* perf: optimize large session reconnect by adding file-size threshold to tail optimization (#6241)
When a sidecar JSON file exceeds 500 KB, the display-path tail optimization
now fires even if the message count is within the raw_budget. This prevents
sessions with few messages but large tool outputs (multi-MB JSON) from
forcing a full-scan merge of all messages on reconnect.
Changes:
- Added _sidecar_file_exceeds_threshold() helper
- Added _SIDECAR_BYTE_TAIL_THRESHOLD = 500_000 constant
- Fall-through to truncation in _state_db_since_timestamp_for_limited_display
when the sidecar file exceeds the threshold, regardless of message count
* CHANGELOG: byte-size reconnect tail optimization (#6260)
---------
Co-authored-by: webtecnica <webtecnica@users.noreply.github.com>
Co-authored-by: nesquena-hermes <agent@nesquena-hermes>
* fix(transparent-stream): remove dead rowIndex param from pushRow (#6189 follow-up)
In #6189 / #6217 the final-segment eligibility was migrated from an
index comparison (rowIndex > lastNonTerminalWorkRowIndex) to a WeakSet
lookup (finalSegmentLiveProseRows.has(row)). The rowIndex parameter on
pushRow became dead code — it's declared but never read, and the
call-site still passes idx from forEach. Remove the unused parameter
and simplify the call-site.
This addresses the greptile review feedback on the original PR.
Closes#6189
* CHANGELOG: dead rowIndex param removal (#6258)
---------
Co-authored-by: webtecnica <webtecnica@users.noreply.github.com>
Co-authored-by: nesquena-hermes <agent@nesquena-hermes>
* fix: split OIDC allowlist on commas only, preserve scope whitespace-split (#6244)
_normalize_text_list is shared with _normalize_scopes — OAuth scopes
are space-delimited per RFC 6749 §3.3. Created _normalize_allow_values
that splits on commas/newlines only, keeping multi-word group names
like 'Hermes Users' intact.
* fix(oidc): filter blank allow_values list elements + add parser-split test (gate follow-up)
The new comma/newline-only _normalize_allow_values() list-path retained empty
strings that the shared _normalize_text_list() had filtered, so a YAML
allow_values: [""] would brick an OIDC-only deployment (every callback 403s).
Filter stripped-empty collection elements. Adds a regression test asserting
allowlist multi-word preservation, comma/newline splitting, blank filtering,
and that scopes stay space-delimited (RFC 6749 §3.3).
Co-authored-by: webtecnica <webtecnica@users.noreply.github.com>
* CHANGELOG: OIDC allowlist whitespace fix (#6244)
---------
Co-authored-by: webtecnica <webtecnica@users.noreply.github.com>
Co-authored-by: nesquena-hermes <agent@nesquena-hermes>
* fix: intercept /sessions and /resume slash commands in WebUI (#6224)
Add a native-intercept branch in the block of
send() alongside the /pet special-case. When the user types /sessions
or /resume, expand the sidebar and refresh the session list instead of
sending the raw slash text to the agent.
Root cause: the agent command registry exposes sessions/resume as
non-CLI-only commands, so the autocomplete popup shows them, but the
WebUI send-time dispatch had no branch to catch them, causing the
literal text to be sent as a prompt.
* fix(commands): use mobile-aware session-browser opener for /sessions /resume (gate follow-up)
The intercept called expandSidebar() directly, which is a no-op on phone-width
layouts, so /sessions and /resume silently did nothing on mobile (composer
cleared, nothing shown). Use the mobile-aware _openProfileSwitchSessionBrowser()
first, falling back to expandSidebar(). Reproduced + specified by the pre-release
Codex gate.
Co-authored-by: webtecnica <webtecnica@users.noreply.github.com>
* CHANGELOG: intercept /sessions /resume slash commands (#6245)
---------
Co-authored-by: webtecnica <webtecnica@users.noreply.github.com>
Co-authored-by: nesquena-hermes <agent@nesquena-hermes>
* fix(reasoning): gate reasoning_effort ladder to GLM-5.2+ on native zai
Z.AI's API (docs.z.ai) defines reasoning_effort as GLM-5.2+ exclusive, but
hemes-webui advertised the full 6-level ladder for all 7 GLM models because
_candidate_supports_reasoning has an unconditional glm token match and
_filter_reasoning_efforts_for_provider had no ZAI branch. Six of seven catalog
models (glm-5.1, glm-5, glm-5-turbo, glm-4.7, glm-4.5, glm-4.5-flash) showed a
selector whose values the endpoint silently ignores, and GLM-4.7 (forced thinking
that cannot be disabled per Z.AI docs) showed a 'none' option with no effect.
Add a ZAI branch to _filter_reasoning_efforts_for_provider mirroring the existing
OpenAI/Gemini/Anthropic ceiling pattern: strip the whole ladder for pre-5.2 GLM
models and for the forced-thinking GLM-4.7 family; preserve the full ladder for
GLM-5.2+ (whose accepted values match VALID_REASONING_EFFORTS exactly). The gate
is scoped to the native zai provider only (aliases glm/z-ai/z.ai/zhipu all
resolve to zai); aggregator providers are untouched because they route through
their own routers, not Z.AI's native endpoint.
The glm family-detection heuristic in _candidate_supports_reasoning is unchanged
— GLM models DO support the thinking on/off toggle at the family level; this fix
is specifically about the reasoning_effort intensity ladder.
State layer: agent.reasoning_effort config + UI dropdown options derived from
resolve_model_reasoning_efforts. Invariant: UI options and coercion now agree
and match Z.AI's per-model docs (max offered only for GLM-5.2+, none never
offered for forced-thinking models). Out of scope: the thinking:{type:...}
request-field translation lives in the external agent/gateway layer.
* fix(reasoning): close ZAI coercion gap + harden test assertions
Address Greptile review on #6219 (round 1):
1. Vacuous test assertion (test_glm_5_2_preserves_none_sentinel): the 'or'
fallback made the assertion always-true, so a regression stripping 'none'
for GLM-5.2 would go undetected. Rewrote to inject 'none' via the raw source
(mocking _resolve_model_reasoning_efforts_impl) and assert it survives — the
test now genuinely exercises the preservation branch.
2. Coercion gap for non-max stored levels on pre-5.2 GLM: the existing
'if ceiling and raw not in ceiling' guard treats an empty ceiling as 'no
rule' (preserving the configured effort verbatim per #3505), so a stored
'high'/'medium'/'low' for glm-5.1/glm-4.5/glm-4.7 on native zai was forwarded
to Z.AI unchanged and silently ignored — contradicting the PR's stated
UI/coercion agreement invariant.
Root cause: the ZAI gate returns [] to mean 'known-empty' (no
reasoning_effort at all), but the coercion path treated all [] as
'ambiguous/unknown, preserve verbatim'. Fixed by extracting the ZAI decision
into _zai_glm_reasoning_efforts_supported (True/False/None sentinel) shared by
both the filter and coercion, then special-casing the known-False result in
coerce_reasoning_effort_for_model to return '' (send no field). The #3505
preserve-verbatim behavior for genuinely-unknown models on non-zai providers
is unchanged.
Added 10 regression tests (all fail before the coercion fix, pass after):
- All 6 levels (max..minimal) coerce to '' for each pre-5.2 GLM + glm-4.7
- All 4 aliases (glm/z-ai/z.ai/zhipu) resolve through the same coercion gate
- GLM-5.2 preserves all 6 levels verbatim
- Regression guard: unknown model on custom: provider STILL preserves verbatim
State layer: agent.reasoning_effort config + the value forwarded to Z.AI.
Invariant now fully holds: UI offers no options for pre-5.2 GLM AND coercion
sends no reasoning_effort field for any stored level on those models.
* fix(reasoning): preserve ZAI GLM 4.5-5.1 thinking toggle when effort ladder empty
Address nesquena-hermes round-2 review on #6219: returning [] for the effort
ladder on sub-5.2 GLM models hid the entire reasoning chip in the composer
(static/ui.js:4932 treats empty supported_efforts as 'no reasoning control at
all'), silently regressing the working thinking on/off toggle for GLM-4.5/4.6/
5.0/5.1 users. Per Z.AI's own docs, those models accept the thinking
{type:enabled|disabled} toggle even though they do not accept the
reasoning_effort intensity ladder.
Fix decouples thinking-toggle capability from the effort ladder:
Backend (api/config.py):
- Refactor _zai_glm_classification returns one of 'effort' (GLM-5.2+), 'thinking'
(GLM-4.5 up to but not including 5.2), 'forced' (GLM-4.7 family), or None
(non-zai / non-GLM). Single source of truth shared by all three consumers.
- _zai_glm_reasoning_efforts_supported now wraps classification for the coercion
contract (unchanged behavior).
- _zai_glm_thinking_toggle_supported returns True for 'effort' or 'thinking',
False for 'forced', None otherwise.
- get_reasoning_status gains a supports_thinking_toggle field = bool(supported)
OR (zai_thinking is True). Non-zai providers default to bool(supported_efforts)
so their chip-visibility behavior is unchanged.
Frontend (static/ui.js):
- New _currentReasoningToggleSupported state var (default undefined = treat as
true so legacy responses without the field do not newly hide the chip).
- _applyReasoningChip shows the chip when hasEffortLadder OR toggleSupported.
Empty efforts + toggle=True keeps the chip visible with just the None/On
control (the existing _applyReasoningOptions already shows 'none' when the
ladder is empty). Empty efforts + toggle=False (GLM-4.7 forced) hides it.
- Profile-transition and fetch-failure resets now pass
supports_thinking_toggle:false alongside the empty efforts so the chip hides
during the unknown-state window, matching the prior reset contract.
Tests (30 new):
- _zai_glm_classification parametrized across all three tiers + aliases + defer
- get_reasoning_status supports_thinking_toggle per tier (GLM-5.2 both, GLM-4.6
toggle-only, GLM-4.7 neither, non-zai defaults to effort capability)
- Frontend _applyReasoningChip behavior via node driver: empty efforts +
toggle=True stays visible, toggle=False hides, effort ladder alone is
sufficient, absent field keeps prior behavior
State layer: agent.reasoning_effort config + supports_thinking_toggle field in
/api/reasoning + composer chip visibility. Invariant: the chip is hidden ONLY
when the model supports neither the effort ladder nor the thinking toggle
(GLM-4.7 forced, or genuinely non-reasoning models); GLM-4.5-5.1 retain the
working On/None control they had before the round-1 effort gate.
* fix(reasoning): make ZAI thinking toggle two-way + force GLM-4.7 stored none
Address nesquena-hermes round-3 review on #6219 — two SILENT gaps in the
thinking-toggle path, plus a click-handler sibling I found while auditing.
Gap #1 — ONE-WAY toggle for GLM-4.5/4.6/5.0/5.1 (api/config.py:4247,
static/ui.js:4902, static/index.html:757):
The round-2 fix kept the chip visible for thinking-tier models but the only
rendered dropdown option was 'None' (the HTML had no Default option, and
set_reasoning_effort rejected empty effort with 400). So a GLM-4.6 user could
turn thinking OFF but never back ON — worse than the original bug.
Fix:
- set_reasoning_effort now accepts empty effort as 'clear the override' (removes
agent.reasoning_effort so the provider default takes effect). Invalid values
still raise ValueError.
- static/index.html gains a <div data-effort=''>Default</div> option.
- _applyReasoningOptions always shows both Default ('') and None alongside the
effort ladder, so a thinking-tier model (empty ladder + toggle=true) renders
an operable Default+None two-state control.
- Click handler (ui.js:5106) checks option presence (if(opt)) not truthiness
(if(effort)) — the old check silently ignored data-effort='' clicks, which
would have left the Default button dead even after the HTML/backend changes.
Gap #2 — GLM-4.7 not forced when 'none' stored (api/config.py:3984, :3857):
When GLM-4.7 had agent.reasoning_effort=none configured, coercion preserved
'none' via the early return at line 3985, so streaming built disabled reasoning
for a model that forces thinking on regardless. Separately, when the raw
capability source listed 'none', resolve_model_reasoning_efforts reattached it
to GLM-4.7's supported_efforts (['none']), leaking an 'off' option to the UI
for a forced-thinking model.
Fix:
- coerce_reasoning_effort_for_model checks _zai_glm_classification == 'forced'
BEFORE the generic 'none' early-return, coercing stored 'none' to '' (default
= thinking on) for forced models.
- resolve_model_reasoning_efforts returns [] early for forced-tier models,
skipping the 'none' reattachment entirely.
Tests (17 new):
- Gap #2: coerce('none', glm-4.7) -> '', regression guard that non-forced GLM
still accepts 'none'; resolve does not reattach 'none' for forced but DOES
for thinking-tier; end-to-end get_reasoning_status for forced+stored-none.
- Gap #1 backend: set('') clears the key (no raise), set('garbage') still
raises, all 7 valid levels still save.
- Gap #1 frontend: three new node-driver tests asserting the dropdown exposes
both Default and None for thinking-tier (two-state), Default+None+ladder for
effort-tier, and the off->on->off round trip keeps both visible throughout.
- Updated test_reasoning_show_hide.test_set_reasoning_effort_rejects_invalid to
reflect the new contract (empty accepted, garbage rejected).
Regression gate: gap#2 coerce tests fail without the forced-tier check (3
failures); gap#1 backend test fails with 'ValueError: effort is required'
without the empty-acceptance change. 219 passed, 1 pre-existing skip across
the full reasoning + chip + config-cache surface.
State layer: agent.reasoning_effort config + supports_thinking_toggle field +
composer dropdown options. Invariant: the thinking toggle is now genuinely
two-way for GLM-4.5-5.1 (Default=on, None=off, both always visible) and
GLM-4.7 forced-thinking never offers 'none' in any path (coerce, resolve,
status, or UI).
* CHANGELOG: GLM per-version reasoning controls (#6219)
---------
Co-authored-by: Ruby Hartono <58564005+rh-id@users.noreply.github.com>
Co-authored-by: nesquena-hermes <agent@nesquena-hermes>
* fix(ui): resolve folder download URL against document.baseURI for subpath support
When Hermes WebUI is served behind a reverse proxy with a path prefix
(e.g. /hermes/), the right-click → "Download Folder" context menu option
navigates to a root-absolute URL (/api/folder/download?...), which resolves
to the server origin instead of the proxy mount point, causing a 404.
This matches the pattern already used by the workspace.js route helper
refactored in v0.52.41 (commit 1a64d7d3).
* CHANGELOG: folder-download subpath baseURI fix (#6227)
---------
Co-authored-by: Steezy <21984836+steezypunk@users.noreply.github.com>
Co-authored-by: nesquena-hermes <agent@nesquena-hermes>
* feat(core): add registerHermesSessionOpenHandler + renderTranscript extension hooks
- registerHermesSessionOpenHandler(fn): register a handler that fires on
session open. Return {cancel:true} to prevent navigation.
- renderTranscript(container, messages, opts): render messages into any
DOM container using core's renderMd pipeline. Skip tool messages.
- Wire _hermesNotifySessionOpen into loadSession: pre-load guard at top,
post-load notification at end for extensions to hook into.
- Follows existing registerHermesTtsEngine extension registration pattern.
This gives extensions like chat-tiling a sanctioned API instead of DOM
hacking to intercept and render session transcripts.
* fix: address 3 gate-fail blockers from PR #5508 review
Fixes the three core issues identified by nesquena-hermes code review:
1. XSS sink in renderTranscript (boot.js)
- Fallback to textContent when window.renderMd is unavailable
- innerHTML exclusively for successful renderMd output
2. Stale _loadingSessionId reset in cancel branch (sessions.js)
- Remove premature nulling of _loadingSessionId
- Cancel path simply returns without touching loading-guard state
3. Pre-open veto bypasses profile/import side-effects (sessions.js)
- Move cancellable preload hook to start of _openSidebarSession,
before external session import and profile switching
- Pass internal _preloadNotified flag to skip duplicate preload
in loadSession while retaining post-load notification
Closes#5508 gate-fail items.
* fix: use module-level flag instead of call argument to avoid test regression
The _preloadNotified approach broke test_static_sessions_js_switches_profile
before_opening_all_profiles_row because it changed the loadSession call
signature from loadSession(sid, loadOpts) to loadSession(sid, Object.assign(...)).
Switch to a module-scoped boolean _hermesSessionOpenAlreadyFired set by
_openSidebarSession before calling loadSession, checked by loadSession's
pre-hook guard. The call signature stays unchanged.
Test 2 (test_load_session_rearms_stream_on_every_early_return) also passes.
* fix: compact pre-hook comment to keep loadSession within test window limits
* fix: replace global _hermesSessionOpenAlreadyFired flag with per-call opts._preloadNotified
The module-level boolean introduced in 381fa0ef had two problems:
1. ReferenceError on direct loadSession() calls — the flag was undeclared
when called outside _openSidebarSession, breaking saved-session restore.
2. Never reset after first sidebar open — all subsequent direct calls
silently skipped the cancellable preload handler.
Per the maintainer's review (PR #5508), replace the global with a per-call
option _preloadNotified passed by _openSidebarSession via Object.assign.
This keeps the call signature stable for existing tests and eliminates
the stale global state leak.
Test adjustments:
- test_issue1611: widened loadSession call literal assertion
- test_session_channel_option_x: body slice 14000→15000 to accommodate
the slightly longer function body
* chore: remove unrelated files from commit
* fix: address 3 review issues — preload-only cancel, drop inner wrapper, pass _preloadNotified on retry
- Only honor {cancel:true} when opts.preload===true (boot.js)
- Drop .msg-body-inner wrapper, render directly into .msg-body (boot.js)
- Carry _preloadNotified:true through cross-profile 409 retry (sessions.js)
* recommit
* Update static/boot.js
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
* fix(extensions): resolve canonical sid before preload hook + pass _preloadNotified on continuation retry
PR #5508 review follow-up (review 4690636760):
1. Move _resolveSessionIdFromSidebarLineage() before the preload hook so
extensions always see the canonical sid, not the raw sidebar click id.
2. Pass _preloadNotified:true on the continuation-session retry path to
prevent duplicate preload events to extensions.
3. Add functional test_extension_session_hooks.py — actually drives the
new hook registration, preload-veto, transcript rendering, and
_preloadNotified bridge in Node (13 tests, all green).
* Remove unused pytest import
Removed unused import of pytest from test file.
* fix: address 2 gate-blocking veto-ordering defects (PR #5508)
Blocker 1 (CORE): cross-profile retry now passes _preloadNotified:true
so the pre-hook doesn't re-fire after destructive side-effects already
ran (stream teardown, message clear, profile switch). A {cancel:true}
on that second fire was stranding the UI profile-switched with a
cleared transcript.
Blocker 2 (SILENT): closeMobileSidebar() was called synchronously
BEFORE _openSidebarSession()'s veto guard in three places (tap-to-open,
child-session, lineage-segment). A {cancel:true} still closed the
sidebar out from under it. Removed the three premature calls; moved
a single closeMobileSidebar() inside _openSidebarSession AFTER the
veto guard so it only runs when the open actually proceeds.
Added 3 regression tests asserting {cancel:true} leaves NO side-effect.
* CHANGELOG: extension session-open handler + renderTranscript API (#5508)
---------
Co-authored-by: Sean <seanos1a@gmail.com>
Co-authored-by: ChonSong <85378550+ChonSong@users.noreply.github.com>
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
Co-authored-by: nesquena-hermes <agent@nesquena-hermes>
* fix(transparent-stream): drop stale final-answer prefix row in multi-segment settle (#5749 follow-up)
A turn with interim assistant messages (prose interleaved with tool calls)
could show the beginning of the final answer TWICE after watching it stream:
once as a settled anchor-scene prose row (the live-token accumulator's last
throttled snapshot) and once as the real assistant segment. The duplicate
persisted until reload.
Root cause: #5758 suppresses the accumulator row only when it sits "after the
last tool row", but _completeSettledAnchorSceneForTurn appends the settled
per-message tool rows AFTER the projected live rows — those re-list tools that
ran EARLIER in the turn, pushing the boundary past the final segment's
accumulator so the guard never fired. The stale prefix snapshot then survived
into the persisted scene and rendered above the settled answer.
Fix: judge final-segment eligibility against the LIVE projection's own
chronology — a live-prose row belongs to the final segment iff no PROJECTED
tool row follows it. Pre-tool narration that happens to prefix the final
answer stays protected (existing #5758 regression tests still pass), and a
new regression test pins the multi-segment shape.
Verified end-to-end by replaying the captured run journal of an affected
session through the real SSE live handlers in headless Chromium: duplicate
before (answer prefix visible twice after settle), gone after; fresh-reload
rendering unchanged.
Rollback: revert this commit; behavior returns to pre-fix (duplicate prefix
row after live-settle of multi-segment turns).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* CHANGELOG: transparent-stream multi-segment prefix dedupe (#6189, @ai-ag2026)
---------
Co-authored-by: ai-ag2026 <m.fuechtenkoetter@posteo.de>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: nesquena-hermes <agent@nesquena-hermes>
* feat: expose msg_limit ceiling via /api/session metadata, frontend reads dynamically (#6177)
Backend exposes _MAX_MSG_LIMIT as _msg_limit_max in every /api/session
response. Frontend reads it dynamically, falling back to _MSG_LIMIT_MAX
for older servers. This removes the hand-mirrored coupling between the
two layers.
Removed test_msg_limit_ceiling_drift.py since the mirror pattern that
required the drift guard is replaced by dynamic metadata.
This is the standalone metadata-decoupling piece from #6206,
without the clamp/paging that already shipped in exp-v0.52.98 via #6152/#6154.
* fix(session): declare _msgLimitMax at module scope + CHANGELOG + gate fixes (#6214 follow-up)
The submitted PR used _msgLimitMax at two read sites (boundedReloadLimit in
_ensureMessagesLoaded, useBeforePaging in _loadOlderMessages) but never declared
it and read it before assignment -> undefined on cold load -> full-transcript
fetch every load (regression) + implicit global. Declared `let _msgLimitMax =
_MSG_LIMIT_MAX;` at module scope so the reload-width paths always read a defined
value (the static fallback) until the server's _msg_limit_max lands.
Also: defined the ceiling globals in the two inline node-harness tests that copy
_ensureMessagesLoaded's body (test_cross_session_message_load_isolation,
test_session_unread_dot_on_visit), updated the source-string assertion in
test_webui_external_refresh_frontend, and added 3 decoupling tests (backend field
present, module-scope declaration + fallback, both paths read the live ceiling)
replacing the deleted drift-guard.
Gate: Codex adversarial SAFE TO SHIP (executable probes: cold-load fallback,
mixed-version omission, live update, cross-session isolation, over-ceiling bare
refresh, msg_before row preservation). Full sharded suite green.
Co-authored-by: webtecnica <webtecnica@users.noreply.github.com>
---------
Co-authored-by: webtecnica <webtecnica@users.noreply.github.com>
Co-authored-by: nesquena-hermes <agent@nesquena-hermes>
* fix(routes): clamp GET /api/session msg_limit to a server-side ceiling
Memory: ?msg_limit= had no upper bound — a client could request
msg_limit=1000000 (or the frontend's msg_limit=9999 outline-jump path) and
force the server to assemble and serialize an unbounded message payload. The
frontend's real pagination grows windows by ~30 at a time (initial load 30,
load-older +30, msg_before paging 30), so only the outline-jump and
pathological/hostile cases ask for huge values.
Add _MAX_MSG_LIMIT (500) and clamp at the query-parse site
(max(1, min(int(_msg_limit), _MAX_MSG_LIMIT))). The ceiling is generous — far
above any legitimate visible-row window — so real pagination is unaffected. The
existing _messages_truncated signal already covers the clamped case: when the
request exceeds the ceiling, _message_window_for_display returns
_messages_offset > 0 and the client learns more rows exist (load-older still
works). The bare no-msg_limit path is intentionally unchanged — it is the
documented "give me everything" escape hatch used by branch/undo/jump-to-start
flows that need the full transcript for correct global indices.
Test: tests/test_session_msg_limit_ceiling.py — the ceiling constant is
reasonable (100..2000), an over-ceiling request is clamped and signals
truncation, and legit sizes (5/30/60/100) are unaltered. Existing window/tail
tests (19) unchanged.
Contract Routing
- Task type: memory-bound fix (unbounded msg_limit request payload).
- Touched areas: GET /api/session query parsing (api/routes.py).
- Relevant public docs: AGENTS.md, CONTRIBUTING.md, docs/CONTRACTS.md.
- State layer mutated: none (read-only endpoint response shaping).
- Contract change (additive): ?msg_limit=N is now clamped to _MAX_MSG_LIMIT=500.
Compatibility: the only frontend caller above the ceiling is outline.js's
msg_limit=9999, which now receives up to 500 rows + _messages_truncated=true
(load-older still works). All other callers (30/60) are unaffected. The bare
no-msg_limit path is unchanged.
Release note: GET /api/session now caps ?msg_limit at 500 rows server-side,
bounding the worst-case payload for oversized requests.
Model Used: builtin:zai-coding-plan/GLM-5.2 (ZCode agent).
* fix(frontend): route msg_limit-heavy paths around the server ceiling
PR #6152 clamps GET /api/session ?msg_limit= to a max of 500 server-side. Two
frontend paths relied on unbounded msg_limit and silently broke for sessions
> 500 visible rows:
1. outline.js _jumpToMessage sent msg_limit=9999 as a "give me everything" hack
so it could scroll to any message by absolute index (msg-user-<rawIdx>).
Clamped to 500, the target row's DOM id doesn't exist for early messages in
long sessions, and the jump silently no-ops (line 119 `if (r)` is falsy).
Fix: drop msg_limit=9999 — use the bare messages=1 path, which #6152 left
unbounded precisely for callers that need the full transcript. This restores
exact pre-#6152 behavior for outline jumps.
2. sessions.js _loadOlderMessages grows msg_limit by +_INITIAL_MSG_LIMIT each
load (30 -> 60 -> ... -> 500 -> clamped). Past 500 the server returns the
same clamped tail, olderMsgs shrinks to 0, and endless-scroll / "Load
earlier" silently stalls — the head becomes unreachable. Fix: once the grown
requestedLimit meets/exceeds the new _MSG_LIMIT_MAX mirror (500), switch the
primary fetch to msg_before paging (a fixed _INITIAL_MSG_LIMIT backward page
keyed off _oldestIdx) — the same bounded paging the existing race-fallback
uses, so the head stays reachable for arbitrarily long transcripts. The
response is handled by the existing prepend path (tailMatches is naturally
false for a page, so the !tailMatches branch reuses the fetched data).
Added _MSG_LIMIT_MAX = 500 constant in sessions.js mirroring backend
_MAX_MSG_LIMIT, documented as kept-in-sync.
Verification: node --check passes on both files; the backend window-contract
tests (test_session_tail_payload, test_session_message_window_renderable_tail,
19 tests) still pass. No JS test harness exists in this repo (vanilla JS,
manual browser verification per TESTING.md) — manual verification needed: open a
> 500-message session, (a) click an outline entry for an early message and
confirm it scrolls+flashes, (b) scroll up past 500 loaded rows and confirm
"Load earlier" keeps producing older messages without stalling.
Contract Routing
- Task type: frontend follow-up to a backend memory fix (#6152).
- Touched areas: outline jump-to-message (static/outline.js), endless-scroll /
load-older (static/sessions.js _loadOlderMessages).
- Relevant public docs: AGENTS.md, CONTRIBUTING.md, docs/CONTRACTS.md.
- State layer mutated: client-side S.messages (the active-session transcript),
same as the existing load-older path. No backend state change.
- Scope boundaries: does not weaken the #6152 clamp. The bare messages=1 path
(outline) was intentionally left unbounded by #6152; the msg_before paging
path (load-older) is bounded by _INITIAL_MSG_LIMIT and never hits the ceiling.
Release note: fixes outline jump-to-message and endless scroll silently
breaking on sessions longer than 500 messages after the server-side msg_limit
ceiling (companion to #6152).
Model Used: builtin:zai-coding-plan/GLM-5.2 (ZCode agent).
* test(frontend): update static-source assertion for refactored _loadOlderMessages fetch
The static-source test test_session_message_loads_keep_explicit_longer_timeouts
asserted the exact multi-line shape of the _loadOlderMessages fetch strings.
The #6154 refactor wrapped them in a useBeforePaging ternary (tail-growth vs
msg_before paging), reindenting them — the contract the test guards (both
fetches keep timeoutMs:120000 for large state.db loads) is preserved, but the
literal source shape changed, so the assertion failed on shard 0.
Updated the assertions to match the new source shape while keeping the test's
intent: assert each fetch URL + its {timeoutMs:120000} survive in sessions.js.
Model Used: builtin:zai-coding-plan/GLM-5.2 (ZCode agent).
* test(msg-limit): stopgap drift guard for _MSG_LIMIT_MAX === _MAX_MSG_LIMIT
Reviewer guidance (#6154 round-2): the stopgap drift-test belongs in THIS PR
(#6152) because _MAX_MSG_LIMIT is defined here. The frontend _MSG_LIMIT_MAX
mirror lives on #6154 (not master yet), so the assertion SKIPS when the JS
constant is absent and activates automatically once both branches land on master.
The guard: extract _MAX_MSG_LIMIT from api/routes.py and _MSG_LIMIT_MAX from
static/sessions.js via regex; assert equality when both present. If they drift,
load-older silently stalls at the wrong value (the exact regression #6154 fixed).
Retired by #6177 (metadata-exposure), which removes the mirror entirely.
Verified: on this branch the cross-check SKIPS (_MSG_LIMIT_MAX absent, as
expected — #6154 not landed) and the backend-existence test PASSES. Extraction
logic confirmed correct against simulated match (500==500 passes) and drift
(999!=500 fails) cases.
Model Used: builtin:zai-coding-plan/GLM-5.2 (ZCode agent).
* fix(session): prevent silent row-loss in msg_limit ceiling paging (gate follow-up)
Adversarial review of #6152/#6154 reproduced two silent row-loss regressions in
static/sessions.js; both fixed here, plus a stale source-string assertion update
and the CHANGELOG entry:
1. _loadOlderMessages: gate the suffix-continuity heuristic on !useBeforePaging so
a raw-row-heavy msg_before cursor page whose visible text repeats the current
tail can't be misclassified as the cumulative tail and wholesale-replace it
(previously dropped 470 of 940 rows). msg_before pages are always prepended.
2. loadSession same-session refresh: only apply the width-hint msg_limit when it
is <= _MSG_LIMIT_MAX; above the ceiling the backend would clamp and silently
shrink an already-loaded >500-row transcript to the last 500. Over-ceiling
reloads now use the bare full-transcript path (no msg_limit / no expand_renderable).
3. tests/test_webui_external_refresh_frontend.py: update the reloadLimitParam
source assertion to the boundedReloadLimit form.
Re-gate: Codex adversarial SAFE TO SHIP (940/940 rows retained on cursor paging,
1000/1000 on refresh, no seam duplication, small sessions still use the fast tail
window); targeted + sessions/pagination/regression suites green.
Co-authored-by: Ruby Hartono <rh-id@users.noreply.github.com>
* test: fix two node-harness tests broken by the boundedReloadLimit change
- test_issue3162_ensure_messages_loaded: replace the brittle fixed 4500-char
window (which the boundedReloadLimit lines pushed the carry-forward past) with
robust brace-balance function-body extraction.
- test_cross_session_message_load_isolation: define _MSG_LIMIT_MAX=500 in the node
harness env; _ensureMessagesLoaded now references it, and an undefined value made
boundedReloadLimit=null -> the fetch URL dropped msg_limit/expand_renderable and
mismatched the ordered api() harness -> stall/timeout.
---------
Co-authored-by: Ruby Hartono <58564005+rh-id@users.noreply.github.com>
Co-authored-by: nesquena-hermes <agent@nesquena-hermes>
Co-authored-by: Ruby Hartono <rh-id@users.noreply.github.com>
Centralize custom-provider model-ID extraction into _configured_model_ids()
+ _configured_model_options(), used by both catalog construction and runtime
routing so the picker and resolver can't drift. Fixes list-shaped
custom_providers[].models allowlists silently staying on the default provider
instead of routing to their own (#6121). Existing mapping / list-of-strings
configs unaffected.
Co-authored-by: nesquena-hermes <agent@nesquena-hermes>
Co-authored-by: franksong2702 <franksong2702@users.noreply.github.com>
Reconcile the gateway pending-mirror in _session_attention_summary()
before counting sidebar approval attention, so a parent session's red
approval dot clears once the underlying gateway approval has drained.
Live approvals stay visible; clarify fallback unchanged. WebUI read-path
mitigation; child-session context-propagation root cause stays in #6100.
Co-authored-by: nesquena-hermes <agent@nesquena-hermes>
Co-authored-by: rodboev <rodboev@users.noreply.github.com>
Scoped to the two stated fixes only: strips the #composerModelDropdown position:fixed
mobile override that #6104's branch had bundled in (that's the #6080 model-picker
mechanism, being reworked separately to reparent to body level — the fixed override
is trapped by .composer-footer's container-type so it doesn't belong here), and drops
the <=360px tool-name-hiding rules (header overflow-clip prevents the spill without
losing tool identity). Thanks @webtecnica.
Cuts the promoted range's bullets from [Unreleased] into a stamped
## [v0.52.28] section (41 bullets, v0.52.0..v0.52.28). Companion to the
v0.52.28 stable tag pushed by the promotion engine.
Co-authored-by: nesquena-hermes <nesquena-hermes@users.noreply.github.com>
Measure-only TTFT (ttft_ms) on the metering session + _firstTokenMs diagnostic. Computed once
on first token (no hot-path latency), all mutation under existing GlobalMeter._lock, existing
token/usage/cost metering unchanged. Codex SAFE (16-stream concurrency probe, no leak/None-deref)
+ stream-gate GREEN all modes + 22 tests.
Local (non-gateway) _resolve_approval_legacy collapsed 'once'+'session' into approve_session()
→ 'approve once' silently persisted session-wide. Narrowed to choice=='session' only; 'once'
still resolves the current call but doesn't persist. Codex SAFE (adversarial: once resolves
current+no residue+next re-prompts, session/always/deny unregressed, gateway consistent, tests
non-vacuous + parked-thread probe of all 4 choices).
Fail-soft git worktree unlock before remove (agent locks worktrees it creates; git refuses
--force remove on locked → orphan accumulation). Dirty-guard still runs first. Codex SAFE,
13 tests pass. Shipped solo (unbundled from #6013 which has a full-suite test-interaction).
MEDIA tokens (image/video/audio/pdf/diff/csv/excalidraw) render inline in real time
during streaming incl split-across-chunk markers. Codex SAFE r4 (both prose regressions
closed — ownership preserved + fade retained, normal prose byte-identical to master, all
4 original findings fixed, security clean, no ReDoS), stream-gate GREEN all modes (hide_all
RED was load flake, isolated re-run clean). 227 tests + 18 real-SMD subtests pass. 4-round
convergence on the crown-jewel render + security path.
Gateway terminal provider errors classified+surfaced (not generic empty turn), session_id-bound,
pending turn persisted through reload incl repeated-prompt (is_exact_checkpoint). Codex SAFE r4 +
stream-gate GREEN all modes. Shipped solo (unbundled from #6002 which has a test-isolation CI fail).
Past committed messages fork during a live response; only the streaming message +
just-sent uncommitted prompt are blocked (toast). Codex SAFE round2 (off-by-one fixed,
_isLastMsg busy-gated, _pending not carried forward), stream-gate GREEN (hide_all RED
was load flake, isolated re-run clean), 99 tests.
The earlier #5979 fix (PR #5980, exp-v0.52.53) resolved the WARM-catalog and
config-declared cases, but b3nw confirmed truncation persists on exp-v0.52.55/56.
Root cause (reproduced on master, confirmed by Codex + Fable):
b3nw's config is provider: custom:llm-proxy, model.default: x-ai/grok-4.5, no
models[] allowlist. His DEFAULT (x-ai/grok-4.5) preserves via config-declaration,
but a model he SELECTS in the picker that isn't the default (x-ai/grok-composer-2.5-fast)
truncates on every COLD send. The send path (api/streaming.py:7857) calls
resolve_model_provider without warming the catalog, and the #1855 chat/start fast
path deliberately skips the catalog when a session already has a persisted
model+provider -- so 'cold at send' is the designed state after any restart /
settings-save / cache-expiry, not a rare race. With provenance None and no config
declaration, the #5980 tri-state fell through to the LEGACY family strip
(if prefix in _PROVIDER_MODELS and _is_first_party_model: return bare), and
grok-composer-2.5-fast had graduated into the x-ai first-party catalog -> stripped.
Fix (two changes, both advisor-verified):
1. Flip the custom-arm cold default from family-strip to PRESERVE-verbatim
(api/config.py). For an explicit custom / custom:<slug> proxy the strip now
requires POSITIVE provenance (endpoint advertises only the bare id, #433). A
wrong strip destroys routing info the user can't recover; a wrong preserve
fails loudly and self-heals on warm. This also honours _endpoint_advertised_model_ids's
own documented 'None -> preserve verbatim' contract and removes the data-driven
flaw where a model graduating into the static catalog silently flips routing.
2. Network-free provenance warm on the send path
(warm_models_catalog_provenance_if_cold, called in streaming.py before the
resolve). prefer_cache=True republishes provenance from the durable disk cache
(no live probe, no _cfg_lock held on the worker thread -> no lock-order
deadlock), so #433's bare-only-strip stays exact whenever a disk cache exists.
Residual (accepted, strictly better than master): a #433-shaped wrong-preserve
now requires memory cold AND disk cache absent/invalid (first run, post-config-save
window, schema/version bump). It fails loudly and heals on the next /api/models or
session-visit rebuild. Under master b3nw's failure was the mirror image but silent,
recurring every send, and unfixable short of declaring every model in config.
Supersedes rodboev's #5996, which targeted the providers.<slug> arm and does not
cover b3nw's actual custom:llm-proxy config (confirmed by Codex).
Tests: rewrite the cold-fallback test (strip -> preserve), sharpen the
foreign-profile test so trusted-strip vs ignored-preserve genuinely diverge, add
b3nw's exact cold non-declared custom:slug case, and add three warm-helper tests
(publishes-from-disk, noop-when-warm, never-live-rebuilds). 64 resolver+sprint40
tests pass; b3nw end-to-end verified; #433 positive-provenance strip verified.
Thanks @b3nw for the CLI-vs-WebUI comparison that localized it.
Final-answer detection understands structured content parts (text after tool-use
boundaries); genuine empty completions still show no_response; terminal provider
error still surfaces as apperror (never masked). Codex SAFE round2 (3 precedence
findings fixed), stream-gate GREEN all modes. Maintainer: updated the #5940 source-
assert test for the _captured_terminal_failure rename.
* test: stabilize cross-platform WebUI suite
* Release exp-v0.52.54: stabilize cross-platform WebUI test suite (#5970)
---------
Co-authored-by: Taha Bakhit <git@tahabakhit.com>
Co-authored-by: nesquena-hermes <agent@nesquena-hermes>
Resolve the model id for custom/proxy providers by ENDPOINT PROVENANCE
instead of a catalog-family guess. The old gate stripped a vendor prefix
whenever the bare id was first-party of the prefix's home vendor, which
could not distinguish a namespace the proxy actually routes on
(x-ai/grok-4.5 -> keep) from a stale cross-provider leftover on a relay
that only serves bare ids (openai/gpt-5.4 -> strip). A model graduating
into the first-party catalog (agent commit adding grok-4.5) silently
flipped a working custom-proxy id from preserved to stripped -> HTTP 400.
New _endpoint_advertised_model_ids() reads the already-published catalog
snapshot (never builds/probes/touches disk; O(1) per send via a
snapshot-identity memo) and answers what the endpoint advertised:
- full vendor/model advertised -> preserve (#5979, #3872, #548)
- only the bare id advertised -> strip the redundant prefix (#433)
- cold/unbuilt catalog -> preserve verbatim (fail-safe toward
the id the user actively selected)
The non-custom first-party-proxy strip arm is unchanged.
Rewrites the two pinned #433 tests to model provenance (bare-only
advertised) and adds #5979 coverage: advertised-full preserve (bare +
named custom), cold-catalog preserve-verbatim, and both-advertised
prefers-full.
Thanks @b3nw for the report and precise root-cause analysis.
* fix(chat): keep transcript pinned across composer auto-resize (#5514)
Reported by @JeffArcLogic.
autoResize() sets the textarea to height:'auto' before measuring scrollHeight
and restoring the height. The transient collapse grows the flex:1 #messages
viewport and the browser clamps a bottom-anchored scrollTop down by the collapse
delta, stranding the reader a few px above the bottom on every keystroke while
the composer is multi-row (delta proportional to composer height) — and the
clamp's async scroll event sticky-unpins the reader, dead-ending the #5516
grow-path re-pin and stream auto-follow. The prior #5516 fix only re-pinned on
NET growth, so a steady-state keystroke was missed.
Fix: snapshot #messages.scrollTop before the height round-trip and restore it
after, undoing the transient clamp within the same synchronous task. Protects
pinned AND near-bottom scrolled-up readers, takes no _programmaticScroll latch,
inert to iOS toolbar reflows. The grow-gated re-pin + ResizeObserver remain for
genuine net growth.
* test(#5514): make near-bottom-unpinned case non-vacuous (Codex gate nit)
The companion unpinned case used scrollTop=7900, below the transient clamp max
(7903), so it never actually clamped and passed on master too. Move it to 7990
(inside the clamp zone: 7903 < 7990 < settled-bottom 8023) so it genuinely
exercises the transient-clamp yank the fix prevents.
* Release exp-v0.52.52: keep transcript pinned across composer auto-resize (#5962, #5514)
---------
Co-authored-by: nesquena-hermes <agent@nesquena-hermes>
