Prior round stripped provider prefixes from both sides and matched bare-only,
which over-matched: openai/gpt-4o would match default openrouter/gpt-4o. Now the
matcher compares bare model ids AND rejects when both sides identify DIFFERENT
providers (from provider/ prefix, @provider: qualifier, or the explicit provider
arg). Same-provider / unknown-session-provider still match. Added cross-provider
rejection regression tests.
Co-authored-by: allenliang2022 <allenliang2022@users.noreply.github.com>
The default-only context_length guard compared model.default to the session
model with exact string equality. But model.default and the session model can
be stored in equivalent-but-different shapes (bare 'claude-opus-4.8',
provider-prefixed 'anthropic/claude-opus-4.8', or '@anthropic:claude-opus-4.8').
An exact compare wrongly treats the actual default model as non-default and
drops its configured context_length cap for provider-prefixed configs.
Add api/routes._model_matches_configured_default(session_model, cfg_default,
provider) that normalizes all three shapes, and use it at all 6 guard sites
(routes resolver + the 5 api/streaming.py sites: live-usage snapshot, persistence
_skip_cc_cl, persistence fallback _apply_cfg_ctx, SSE-done _dropped_stale_cap_sse,
SSE fallback _apply_cfg_ctx). Imported function-scoped in streaming to avoid the
routes<->streaming module-level circular import. 10 helper unit tests + a
behavioral test that a prefixed default still receives its cap.
Co-authored-by: allenliang2022 <allenliang2022@users.noreply.github.com>
The threshold-rescale block runs unconditionally after the fallback and
references _skip_cc_cl/_cc_cl, which were only defined inside 'if _cc_for_save:'.
On the no-compressor path (fresh agent / interrupted stream) that raised
UnboundLocalError (caught by test_issue1857_usage_overwrite). Hoist both inits
above the block (no-op rescale when no compressor). Also widen the #1318
source-assertion test to accept the widened fallback gate (still asserts the
falsy-check invariant).
Co-authored-by: allenliang2022 <allenliang2022@users.noreply.github.com>
Codex regression gate (+Opus, both independently) found the default-only guard
dropped the stale compressor cap but two sibling paths stayed inconsistent:
1. Per-turn persistence: fallback resolver only ran when context_length was
falsy, so a previously-persisted stale 232K survived forever on non-default
sessions. Now also runs when _skip_cc_cl, and rescales threshold_tokens to
the recomputed real cap (or clears it).
2. Terminal SSE 'done' payload: re-emitted the stale compressor threshold, so
messages.js overwrote S.lastUsage and the indicator reverted on stream end.
Now rescales threshold to the resolved window when the stale cap was dropped.
Added 3 source-structure regression tests pinning both fixes; bumped the brittle
test_pr1341 distance limit 13000→15000 (+ noted it should become structural).
Co-authored-by: allenliang2022 <allenliang2022@users.noreply.github.com>
- test_pr1341 distance limit 9000→13000 (the PR legitimately added the
default-only pre-save guard block; the test is designed to be bumped when a
new pre-save mutation block lands — this was the only CI-red cause on shard 1).
- new tests/test_issue3256_context_length_default_only_guard.py: verifies the
global model.context_length cap applies ONLY to model.default (revert-fix-
verified — fails on master, passes with the fix).
Co-authored-by: allenliang2022 <allenliang2022@users.noreply.github.com>
The default-only guard corrected context_length to the real per-model cap
(e.g. 1M for claude-opus-4.7-1m) but left threshold_tokens pointing at the
ContextCompressor's stale value (computed from the global 232K cap → 197.2k
@ 85%). UI then showed 'auto-compress at 197.2k / 1M' which is misleading.
Rescale threshold_tokens by the real/orig ratio so the displayed trigger
reflects the actual window (e.g. ~850k @ 1M).
NOTE: this only corrects the SSE display payload. The real auto-compress
trigger lives inside ContextCompressor in hermes-agent (agent_init.py:1446
constructs it with the global cap). A full fix requires a parallel change
upstream — tracked separately.
_live_usage_snapshot() runs on every metering tick (~10x/sec while streaming).
The #3256 default-only guard recomputed get_model_context_length() there on
every tick for non-default models, which does a config read + potential
metadata/network probe — freezing claude-opus-4.7-1m streams while the default
model (4.8) stayed fast (guard not triggered for it). Resolve the real cap at
most once per stream via _real_ctx_cache. Backend-only, no frontend changes.
Re-applies cb0065eb + b34311b3 (context_length default-only guard) which were
dropped by the upgrade reset to v0.51.185. Fixes 4.7-1m context window showing
as stale global cap (232K) instead of real 1M metadata. Backend-only: touches
api/routes.py + api/streaming.py, zero frontend/render changes.
Codex regression gate found the launchd guard blocked ANY ctl.sh start while a
launchd job was live — including a legitimate second instance on a different
port (HERMES_WEBUI_PORT=8788). Now _launchd_webui_pid only treats the launchd
job as a conflict when its PID is actually listening on the requested CTL_PORT
(via a new best-effort _pid_listens_on_port helper using lsof); a different-port
start is allowed. When port ownership can't be determined (no lsof), falls back
to guarding only the default 8787 port so non-default ports are never wrongly
blocked. Adds a different-port-allowed regression test + makes the existing
block test deterministic. Co-authored-by preserved.
Both pre-release reviewers (Opus advisor + Codex regression gate) converged on
the same MUST-FIX:
- #3194: treating gateway_stale_stopped_state as 'configured' contradicted
#1944 (a stopped root gateway should read like 'not configured' so the
banner doesn't nag). Now ONLY stale-RUNNING metadata
(reason=gateway_stale_running_state or gateway_state=='running') flips
configured=True; stale-stopped falls through to bool(identity_map) like the
genuinely-unconfigured case. Updated the test accordingly + added a
stale-stopped no-regression test.
Opus follow-ups also applied:
- #2905: narrowed the populated-home markers to WebUI-only artifacts
(webui/, webui/sessions, webui/settings.json), dropping config.yaml/auth.json
so a long-time agent user installing WebUI fresh isn't wrongly diverted to
the legacy %USERPROFILE%\.hermes (auth.json predates #2897 there).
- profiles._resolve_base_hermes_home(): narrowed except Exception -> ImportError
so a real bug in the config helper still surfaces.
Adjacent suites green: #2840, #1879, gateway_status_agent_health (66 tests).
Two urgent breaking bugs that had no PR, combined into one hotfix.
#2905 (data-loss-class): v0.51.134 moved the Windows default Hermes home
from %USERPROFILE%\.hermes to %LOCALAPPDATA%\hermes (PR #2897) with no
migration, so upgrading Windows users opened the app to empty
sessions/pins/settings (data intact on disk, at an address the new build
no longer read). _platform_default_hermes_home() now prefers the populated
legacy home ONLY when the new location is not yet established —
non-destructive (no file moves) and self-healing on next launch.
profiles._resolve_base_hermes_home() delegates to the same config helper so
the active-profile pointer can never drift from STATE_DIR.
#3194: GET /api/gateway/status reported 'Gateway not configured' on a fresh
two-container Docker deploy because an alive=None + gateway_stale_running_state
health payload with an empty identity_map fell through to
configured=bool(identity_map)=False. The alive=None branch now treats a
payload carrying gateway metadata (gateway_state detail, or a stale-running/
stale-stopped reason) as configured.
Tests: +17 regression tests (11 for #2905 incl. full truth table + non-destructive
guard + POSIX no-op; 6 for #3194 incl. 5 no-regression guards). Full suite
7090 passed, 0 failed.
Closes#2905Closes#3194
Shift from backend mtime-based detection to frontend SSE deduplication.
Backend: Revert gateway_watcher.py to original pure hash-based polling.
Remove _get_db_mtime, _detect_gateway_restart, and mtime tracking.
This is a no-op in behavior — the original was already hash-only.
Frontend: Add deduplication at the SSE event handler level.
- _gatewaySessionSnapshotKey(sessions): deterministic key from
session_id + updated_at + message_count (same fields as backend hash)
- _isGatewaySessionForSnapshot(session): classify non-webui sessions
- _isDuplicateGatewaySessionSnapshot(sessions): compare SSE payload
against current _allSessions, filtered to gateway subset
- SSE sessions_changed handler wraps renderSessionList() in dedupe:
identical data → skip refresh
This directly addresses the real root cause: the SSE reconnect snapshot
(routes.py:7735) unconditionally pushes an initial snapshot, and the
frontend always re-renders. After this fix, a reconnect with unchanged
session data is correctly detected and the redundant redraw is skipped.
Previously submitted as #3259 (backend mtime approach, now closed per
maintainer review).
Defense-in-depth flagged by both pre-release reviewers (Opus + Codex), both
non-blocking but cheap on a credential-handling surface:
- reject any non-http(s) HERMES_WEBUI_RUNNER_BASE_URL scheme at construction
(a misconfigured file:// / ftp:// can never reach urlopen);
- route requests through an opener that does NOT follow 3xx redirects, so a
misbehaving/compromised runner cannot smuggle the Bearer token to another host.
Both operator-misconfiguration-only (not user-reachable). +2 regression tests.
Co-authored-by: AJV20 <AJV20@users.noreply.github.com>
Opt-in HTTP runner-client boundary for the runner-local runtime adapter
(RFC hermes-run-adapter-contract / #1925, Slice 4c/4d). Default-OFF:
without HERMES_WEBUI_RUNNER_BASE_URL the factory preserves the bounded
'not configured' path; when set, WebUI acts only as a JSON HTTP client +
SSE bridge for start/observe/status/controls without owning runner maps.
New api/runner_client.py + additive _runner_* helpers in api/routes.py;
no change to the live _run_agent_streaming legacy path.
Co-authored-by: AJV20 <AJV20@users.noreply.github.com>
Pre-release Codex regression gate caught that _normalizeArtifactPath()
did not strip ./ or ~/ prefixes, so a tool arg recorded as ./foo.md did
not match a file-tree-opened foo.md in _turnMutatedPreviewPaths — the
open preview was left stale after an agent edit via a ./-prefixed path.
Strip ~/ and leading ./ before ignore/membership checks. Node-driven
regression test pins foo.md == ./foo.md == ~/foo.md and confirms the
existing ignore-dir / URL / empty rejections still hold.
Co-authored-by: Pamnard <pamnard@users.noreply.github.com>
Track write/edit tool paths per turn, refresh the open preview on
tool_complete and after preservePreview loadDir on stream done, without
closing preview for unrelated responses or wiping unsaved local edits.
Increase MESSAGE_UPWARD_INTENT_MS from 450ms to 2000ms to fix a race
condition where the user scrolls up during streaming, pauses to read
for >450ms, and then gets snapped back to the bottom.
The root cause: after the 450ms upward-intent window expires, DOM layout
changes from the streaming markdown parser (smd), tool card insertions,
or code re-highlighting can trigger scroll events that the handler no
longer recognizes as user-initiated. When the resulting position lands
inside the 250px near-bottom zone for two consecutive samples, the
hysteresis counter re-pins (_scrollPinned=true) and the next streaming
token's scrollIfPinned() call forces scrollTop to the bottom.
With a 2-second window, the user's upward intent persists through typical
streaming DOM churn. Downward scrolling and the scroll-to-bottom button
are unaffected — movedUp requires top < _lastScrollTop-2 which is false
for downward movement regardless of the intent timeout.
Refs: #1360 (macOS momentum protection), #1731 (direction-aware unpin)
When agent checkouts track main past an older tag but the newest published
tag is on a divergent side branch, stop advertising tag-based updates and
route check/apply through the upstream branch instead.
_run_git used subprocess.run(text=True) without an explicit encoding, so on
Chinese Windows (and other non-UTF-8 codepages) git stdout was decoded with the
locale codepage. _dirty_suffix() runs `git diff --binary HEAD`, whose binary
bytes are not valid GBK, raising UnicodeDecodeError in the subprocess reader
thread. That left r.stdout = None, so `r.stdout.strip()` raised AttributeError
during module import of api.updates, crashing server.py before it could bind
its port.
Force UTF-8 decoding with errors=replace and guard against None defensively.
Chrome's password manager aggressively autofills the clarify card's
input field with saved credentials (e.g. provider base URLs) despite
autocomplete='off'. This causes two bugs:
1. 'Clarification closed. Your draft was kept in the composer.' appears
on every session completion because _stashClarifyDraft reads the
autofilled value and treats it as a user draft.
2. The autofilled URL gets injected into the main composer, confusing
the user.
Fix: add readonly attribute to the clarify input element so Chrome's
autofill ignores it. When showClarifyCard() makes the card visible,
readonly is removed programmatically so the user can type normally.
Both the static HTML (index.html) and the dynamic DOM creation
(_ensureClarifyCardDom in messages.js) are patched.
