flat_dm_status was speculative config (rubric violation): no user wants
'make my agent look dead', and the only real consumer of status
suppression was native's placement-contamination guard — which the relay
lane handles structurally (QA-6/7 send-side anchor strip, leak-guard
test), not via preference. Status now anchors whenever an inbound ts
exists, in both modes.
Docs: new website/docs/user-guide/messaging/relay.md — enterprise-only
relay lane page documenting the platforms.relay.extra.<platform> subset
shape (nested wins, flat fallback), the Slack reply_in_thread control,
and always-on liveliness. Kept out of the native slack.md on purpose:
relay controls are not Slack config.
Victor's correction: flat DMs CAN have a live thinking status. setStatus
on the triggering message's ts renders '… thinking'/per-tool phrases in
that message's thread-footer space and clears without leaving a message
artifact. Native suppresses this because ITS reply routing could inherit
the activated thread; the relay lane's flat-mode sends strip their
anchors explicitly (QA-6/7), so the status anchor cannot leak into reply
placement — proven by the new leak-guard test.
send_typing/stop_typing now anchor the status in flat mode too, gated by
platforms.relay.extra.slack.flat_dm_status (default ON; false restores
the fully anchorless posture). Thread mode unchanged.
A 2nd top-level DM while a turn was in flight resolved to the SAME
session key and steered the running turn ('Redirected current run')
instead of starting its own. Native SlackAdapter stamps thread_ts =
event.thread_ts or ts on EVERY inbound, so build_session_key isolates
each top-level message; the connector normalizes top-level messages with
thread_id=null and the relay lane never reproduced the stamp.
_stamp_slack_session_thread applies native parity on the inbound bridge:
top-level Slack message + thread-per-message mode => source.thread_id =
its own ts (fresh session, parallel turns). Real thread replies and flat
mode untouched (flat keeps the shared rolling DM session on purpose).
Also introduces the enterprise config shape for relay-fronted Slack:
platforms.relay.extra.slack.<subset of native Slack fields> (nested
object wins; legacy flat extra.reply_in_thread still honoured). All
reply_in_thread reads (send/typing/stop_typing/run.py progress) now
route through one resolver.
The connector's Slack sender threads on metadata ONLY: threadTs() reads
metadata.thread_id/thread_ts and never the frame's reply_to. base.py's
final-reply lane (and its stream-fallback 'first response' resend) builds
metadata from source.thread_id — None for a top-level DM — so its sends
carried reply_to as the sole threading signal and posted to the home
channel (2026-07-27 post-approval report; the 15:17:03 frame showed
meta_keys=['notify','user_id']).
After the QA-6 mode gate keeps the anchor, copy it into
metadata.thread_id so the wire carries the signal where the connector
reads it. Flat mode unaffected (anchor already nulled); explicit thread
metadata wins; non-Slack untouched.
Slack's thinking-status line (thread replies footer, plain chat:write —
no assistant scopes needed) is thread-only: the connector's typing case
no-ops without thread_ts. The typing lane's metadata has no anchor for a
top-level DM (base.py builds from source.thread_id = None), so every
status heartbeat was silently dropped — the trace showed typing frames
with meta_keys=['user_id'] only.
Cache the triggering message ts per chat on inbound (_capture_scope) and
synthesize metadata.thread_id on send_typing/stop_typing in
thread-per-message mode, mirroring native send_typing's
_resolve_thread_ts(metadata.message_id). Flat mode unchanged (#18859);
real-thread metadata wins over the cache; the clear frame targets the
same synthesized thread so the status never sticks.
Native Slack shows dynamic assistant-status text ('Finding answers…',
'is running pytest…') because SlackAdapter sets supports_status_text=True
and renders the set_status_text() phrase in send_typing. The relay lane
advertised nothing, so run.py's live-status lane never fed it phrases and
the connector fell back to the static default.
- supports_status_text: descriptor-gated property (Slack only; other
fronted platforms keep textless bubbles)
- send_typing: carry the stashed phrase as the typing op's content; omit
when unset (empty string is Slack's explicit clear, reserved for
stop_typing). Connector already renders content via
assistant.threads.setStatus (#154).
_resolve_reply_to_for_send dropped the triggering-ts reply_to on every
Slack DM with no metadata thread_id. But the final-reply lane (platforms/
base.py) builds metadata from source.thread_id only — None for a top-level
DM — so in thread-per-message mode that reply_to is the final reply's ONLY
threading signal, and stripping it exiled the final message to the DM root
while progress bubbles stayed threaded (sibling of the QA-5 prompt bug).
Mirror native _resolve_thread_ts: suppress the synthetic anchor only when
platforms.slack.extra.reply_in_thread=false. Flat mode behavior unchanged;
real threads and channels unchanged.
The threading mode (flat vs thread-per-message) is decided once, in run.py's
_resolve_progress_thread_id (reply_in_thread knob), and encoded in the
metadata stamp: flat => no thread_id, threaded => thread_id for the turn
(first turn: == message_id, the synthetic root IS the thread).
_strip_synthetic_dm_thread re-derived the mode with an unconditional
thread_id == message_id strip, exiling approval/clarify cards (and their
resolved-state swaps) to the DM root while progress bubbles honoured the
thread (2026-07-27 mixed-placement report). Trust the stamp instead; flat
mode is unaffected because flat metadata never carries an anchor.
Slack Block Kit buttons only support style primary (green) / danger (red) /
default (white). The relay approval + slash-confirm prompts emitted an invalid
style 'success' (Slack silently drops it → white/stroke button) and baked
emoji into the labels (non-native). Native Slack Hermes uses plain labels with
primary/danger. Map to valid styles (once→primary, deny/cancel→danger,
session/always→default) and drop the emoji from labels. The connector already
compensates success→primary, but emitting valid values at the source is correct
and removes the fragile dependency on that compensation.
check-windows-footguns (blocking CI) flagged a bare Path.write_text() without
encoding= at scripts/tool_search_livetest2.py:190, which uses the platform
locale encoding on Windows. Pin utf-8. Pre-existing on main; unblocks the
required-checks gate for this PR.
A prompt (approval/clarify) is emitted in reply to the triggering inbound event,
so its metadata carries that event's synthetic DM thread anchor; forwarded to
the connector it threads the Block Kit prompt under the user's message instead
of posting flat at the DM root. Main routes all prompts through the single
_send_prompt prompt-op choke point, so strip the synthetic DM thread anchor
there via _strip_synthetic_dm_thread — preserving real threads (distinct
thread_id), tenant scope (scope_id/slack_team_id), and non-DM/non-Slack chats.
Preserves main's hp1 prompt-codec; no competing ap:/cl: encoding.
On the relay lane a Slack DM's streamed reply was sent with reply_to=the
triggering message ts; the connector maps a raw reply_to to a Slack thread_ts,
so the DM reply posted threaded under the user's message and lost progressive
edit-streaming (flat reply, no thinking status). Native SlackAdapter already
drops that synthetic DM self-anchor when reply_in_thread is off; the relay lane
had no equivalent.
Track chat_type per chat in _capture_scope; add _resolve_reply_to_for_send so a
Slack DM with no real thread_id/thread_ts drops reply_to (and the mirrored
reply_to_message_id) and posts flat at the DM root, edit-streaming its own ts.
Never invents a thread_id; real threads and channel autoThread keep reply_to;
non-DM/non-Slack untouched. Adapted to main's phase-3 prompt architecture.
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.
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.
`/work` typed into a fresh Cmd+T tab loaded the skill in that tab and
printed "⚡ loading skill: work" there, then fired the skill's kickoff
prompt as a user message into whatever conversation was on screen.
The dispatcher resolves its target once, through resolveTargetSessionId,
and every other consumer of that answer already honors it: the output
writer binds to the target's stored id, and the busy gate reads the
target's own state. The send did not — `submitPromptText(message)` passed
no target at all, so submit fell back to `activeSessionIdRef`, which
names the foreground chat. #71805 fixed the two sibling leaks in this
same function; this is the third and the one that actually moved the
user's prompt.
Forward the resolved pair instead. Every target the dispatcher serves —
a tile, a background queue drain, a session this very call created —
was hitting the same fallback, so the fix covers the class rather than
the tab case that surfaced it.
Technical mode rendered the raw payload two different ways — a bare
block for most rows, a native `<details>` for file edits, whose
browser-drawn marker matches nothing else in the app. Both are now one
collapsed chevron disclosure at a smaller type size, with even padding
against the row body.
The strip's rule is an inset shadow painted in the container's last pixel
row, and full-height tabs covered it — so each tab read as overhanging
the bar by 1px. Inactive tabs compensated with their own border, stacking
a second translucent line that darkened the seam.
Inactive tabs now stop 1px short and draw no bottom border, leaving the
container as the sole owner of one continuous rule; the active tab keeps
full height so it alone cuts through. Hover also darkens rather than
lightens, since lightening moved a hovered tab toward the active
surface's look.
The arc reduced to a few faint dots on session rows. Two causes: the ring
was outset by 2px into a scroller that clips horizontally, losing its
left and right runs; and its tail color defaults to the chrome
background, which is invisible against the sidebar, leaving only the
bright stop of each gradient pass.
An `arc-row` variant sits flush and ties the tail back to the ring
color. The ring's radius is now derived from its standoff (r_host + gap)
rather than inherited, which keeps any outset host concentric instead of
pinched.
Content links read as a small primary-tinted chip instead of an
underline, dropping the trailing external-link arrow. The tint is
currentColor-relative, so one class carries text and fill in the same
hue across every theme, and `box-decoration-break: clone` gives a
wrapped link a chip per line fragment.
An agent told to work in a fresh git worktree does exactly that — creates
it, cds in, and runs every later command there — but the session stayed
pinned to the checkout it started in. The desktop kept labelling the chat
with the primary branch while all the work landed somewhere else.
The desktop half already existed: session.info carrying a moved cwd runs
followActiveSessionCwd, which refreshes the project tree and scopes the
sidebar into the new project. The backend just never reported the move.
Reconcile the session's cwd against terminal_tool's per-session record at
the end of a turn, when the agent has stopped moving and its recorded cwd
is a stable answer. A plain cd stays what it always was — not a workspace
move — so the reconcile only fires when the recorded cwd sits in a
different git working tree than the session's workspace.
The sidebar "+" now stacks a tab instead of replacing the surface, so the
prior session stays mounted and several chat surfaces can be on the page at
once. Helpers that waited for the old transcript to disappear from the page
timed out, and `.first()` locators / bare `document.querySelector` calls
started resolving against the wrong session (CI's "resolved to 2 elements"
strict-mode violation).
Target the most recently mounted `[data-composer-target]` surface instead,
and assert the NEW surface is empty rather than waiting for the old text to
vanish.
CuaDriverBackend caches a long-lived cua-driver subprocess for the life of
the Hermes process, and stop() was never called from anywhere — the driver
outlived the session that spawned it. #69903 stopped the orphan from pegging
a core by disabling the cursor overlay, but left the process behind; this is
item 3 of #28152 ("Hermes does not keep the driver alive after tool
completion").
Register an atexit hook, mirroring browser_tool's
atexit.register(_emergency_cleanup_all_sessions). atexit only, no signal
handlers, for the prompt_toolkit reason documented there. reset_backend_for_tests
now reuses the same teardown instead of repeating it.
