crying wolf.
Closes#71211.
Drives the same synthetic multi-tab streaming workload as `multitab`
(publishSessionState per session per flush, no backend, no credits) but
reports render attribution instead of frame pacing — sidebar_renders,
wasted_renders, and wasted_notifies.
Answers 'does the sidebar re-render while an agent is typing' directly.
Adds two dev-only counters that attribute re-renders and store
notifications during an interaction, so a perf claim can be answered with
a number instead of a hunch:
window.__RENDER_COUNTS__ what re-rendered, and why (props/state/parent)
window.__ATOM_CHURN__ which store published it, and whether it mattered
The `wasted` column in each is the fix list — components that re-rendered
with no changed input, and stores that published a value equal to the
last one. Both are inert until start(), so idle cost is one branch per
commit and per notify.
React 19.2 removed injectProfilingHooks from react-dom, so the mark*
profiling family is unavailable and onCommitFiberRoot is the only channel
left. <Profiler> can't answer the question either: React invokes onRender
for every Profiler in a committed tree including subtrees that bailed
out, and a bailed-out subtree still reports nonzero actualDuration. This
uses bippy's didFiberRender instead. bippy over react-scan because
react-scan/lite is a thin wrapper over it while the package pulls ~217
transitive deps and floats two on latest.
main.tsx imports the entry statically above react-dom because react-dom
captures the devtools hook at module init — a late install reports
renderers=0 and observes zero commits. Production exclusion is handled by
a build-time alias to a no-op module rather than tree-shaking, since a
static side-effect import can't be eliminated.
While the workspace pane shows a full page (Artifacts, Skills, Messaging, a
plugin route), a sidebar click on the ACTIVE session did nothing: onResumeSession
took focusOpenSession's `true` for the main-session branch as "already on
screen" and skipped the navigate, but fronting the workspace tab doesn't put the
chat back — the page is still routed. The user had to click some other session
and then the active one to get back.
focusOpenSession now reports WHICH surface it fronted ('main' | 'tile' | null),
and focusedSessionNeedsRoute decides: a tile never needs a route (its pane
renders the chat regardless), a main hit does while a page covers the workspace.
