`_reap_stale_sidecar` is `async`, but it identified the processes holding
the sidecar port with two blocking helpers called inline:
* `_find_listener_pids` -> `subprocess.run(["lsof", ...], timeout=5.0)`
* `_pid_is_sidecar` -> `subprocess.run(["ps", ...], timeout=5.0)`, once
per candidate pid
so the inspection can hold the shared gateway loop for 5 + 5·N seconds
while nothing else on it is serviced. It only runs once the /healthz probe
finds something already listening — the orphaned-sidecar recovery path —
and `_reap_stale_sidecar` is awaited from `_start_sidecar`, which runs on
every reconnect (`connect(is_reconnect=True)`). The stall therefore lands
on a live gateway that is still serving every other platform, right when a
crashed sidecar has already left an orphan behind.
Move the whole inspection to one `asyncio.to_thread` hop (one hop rather
than N+1 round trips). The reaping semantics are untouched: SIGTERM for
verified orphans, SIGKILL escalation, and both foreign-listener
RuntimeErrors behave exactly as before.
Same off-the-loop class as the inbound-image decision (#66688) and the
cron-fire verifier.
Adds a regression test asserting both the lsof lookup and the per-pid ps
check execute on a worker thread rather than the loop thread.
Follow-up widening of the #73170 pattern: apply the same
asyncio.current_task() guard used for the health task (and now the
supervisor task in _stop_sidecar) to the inbound-task cancel path in
disconnect(), so an inbound task that triggers disconnect() cannot
cancel-and-await itself.
Fixes#73159.
When the Photon sidecar exits unexpectedly, _supervise_sidecar()
(running as self._sidecar_supervisor_task) correctly detects
SIDECAR_CRASHED and calls self._notify_fatal_error(). The Gateway's
fatal-error handler answers that by calling adapter.disconnect(),
which calls _stop_sidecar() -- from INSIDE the very task that's
currently executing this whole chain.
_stop_sidecar()'s cleanup unconditionally cancelled
self._sidecar_supervisor_task. Cancelling the currently-running task
raises CancelledError at its own next await point (inside
_notify_fatal_error() or _stop_sidecar() itself). Since
asyncio.CancelledError inherits from BaseException (not Exception),
the Gateway's `except Exception` guards around the fatal-error handler
don't catch it -- the handler aborts before ever reaching the
"queue platform for background reconnection" step. Photon then stays
permanently in `retrying` state until the whole process is manually
restarted, even though detection worked correctly and the underlying
transient upstream outage had long since recovered.
Fix: in _stop_sidecar()'s cleanup, check whether
self._sidecar_supervisor_task is asyncio.current_task() before
cancelling it. A task cannot legally cancel itself in any useful way
anyway (the cancellation only takes effect at its own next await,
which is exactly the corruption described above) -- when we're inside
the supervisor's own call stack, it's already in the process of
finishing on its own once _notify_fatal_error() returns, so skip the
cancel and just clear the reference.
This mirrors an existing precedent in the same file: disconnect()
already guards its OTHER task-cancellation (self._sidecar_health_task)
the same way (`if task is not asyncio.current_task()`), just not the
supervisor task in _stop_sidecar().
Per the issue's own note that existing tests mock out
_notify_fatal_error() entirely (so this integration chain was never
exercised), added two tests that drive the REAL chain: one runs
_supervise_sidecar() as an actual asyncio task with a real
_notify_fatal_error() that calls the real disconnect() -> _stop_sidecar(),
confirming the task completes without CancelledError and that the
post-disconnect reconnect-queue step actually executes; a second
confirms the OTHER call path (external cleanup, a different task)
still correctly cancels a running supervisor exactly as before.
Reverting only the adapter.py fix (keeping the new test) reproduces
the exact CancelledError from the bug report, confirming this is a
genuine regression test.
2 new tests pass; 113/113 in the full tests/plugins/platforms/photon/
directory (no regression to sidecar lifecycle, overflow recovery, or
health-monitoring behavior).
Follow-up to #69112. That PR hardened the shared gateway dispatch path in
gateway/run.py against the *caller* being cancelled. A second, self-referential
cancellation specific to PhotonAdapter sits one layer underneath it and survived
that fix.
PhotonAdapter is the only platform adapter that awaits _notify_fatal_error()
inline, on the same task that detected the fault. Both _monitor_sidecar_health
and _supervise_sidecar run as self._sidecar_health_task /
self._sidecar_supervisor_task, and the notification routes into
GatewayRunner._handle_adapter_fatal_error_impl, which tears the adapter down via
_safe_adapter_disconnect -> disconnect(). disconnect() then cancels
self._sidecar_health_task and awaits it -- which, when the health task is what
raised the notification, means disconnect() cancels its own caller several
plain-await frames up.
disconnect()'s `task is not asyncio.current_task()` guard does not catch this.
The current task where that guard evaluates is the wrapper
_await_adapter_cleanup_with_timeout creates around disconnect() via
asyncio.ensure_future, not the health task further up the chain, so the guard
passes and the cancel lands.
CancelledError stopped subclassing Exception in Python 3.8, so the
`except Exception` that wrapped the inline notify call never saw it. The health
task died silently mid-handoff: no log line, no "exception never retrieved"
warning (cancellation is normal asyncio), and no retry. The platform stayed
stranded until the gateway was restarted by hand.
Fix: dispatch the notification onto a new task, the same pattern
DiscordAdapter._handle_bot_task_done already uses for this reason. disconnect()
can then cancel the health/supervisor task freely without that cancellation
reaching the code still running the handoff, so the handoff always reaches the
reconnect queue. Both Photon fatal call sites are converted: the health-poll
path (observed wedging) and the sidecar-crash path (same shape, not yet
observed). gateway/run.py is untouched.
Observed twice on a self-hosted gateway, ~4h38m and ~52min of silent inbound
outage, both cleared only by a manual restart, both post-dating #69112's merge.
In each case the fatal log line appears and `queued for background reconnection`
never does.
Tests: new tests/plugins/platforms/photon/test_fatal_notify_self_cancel.py
covers the self-cancellation (fails with CancelledError without this change),
that the dispatch does not block its caller, that a failing notification warns
rather than raising, and a source guard against reintroducing either inline
await. Two assertions in test_overflow_recovery.py that drove these coroutines
directly now drain pending tasks before asserting delivery, since the
notification is deliberately no longer awaited inline.
Prepared with agent assistance and reviewed before submission.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The multiplexed listener now rejects the default API_SERVER_KEY on
/p/<profile>/ prefixes (fail-closed per-profile keys). Add the
multi-profile routing section with an explicit breaking-change callout
for the next release notes.
Sibling of #64686: _run_agent's empty-final_response early return dropped
failure_reason (and #64686 only fixed the non-empty path), so downstream
consumers (TUI billing surface, transient-failure persistence) lost the
structured reason exactly when a failed run produced no text.
Also hardens the two BasePlatformAdapter identity checks (edit_message /
delete_message) with getattr so duck-typed adapters without the attribute
mean 'capability absent', not AttributeError — this was crashing the
send_progress_messages path for minimal adapters and test fakes.
Sibling of #65254 (main-slot endpoint preservation): the auxiliary scope of
POST /api/model/set dropped the request's base_url/api_key on the floor, so
an aux slot pinned to a custom/local endpoint silently depended on
model.base_url — and broke the moment the main slot switched away and
cleared it. The aux resolver already reads auxiliary.<task>.base_url/api_key
(_resolve_task_provider_model); this persists them.
Desktop side: setAuxiliaryToMain / applyAuxiliaryDraft now carry the
user-defined provider's api_url as base_url, mirroring applyMainModel.
Consolidates the native-transport half of PR #73636 by @ScaleLeanChris
onto the merged adapter: persistent NIP-42-authenticated Nostr WebSocket
subscription as the default inbound path (transport=auto|websocket|poll),
kind-44100 membership events for live DM discovery, since-timestamp
resume on reconnect with bounded exponential backoff, and automatic
fallback to CLI polling when the WS can't be established. Events route
through the same _handle_event() pipeline as the poll loop, so de-dupe,
mention gating, p-tag DM latching, and allow-lists behave identically on
both transports. Outbound stays on the CLI (one-shot sends never race a
WS auth handshake — his design).
E2E verified against a real in-process websockets relay: NIP-42
challenge -> signed kind-22242 AUTH (event id re-derived server-side) ->
REQ subscription -> EVENT dispatch -> clean disconnect.
Co-authored-by: ScaleLeanChris <chris@scalelean.com>
Extracted verbatim from PR #73636 by @ScaleLeanChris — nsec/hex key
decoding, secp256k1 point math, BIP-340 Schnorr signing, and NIP-42
AUTH event construction with optional NIP-OA owner attestation tag.
Two independent cross-PR collisions red on main (slice 8/8):
1. fd4f756492 (salvaged from stale #54514) added an early 'drop U+FFFC
placeholder' return at the top of _dispatch_inbound — written before
the deferred-wait handler (6b91b50c6e/afab7ed46e) existed further
down the same function. The early return shadowed it: _pending_fffc
never populated, no attachment-timeout tracking, 4 tests red. Remove
the duplicate block; the deferred handler already drops the
placeholder AND tracks/cancels/warns.
2. 9cf2046081 tightened sidecar_deps_installed() to require
node_modules/spectrum-ts, but test_runtime_record's _patch_spawn
fixture still created only bare node_modules/ — 2 tests red. Mirror
a real completed install.
tests/plugins/platforms/photon: 164/164 after; 158/164 before.
Two follow-ups from the voice PR (#70509).
1. macOS ARM64 onnx migration. Existing users who pinned
openwakeword.inference_framework=onnx before the tflite fix landed kept a
wake word that arms but never fires (ONNX's embedding model is broken on
Apple Silicon, upstream #336). New resolve_inference_framework() honors an
explicit framework everywhere ONNX actually works, but coerces the one
provably-dead combination (explicit onnx + macOS ARM64) to tflite with a
one-time warning. No config mutation; empty still falls back to the platform
default. Both read sites (engine init + requirements check) route through
the shared resolver.
2. Voice turn-timeout leak. Each listen cycle reassigned turnTimeoutRef
without clearing the prior 60s timer, so a stale timer from an earlier
cycle could fire handleTurn() mid-way through a later listen — after enough
idle re-listens this wedged the loop into a non-re-arming state (the
'voice chat deactivates after ~a minute' report). Clear before re-arm.
Tests: 64 wake tests (added onnx-coercion / intel-kept / tflite-kept /
empty-default cases; updated the stale 'explicit onnx kept on ARM64' test
that encoded the old broken behavior), 39 desktop voice/wake vitest, tsc +
eslint clean.
The sidebar's messaging section renders one PlatformAvatar (labelIcon) and
StatusDot per platform group. The sidebar re-renders on every streaming tick
($sessions/$workingSessionIds/$messagingSessions churn), and both leaves were
unmemoized — so every platform's avatar + dot re-rendered on each delta even
though their props (platformId/tone/className) never changed.
- PlatformAvatar: memo(forwardRef(...)) — pure fn of platformId/name/class/style
- StatusDot: memo() — pure fn of tone/class
Measured (Messaging open, settled, 2 sessions streaming, 3s window):
before: 96 wasted (PlatformAvatar 32, StatusDot 32, + brand icons)
after: 0 wasted
Both are shared primitives; the memo also helps every other consumer
(session rows, gateway menu, session tiles) that renders them under a hot
parent.
- command-center/command-palette/skills: import ordering + drop unused
HermesGateway type
- settings: wrap openSubView/openProviderView/openKeysView in useCallback so
the navGroups memo deps are honest and stable
- command-center: add setSection to navGroups memo deps
The sidebar stays mounted beneath every overlay/page, and it subscribes to
$sessions + $workingSessionIds — both tick on every streaming token. An
unmemoized SidebarSessionRow re-rendered the whole list (Codicon, labels,
status dots) on each delta, and that churn bled into every overlay opened on
top: Cron, Profiles, Agents, Starmap, Webhooks, Command Center, Settings.
- SidebarSessionRow: memo() with a custom comparator that ignores the pure
id-forwarding callbacks (fresh closures by design) and compares only the
data that changes what the row paints. Rows bail out while siblings stream.
- SessionStatusDot: the 5 $...SessionIds arrays now read via useStoreSelector
returning this session's boolean, so a dot repaints only when ITS OWN
membership flips, not on every array tick.
- Artifacts: stable cellCtx (useMemo) + memoized Primary/Location/Session
cells so a link-title fetch on one row stops re-rendering the whole table.
Measured before/after (2s idle, sessions streaming), sidebar-fed overlays:
Cron 407->~30 wasted, Profiles 732->~80, Agents 132->9, Starmap 188->56,
Webhooks 154->22.
'photon:any;-;+1555...' targets matched no parser pattern, so
_handle_send bounced them off the channel directory and failed
resolution even though the adapter accepts the GUID verbatim (the
react handler already passed them through). Recognize the DM chat
GUID shape (mirrors the adapter's _DM_CHAT_GUID_RE) in
_parse_target_ref for photon only.
The sidecar auth token is generated at spawn (secrets.token_hex) and
existed only in the gateway process memory + sidecar child env, so
_standalone_send from cron subprocesses, hermes send, or the dashboard
structurally could not authenticate (#69960).
The adapter now writes <hermes-home>/runtime/photon-sidecar.json
({port, token, pid}, 0600, atomic tempfile+os.replace) once the sidecar
passes its /healthz readiness check, and deletes it in _stop_sidecar,
on every startup-failure path, and at disconnect so a stale record
never outlives a dead sidecar. _standalone_send falls back to the
record when PHOTON_SIDECAR_TOKEN is unset, validating the recorded pid
is alive first; a stale record yields a clear 'gateway appears to be
down' error. Docs note the gateway-must-be-running requirement and the
Photon-side shared-line initiation policy (#51897).
Maintainer follow-up: _cmd_setup now validates existing tokens (#72763
salvage); the pre-existing setup tests monkeypatch a stored token, so
without stubbing check_photon_token_valid they'd hit the real dashboard
API and hang.
All Photon sidecar HTTP requests target 127.0.0.1 — they should
never be routed through a system HTTP proxy. When trust_env=True
(the default), httpx picks up macOS system proxy settings and
routes localhost requests through the proxy. If the proxy returns
a spurious response (e.g. 502), _reap_stale_sidecar() interprets
it as 'port in use by a non-sidecar process' and refuses to start,
yielding: 'pids: unknown, not a Photon sidecar'.
Set trust_env=False on all five httpx.AsyncClient call sites in
the Photon adapter so localhost sidecar communication bypasses
the system proxy entirely.
Addresses teknium1 review on #50983:
- cli.py's `hermes photon status` and adapter.py's _start_sidecar() still
used the old node_modules/-existence check while check_requirements()
had moved to a spectrum-ts content check. Extracted the check into a
shared sidecar_deps_installed(), used by all three, so an
empty/partial node_modules/ (aborted npm install) is rejected
consistently instead of only in check_requirements().
- _install_sidecar()'s success-path _NPM_ERROR_LOG.unlink() only caught
FileNotFoundError, so a PermissionError/OSError on a locked file would
propagate. Broadened to OSError.
- npm stderr was truncated only when read back in check_requirements();
an unbounded stderr was written to disk on every failed install. Now
truncated to _NPM_ERROR_LOG_MAX_CHARS before write_text().
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Problema
--------
Quando o npm install do sidecar Photon falhava, o Hermes descartava toda
evidencia e continuava normalmente - deixando o adapter de iMessage
silenciosamente ausente, sem nenhuma mensagem de erro acionavel.
Tres falhas independentes formavam o caminho de falha silenciosa:
1. check_requirements() sem logging
Cada branch de return False retornava sem emitir nenhum log. O core em
platform_registry.py so consome o bool de check_fn() e loga uma mensagem
generica com o install_hint - sem acesso ao motivo real da falha.
if not HTTPX_AVAILABLE: return False # sem log
if not shutil.which(node): return False # sem log
if not node_modules.exists: return False # sem log
2. node_modules/ parcialmente criado passava o guard (Risk 2)
npm cria node_modules/ antes de abortar em ENOSPC, timeout de rede ou
EACCES. O diretorio existia, check_requirements() retornava True (falso
positivo), o adapter era registrado, e o crash acontecia em runtime com
um erro de modulo ausente aparentemente nao relacionado ao setup.
3. stderr do npm descartado (Risk 3)
subprocess.run sem stderr=PIPE. O output de erro aparecia no terminal
durante o setup e sumia depois - diagnostico impossivel em CI/CD, Docker,
VPS headless, e qualquer reinstalacao posterior.
Correcoes
---------
adapter.py - check_requirements() agora loga por branch:
- httpx ausente -> logger.warning com nome do pacote
- node nao no PATH -> logger.warning com nome do binario e env var
- spectrum-ts ausente -> logger.debug com path do sidecar + ultimo erro npm
(DEBUG nao WARNING: estado normal pre-setup; check_fn() e chamado de
5 hot paths do core incluindo polling do /api/status)
adapter.py - content check em vez de existence check (Risk 2):
antes: if not (_SIDECAR_DIR / node_modules).exists()
depois: if not (_SIDECAR_DIR / node_modules / spectrum-ts).exists()
spectrum-ts e a unica dependencia do package.json. Checar sua presenca
garante que instalacao parcial/abortada e detectada no boot do gateway,
nao na primeira mensagem recebida via gRPC.
cli.py - stderr capturado e persistido (Risk 3):
subprocess.run passa agora stderr=subprocess.PIPE, text=True em ambas as
chamadas (npm ci e npm install fallback). O stderr capturado e:
- impresso em sys.stderr imediatamente (output visivel no terminal)
- persistido em _NPM_ERROR_LOG = sidecar/.photon-npm-error.log se
returncode != 0, limitado a 300 chars
- apagado de _NPM_ERROR_LOG se returncode == 0 (evita erro stale)
check_requirements() le _NPM_ERROR_LOG quando spectrum-ts esta ausente e
inclui o conteudo no DEBUG log - o erro do npm sobrevive ao terminal, ao
restart do gateway e a reinicializacao da maquina.
sidecar/.gitignore - adicionado node_modules/ e .photon-npm-error.log.
Isolamento - sem impacto no core:
- check_fn() continua retornando apenas bool; core nao e modificado
- Logging usa namespace plugins.platforms.photon.adapter, isolado de
gateway.* e hermes_cli.*
- Cada plugin tem seu proprio check_requirements() independente
- OSError no write/read de _NPM_ERROR_LOG e silenciado - nunca propaga
Testes - 24/24 passando:
test_check_requirements_risks.py (7 testes):
WARNING emitido quando httpx ausente
WARNING emitido quando node nao no PATH
DEBUG emitido (nao WARNING) quando spectrum-ts ausente, com path
node_modules/ vazio agora retorna False (Risk 2 resolvido)
_NPM_ERROR_LOG escrito no stderr do npm em falha
_NPM_ERROR_LOG apagado apos npm bem-sucedido
erro npm aparece no DEBUG log quando node_modules ausente
test_npm_error_log_regression.py (9 testes - vetores de falha da solucao):
return code contrato intacto (0 em sucesso, nao-zero em falha)
OSError no write do log silenciado, exit code ainda propagado
OSError no read do log silenciado, check_requirements() retorna False
stderr vazio nao cria arquivo de log
proc.stderr=None nao lanca AttributeError
log stale apagado apos reinstall bem-sucedido
DEBUG emitido mesmo sem log de erro (setup pela primeira vez)
Closes#50981
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
hermes photon setup unconditionally called regenerate_project_secret()
on every re-run, invalidating the credential held by a running sidecar
and causing all outbound sends to fail with AuthenticationError.
Now validates existing credentials via a lightweight list_users call
before deciding to regenerate. Only rotates when no credentials exist
or the existing ones are invalid, and warns the user to restart the
gateway when rotation occurs.
Fixes#50755
Maintainer follow-up to the #72763 salvage: check_photon_token_valid()
now delegates to the existing validate_photon_token() (session lookup +
/api/projects/) instead of a bespoke get-session-only probe, since the
device flow can mint tokens that pass the session check but fail the
project APIs that setup actually uses. Semantics preserved: definitive
auth rejection = stale, transient errors = probably-valid.
Two related bugs in hermes photon setup / gateway_setup:
Bug 1 -- stale token reused (401)
----------------------------------
_cmd_setup reused an existing dashboard token without validation.
The device token has a short TTL (~3-4 days observed); reusing a
stale token caused every management API call (find_project_by_name,
regenerate_project_secret, etc.) to fail with 401. The operator
saw confusing "spectrum provisioning failed: 401" errors.
Fix: check GET /api/auth/get-session before using the stored token.
On 401/403, clear the stale token with clear_photon_token() and
fall back to a fresh device-login flow automatically.
Bug 2 -- channel left disabled after successful setup
-----------------------------------------------------
After all five provisioning steps completed, config.yaml still had
photon.enabled: false, so the gateway never loaded the Photon
adapter. Every inbound iMessage hit Photon's offline auto-responder
without the operator being notified.
Fix: call write_platform_config_field('photon', 'enabled', True,
raw=True) as a final setup step so the gateway picks up the freshly
configured channel on its next start.
New public API in auth.py:
- clear_photon_token() -- discard stored token from auth.json
- check_photon_token_valid(token) -- lightweight session-check test
References: #72763
store_photon_token/store_project_credentials/store_user_numbers read-modify-
wrote auth.json without hermes_cli/auth.py's _auth_store_lock(), the
cross-process flock every other writer of that file (credential_pool
refresh, model_switch, fallback_cmd, nous_portal adapter, etc.) already
holds. A concurrent write from either side during the unlocked
load-mutate-save window silently drops the other side's update.
Review follow-up: if os.fdopen() raised before taking ownership of the
descriptor returned by os.open(), the cleanup handler unlinked the temp
file but leaked the fd. Close it explicitly on that path, mirroring the
credential-writer cleanup from #62837.
Strengthen the tests so the old writer could not pass them: an os.open
spy asserts O_CREAT | O_EXCL and an explicit 0o600 mode (the final-mode
check alone was also satisfied by the post-write chmod), and a forced
fdopen-failure test asserts the raw fd is closed and no temp file is
left behind.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
_save_auth() wrote the bearer token with tmp.open('w') — created at
process umask (typically 0o644) — and only chmod'ed to 0o600 after the
write, leaving a window where the token sat world-readable. The temp
name was also fixed and predictable (auth.json.tmp), so it could be
pre-planted (symlink attack).
Create the temp file with os.open(O_WRONLY|O_CREAT|O_EXCL, 0o600) and a
per-process random suffix, fsync before the atomic replace, and clean
the temp file up on failure. Mirrors hermes_cli/auth.py:_save_auth_store
(#19673, #21148), which hardened the same pattern in the core writer.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Assert isMessagingSource('photon') is true and that Photon/iMessage search aliases resolve, so a silent removal of the photon entry from MESSAGING_SESSION_SOURCE_IDS fails CI instead of regressing the sidebar section added in #47395.
Register Photon as a messaging platform in the Desktop sidebar:
- Add Photon brand icon (three-bar mark) to PlatformAvatar catalog
- Add photon to SOURCE_LABELS, SOURCE_ALIASES, and MESSAGING_SESSION_SOURCE_IDS
Closes#46761
Follow-up on top of #65554 (@the3asic):
- Lower FTS5 'usermerge' to its minimum of 2 (persisted in the config
shadow table, applied once per SessionDB instance). Without this a
positive-rank 'merge' skips any level holding fewer than 4 segments
(SQLite FTS5 §6.8), so the fragmented-index case the cadence targets
never converges.
- Run up to _FTS_MERGE_COMMANDS_PER_PASS (4) bounded merge commands per
index per cadence, stopping early on the documented no-progress
signal (total_changes delta < 2). Each command is its own implicit
transaction, so the write lock is released between commands.
- Skip a missing messages_fts instead of raising: the chunked
optimize-storage rebuild legitimately drops + backfills FTS tables
while writers keep running; warning every 1000 writes for the whole
backfill window would be noise, and optimize_fts() has always
treated missing tables as skippable.
- Replace traced-SQL shape assertions with behavioral tests: real
fragmented-index convergence (automerge suppressed, 60 segments) and
a 3-segment below-default-usermerge compaction test that fails
against a bare positive-rank merge (sabotage-verified).
Validation on a sqlite3.backup() copy of a real 10.7 GB production
state.db (1.49M messages, 1.3 GB + 2.9 GB FTS shadow tables):
worst per-command write-lock hold 41.8 ms (was 9.2 s / 18.1 s per
index with 'optimize'), search results byte-identical, fts5
integrity-check and PRAGMA integrity_check clean, steady-state pass
0.0 ms.
Sibling of the test_auxiliary_client.py cleanup — these two assertions
pinned the OpenRouter/Nous aux default as a frozen literal and broke on
the 3.6-flash bump (CI slice 5). Reference _OPENROUTER_MODEL/_NOUS_MODEL
instead so the next default rotation can't break them.
gemini-2.5-flash shuts down Oct 16 2026 (Google deprecation schedule)
and gemini-3-flash-preview is superseded. Update every hardcoded
default to the current GA flash model:
- gemini_native_adapter: probe_gemini_tier + _create_chat_completion
default params; free-tier guidance de-pinned from a specific model's
RPD number so it doesn't stale again
- auxiliary_client: gemini/kilocode fallback aux models,
_OPENROUTER_MODEL, _NOUS_MODEL -> google/gemini-3.6-flash
(verified live on both OpenRouter and Nous portal /models)
- provider plugins: kilocode + vertex default_aux_model
- hindsight memory plugin: gemini provider default
- setup wizard gemini list: 3-flash-preview -> 3.6-flash (matches the
curated picker catalog)
- tests: aux-client assertions that pinned the old default literal now
reference the _NOUS_MODEL constant, so the next default bump can't
break them (change-detector cleanup)
Fixes#32360.
