The msg_type_str == "richText" branch reset msg_type to PHOTO/TEXT after
the rich-text item scan had already promoted it (e.g. a native voice item
→ VOICE), dropping voice notes from the auto-STT path. Only re-derive when
the scan left the type at TEXT.
Ports the root-cause analysis from PR #38276 (stale, targeted the deleted
gateway/platforms/dingtalk.py) onto the live plugin adapter, with
regression tests.
Refs #38211#38219#38276
- Promote EXT_MAP to a module-level constant for reuse
- Classify DingTalk image messages and image file attachments as PHOTO
- Extract DingTalk card/interactiveCard document link content with defensive parsing
- Handle None, empty, JSON, and plain-string card content fields
- Add coverage: 8 card + 4 interactiveCard + 5 _extract_media = 17 test cases
- Remove a shadowed duplicate TestExtractText class (pytest collected only the later one)
3 fixes for the dingtalk-platform plugin (plugins/platforms/dingtalk/adapter.py):
1. _extract_text: parse ASR recognition text from voice messages via
extensions['content']['recognition'], fall back to fileName for
file messages ("[文件] xxx")
2. _extract_media: handle audio/file/image msgtype_str with correct
MIME mapping; exclude audio from media_urls to preserve DingTalk's
built-in ASR result (avoids failed whisper re-transcription)
3. _resolve_media_codes: parse downloadCode from extensions['content']
for file/image message types
All tested with live DingTalk group chat: voice → recognition text,
PDF/Markdown file → filename + download.
Lark's native "audio" msg_type is an in-app voice recording — uploaded
audio files arrive as "file"/"media". But _resolve_normalized_message_type
resolved the "audio" preferred type to MessageType.AUDIO, which the gateway
treats as a non-transcribed file attachment (run.py: AUDIO -> audio_file_paths,
"never STT"; VOICE -> audio_paths, "always STT"). Result: a Feishu voice
note reached the agent as an untranscribable audio attachment and was
silently ignored — the user's spoken message never became text.
Every other platform that receives native voice notes (Telegram, Discord,
Slack, WhatsApp, Signal, Matrix, WeChat, WeCom, DingTalk, QQ, BlueBubbles,
Mattermost, Yuanbao) classifies them as MessageType.VOICE. Feishu was the
only one classifying them as AUDIO. This is the follow-up to #28993, which
added native voice-note transcription for Discord + DingTalk but did not
cover Feishu.
Return MessageType.VOICE for the "audio" branch. The branch is reached only
for Lark's top-level audio msg_type (set in the normalizer; file uploads map
to "document"), so VOICE is unconditionally correct here — no risk of
auto-transcribing an uploaded music/audio file.
- plugins/platforms/feishu/adapter.py: _resolve_normalized_message_type audio
branch returns VOICE instead of resolving to AUDIO via mime.
- tests/gateway/test_feishu.py: test_extract_audio_message_downloads_and_caches
asserted the old AUDIO behavior on a fixture literally named voice.ogg —
updated to expect VOICE (the corrected classification).
- tests/gateway/test_feishu_voice_message_type.py: new focused regression
tests (audio->VOICE with and without mime; photo/document/text unaffected).
Rebased onto current main: the Feishu platform moved from the single-file
gateway/platforms/feishu.py into the plugins/platforms/feishu/ package; the
fix applies to the same _resolve_normalized_message_type logic at its new home.
Verified the new voice tests fail when the branch resolves to AUDIO and pass
with VOICE, while the photo/document/text cases are unaffected either way.
Note: classification is mock-tested here; the downstream STT pipeline is the
shared, already-proven path (#28993). End-to-end verification on a live
Feishu account would be a welcome confirmation.
Refs #28993 (sibling-gap: Feishu was the platform left uncovered).
Keep one owner for PATH/local-prefix ffmpeg discovery: ffmpeg_utils now
delegates to tools.transcription_tools._find_ffmpeg_binary and only adds
the Discord-specific FFMPEG_PATH override and Windows winget fallback on
top (follow-up to PR #60627 by @LauraGPT, fixes#60624).
Discord's voice socket needs a brief warm-up before receiving clients
actually hear audio; the first ~100-200ms is lost, clipping the first
word/syllable of TTS playback. Prepend a configurable lead of silence to
speech on both playback paths:
- Mixer path: new _lead_silence_bytes() helper prepends PCM silence
(BYTES_PER_MS constant added to voice_mixer.py) before play_speech, on
both the reply and the pre-tool ack.
- Legacy FFmpegPCMAudio path: apply -af adelay=<ms>:all=1.
Tunable via discord.voice_fx.lead_silence_ms (default 200, 0 disables).
Fixes#66827
VoiceMixer duck-typed the discord.AudioSource interface (is_opus, read,
cleanup) but never inherited from it. discord.py's vc.play() does an
isinstance check and rejects non-AudioSource objects, causing the voice
fx mixer to silently fail with:
"Voice mixer failed to start: source must be an AudioSource not
VoiceMixer"
Added missing `import discord` and changed the class definition from
`class VoiceMixer:` to `class VoiceMixer(discord.AudioSource):`.
- Wire adapter._voice_input_callback at connect and reconnect so voice
transcription is forwarded without requiring /voice join (#60623).
- Add optional text_channel_id and source params to DiscordAdapter
.join_voice_channel() so automatic/programmatic voice joins can
establish the text-channel binding needed by _handle_voice_channel_input.
- Add TestVoiceInputCallbackWiring: asserts callback wiring on startup
and reconnect for Discord adapters with voice attributes.
Closes the #32029 gap: TelegramAdapter.send_voice passed captions raw with
no parse_mode, so auto-TTS captions (which carry the agent's markdown
reply) showed literal *asterisks*, backticks and [links](...). Captions
are now formatted to MarkdownV2 via the adapter's format_message when the
formatted text fits Telegram's 1024-char caption cap, with fallback to the
plain truncated caption when formatting overflows or the Bot API rejects
the entities (mirrors the text-send markdown fallback ladder).
Fixes#32029
Salvaged from PR #53157 (@LLQWQ). Three changes for native Feishu voice
bubble delivery:
1. Include audio duration in the file-upload body when uploading opus
files — Feishu renders 0:00 bubbles without it. Duration is extracted
by parsing the OGG container's last granule position (pure Python,
no ffprobe dependency).
2. Thread routing fallback: Feishu's create-message API rejects
msg_type='audio' with receive_id_type='thread_id' (error 99992402);
retry via the reply API against the thread's last message, then fall
back to chat_id routing.
3. (dropped) tools/tts_tool.py want_opus hunk — superseded by main's
OPUS_VOICE_PLATFORMS, which already includes feishu. The PR's stray
scripts/release.py hunk was also dropped (frozen AUTHOR_MAP policy;
mapping added under contributors/emails/ instead).
Fixes#45557
Refs #18831#16524
Salvaged from PR #18861 (@shellybotmoyer). Switch the central Opus
transcoders from CBR 64k (-vbr off) to VBR 48k with -application voip and
-compression_level 10. The old CBR settings produced lower-quality speech
audio and occasionally failed to trigger Telegram's native voice-bubble
rendering. Applied to _ffmpeg_transcode_to_opus (the single transcoder
behind _convert_to_opus and _repair_ogg_container), the Gemini WAV→OGG
path, and the Matrix adapter boundary transcoder added in this branch.
Fixes#18818
Salvaged from PR #68063 (@malaiwah). MatrixAdapter.send_voice now transcodes
any non-Ogg audio to Ogg/Opus at the adapter boundary (best-effort — the
original file is sent unchanged when ffmpeg is unavailable), so MSC3245
voice bubbles render even when a caller hands the adapter MP3/WAV audio.
_matrix_voice_metadata_for_file probing now runs via asyncio.to_thread so
ffprobe/ffmpeg subprocess timeouts can't stall the adapter event loop.
The PR's tools/tts_tool.py want_opus hunk was dropped: main's
OPUS_VOICE_PLATFORMS set (PR #73072) already includes matrix; the Matrix
opus-routing test is kept.
Refs #14841
Salvaged from PR #68063 (base commit, runner-path hunks superseded by the
platform-aware OPUS_VOICE_PLATFORMS fix in this branch). Element and other
Matrix clients render voice bubbles more reliably when m.audio events carry
duration and waveform metadata; probe both best-effort via ffprobe/ffmpeg.
The native Gemini provider profile's default_aux_model and the curated
model picker catalog were still pinned to gemini-3.5-flash, a stale
generation now superseded by gemini-3.6-flash (documented GA). Bump
both so the auxiliary-task default and the picker stay in sync with
the current model.
Contract test asserts the durable lockstep invariant only
(default_aux_model is a member of _PROVIDER_MODELS["gemini"]) rather
than pinning either side to a frozen model-name string, so it doesn't
need updating on the next model-generation bump.
- Change await self._send_read_receipt to asyncio.create_task to avoid
blocking message dispatch on slow bridge responses (matches BlueBubbles
pattern). Up to 5s per-message latency eliminated.
- Update tests: assert_called_once_with instead of assert_awaited_once_with
since the receipt is now scheduled, not directly awaited.
- Add send_read_receipts documentation to whatsapp.md following the
BlueBubbles docs pattern.
Direct tool HTTP calls already identified as Hermes-Agent, but the main
OpenAI-SDK chat path still sent OpenAI/Python. Set Hermes-Agent/<ver> for
api.x.ai clients (xai + xai-oauth) so normal text traffic is attributed correctly.
The per-IP httpx transports were built once in __init__ and never torn
down. A connect that reached ESTABLISHED and was then closed by the peer
left its socket in CLOSE_WAIT inside the pool, and the failure path only
logged and continued — so the poisoned pool was retained and leaked one
descriptor per retry.
With DNS for api.telegram.org failing, every poll fell through to the
seed IP and leaked another fd every ~2.5s. The bot gateway reached 177
CLOSE_WAIT sockets against launchd's 256 soft limit and wedged: accept()
on the gateway port, config reads and DNS resolution all failed with
EMFILE, which in turn made the primary path fail and fed the loop.
Build fallback transports lazily and discard them on a retryable connect
failure, and bound every pool at 8 connections (httpx defaults to 100,
so two seed IPs plus primary could alone exceed the fd ceiling).
Generated with [Claude Code](https://claude.ai/code)
via [Happy](https://happy.engineering)
Co-Authored-By: Claude <noreply@anthropic.com>
Co-Authored-By: Happy <yesreply@happy.engineering>
_check_local_runtime() only imported 'hindsight' and
'hindsight_embed.daemon_embed_manager', a strictly weaker import surface
than the embedded daemon actually needs: the daemon imports
sentence_transformers at startup (embeddings + reranker). When the
embedding stack is broken (e.g. a dependency conflict on a shared package
like huggingface-hub), the daemon can never start, yet is_available() and
'hermes memory status' still report Hindsight as available — every
retain/recall then fails silently.
Import sentence_transformers in the same probe so local/local_embedded
availability goes red with the real ImportError as the reason, letting
the agent degrade loudly instead of silently dropping memory. Local path
only; cloud mode is untouched and no network or model download is
triggered by the import.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MN8RMDLwxCfFxwtADoEJJf
Stop offering deepseek-chat/reasoner in the static catalog and point
fallback/aux defaults at the permanent v4 IDs. Keep retired aliases in
a detection-only map so /model deepseek-chat still resolves to deepseek.
MEDIA_DELIVERY_EXTS in gateway/platforms/base.py omitted .3gp, causing
MEDIA: tags with .3gp files to leak as plain text instead of being
extracted for native video delivery. _VIDEO_EXTS in
tools/send_message_tool.py and _MIGRATION_VIDEO_EXTS in the Feishu
adapter omitted .webm, causing .webm files to be classified as
documents instead of video on Telegram and other platforms.
Both extensions are already present in every gateway-side _VIDEO_EXTS
definition (run.py, kanban_watchers.py, weixin.py, base.py local).
Closes#71621, Closes#71603
Surface a user-visible delivery notice when non-streaming media dispatch
gets success=False after MEDIA tags were stripped, and validate forum
starter-message attachments the same way as direct channel sends (#66797).
Outbound MEDIA video/document tags on the Discord non-streaming path
were extracted (stripped from the visible text) but never delivered:
no attachment, no error, and no dispatch log line (#66797). The
open-handle discord.File plus singular file= form could race Discord's
multipart encoder after an earlier image batch on the same channel and
return a successful message carrying zero attachments.
Fix _send_file_attachment (used by send_video/send_document/
send_image_file) to:
- use a path-based discord.File via the plural files=[...] kwarg, the
same pattern as the working send_multiple_images batch path;
- pre-flight os.path.isfile and return a File not found result instead
of raising deep inside discord.File;
- fail loud when Discord accepts the message but attaches nothing, so
the dispatch loop surfaces a warning rather than a silent drop;
- add INFO dispatch logs (base non-image MEDIA fan-out, video send, and
file attachment) so a MEDIA:.mp4 cannot vanish without a trace.
Tests (tests/gateway/test_discord_send.py):
- path-based files=[...] kwarg used, singular file= unset;
- fail-loud when the returned message has no attachments;
- missing file fails fast without resolving the channel;
- forum-parent delivery routes through create_thread with files=[...];
- end-to-end image+video response routes the mp4 to send_video while
images still batch.
Fixes#66797
The identity-refresh TTL used 0.0 as the "never checked" sentinel and
compared it against time.monotonic(). That epoch is arbitrary and starts
near zero on a freshly-booted host, so on CI runners and containers
`monotonic() - 0.0` was itself below the TTL — "never checked" read as
"checked just now" and the first identity refresh was suppressed for the
first 5 minutes of uptime. The stale-handle recovery therefore did nothing
on exactly the machines most likely to be freshly booted.
Invisible on a long-lived dev box (uptime >> TTL); caught by CI.
- Sentinel is now None, meaning never checked and always stale.
- Both TTL comparison sites route through _bot_identity_is_fresh().
- Regression test pins the invariant under a faked 12s-uptime clock.
Verified by re-running the recheck tests against a simulated 3s-uptime
host: green with the fix, and restoring the 0.0 sentinel reproduces the
exact CI failure locally.
Renaming a Telegram bot's @username in BotFather silently stopped the
gateway from answering in groups.
PTB caches getMe() in Bot._bot_user and only rewrites it inside get_me(),
so after a rename the adapter kept comparing mentions against the OLD
handle. The exclusive-mention gate then saw the new @handle, failed to
match itself, and concluded the message was addressed to a different bot
— dropping it before the reply and wake-word fallbacks could run. Native
replies to the bot were discarded too. Polling mode recovered on the next
90s heartbeat; webhook mode never calls get_me() again, so it stayed dead
until restart.
Separately, the bot-handle pattern assumed every bot username ends in
"bot". Collectible (Fragment) usernames can be assigned to bots and drop
that suffix (@jarvis, @pic), so such a bot could not recognise its own
handle in the entity-less fallback and was suppressed by any message that
also named another bot.
- Route every mention comparison through _current_bot_username(), which
prefers the last observed handle over PTB's cache.
- Learn the live handle from inbound updates: Telegram stamps the current
username on our own messages and on reply_to_message. Guarded by user
id, so another account's handle is never adopted.
- Re-check identity out of band (TTL-bounded, one getMe per 5 min) when
the exclusive gate is about to drop a message — the exact stale-handle
symptom — so the mistake self-corrects instead of persisting.
- Refresh identity in webhook mode via a dedicated low-frequency loop,
cancelled on the same teardown fence as the heartbeat.
- Match our own handle by identity rather than shape. Foreign handles keep
the deliberate "...bot" narrowing so human @handles still never act as
routing hints (the intent behind ce4d857021).
Validation: 11 regression tests; sabotage runs confirm each behavioral
test fails with the fix reverted. 157 tests green across the Telegram
gating, reconnect, and topic-mode suites.
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.
Auxiliary LLM calls routed through OpenRouter now carry the `session_id`
sticky routing key, so they pin to the same provider endpoint as the
conversation they belong to instead of routing independently.
`OpenRouterProfile.build_extra_body` sourced the key only from its explicit
argument. Auxiliary call sites — compression, title generation, vision,
web_extract, session_search, MoA slots — funnel through
`agent.auxiliary_client`, which has no session handle and never passes one.
Those calls sent NO sticky key at all, so OpenRouter fell back to hashing
the opening messages and each aux call could land on a different provider
than the conversation it served. Per OpenRouter's prompt-caching docs a
present `session_id` is used directly as the routing key and activates
stickiness on the first successful request rather than only after a cache
hit, so the missing key also delayed pinning for the main loop.
Resolve it from the ambient conversation contextvar first, explicit argument
as fallback — the same resolution the Nous Portal profile uses (f2f4df064d)
and the same one `nous_portal_tags` already used for the `conversation=` tag.
The ambient value is the session-lineage ROOT, so the key also stays stable
for installs that opt out of the default `compression.in_place: true` and
across delegate-subagent trees.
The xAI `x-grok-conv-id` cache-affinity header in `build_api_kwargs_extras`
had the identical gap and is fixed the same way; it remains model-gated to
`x-ai/grok-*` / `xai/grok-*`.
Fixes#70820. Credit to @webtecnica, who reported it and submitted #70883
first; that PR threaded `session_id` through `call_llm` to reach the same
goal. This takes the ambient-contextvar route instead because none of the
34 `call_llm` call sites currently pass a `session_id`, so the parameter
alone would not have changed any live request.
Tests: 2 cases in tests/providers/test_provider_profiles.py, both verified
to fail against the pre-fix behavior via a sabotage run. E2E-validated on a
real AIAgent driving the out-of-turn compaction path.
The Nous Portal profile publishes a top-level `session_id` that the Portal
uses as its sticky routing key, pinning a conversation to one upstream
endpoint so explicit Anthropic `cache_control` breakpoints stay warm
(those caches are instance-local, so a reroute cold-writes instead of reads).
`NousProfile.build_extra_body` sourced that key only from the explicit
`session_id` argument. Auxiliary call sites — compression, title generation,
vision, web_extract, session_search, MoA slots — funnel through
`agent.auxiliary_client`, which has no session handle and never passes one.
They carried the `conversation=` tag but NO sticky key at all, so each one
routed independently of the conversation it belonged to.
Resolve the key the way `nous_portal_tags` already resolves that tag:
ambient lineage-ROOT contextvar first, explicit argument as fallback. Fixes
every aux call site at once with no per-call-site plumbing.
Also publish the root in the `_compress_context` forwarder when nothing is
ambient. Out-of-turn entry points (`/compact`, the gateway `/compress`
command and its hygiene sweep, partial head compression) call it outside
`run_conversation`'s scope, so the summarizer's aux call ran untagged and
unkeyed; the caller's value is restored in `finally` so a compaction never
leaks its tag.
Scope note: this does NOT keep the compaction turn's own prompt cache warm.
Compaction replaces the history with a summary and rebuilds the system
prompt, so that request is a cold write on any endpoint — what a stable key
buys is the turns AFTER compaction reading the cache it wrote. Under the
default `compression.in_place: true` (#38763) the session id does not rotate
at compaction, so for the main loop the ambient root and the explicit
argument already agree; the ambient resolution additionally holds the key
stable for installs that opt back into rotating compaction and across
delegate-subagent trees.
Salvaged from #71747. Subsumes the aux-call half of #70883 (@webtecnica),
which threaded `session_id` through `call_llm` to close the same gap.
Tests: 4 cases in tests/agent/test_portal_tags.py; both fix halves verified
to fail against the pre-fix behavior via sabotage runs.
The Codex image backend rejected our own request shape for every account, and
we then translated that rejection into "Image generation is not enabled for the
current Codex account. Switch the image provider to OpenAI API key, FAL, or
xAI." — telling every affected user to abandon a provider that had never
actually been tried. That message is why this reads as a setup failure rather
than a bug: the wire error was replaced with a confident, wrong diagnosis.
Removes the classifier and its exception, so any HTTP failure surfaces
verbatim. The paired request-shape fix (previous commit) is what makes the
400 stop happening; this commit makes the next one diagnosable.
Also fixes error-body truncation: bodies were head-truncated at 500 chars, and
Codex error payloads can carry hundreds of bytes of leading metadata, so the
user got a wall of padding and no message. _summarize_error_body() prefers the
parsed error.message and falls back to a truncated raw body.
Docs: drop the unqualified image-to-image claim for the Codex backend and note
that the hosted tool call cannot be forced, so it is best-effort.
Verified E2E against a local fake Codex backend: success path writes a real PNG
with no tool_choice on the wire; the 400 path now returns api_error carrying
"Tool choice 'image_generation' not found in 'tools' parameter" (148 chars)
instead of the entitlement message. Sabotage run confirms all 4 regression
tests fail when the old behavior is restored.
Refs #19505, #49008, #31335.
The chatgpt.com/backend-api/codex backend 400s on every tool_choice
shape for the hosted image_generation tool — it looks up tool_choice as
a function name and never recognizes hosted-tool entries. Removing the
field from _build_responses_payload() lets the host model decide; the
instructions field nudges it toward the tool.
Salvaged from PR #19979 (originally targeted the old
client.responses.stream call, which no longer exists on upstream/main;
the live request now flows through _build_responses_payload + httpx in
_collect_image_b64).
Pair conditional approval buttons into rows of two so the full Allow Once /
Session / Always / Deny set stays readable instead of one truncated 4x1 row.
Follow-up hardening for the salvaged #69240 readiness gate (#67498):
- _start_polling_once now returns its (generation, progress_event) pair
so the strict cold-start gate binds to exactly the generation it
started, instead of re-reading self._polling_progress_event which a
concurrent recovery task may have replaced with a newer generation's
event (the G1/G2 race flagged in the #69240 review).
- Strict cold start no longer schedules background polling recovery: a
polling error during the readiness wait is captured by a strict
callback and fails the connect attempt immediately with a loud
OSError, so GatewayRunner disposes the partial adapter and retries
with a fresh one — no more waiting out the full readiness deadline on
a generation that already errored, and no G2-on-partial-app healing.
- After readiness is proven the strict callback delegates every later
polling error to the real background-recovery callback, preserving
the existing degraded/reconnect semantics for the polling lifetime.
- The readiness-timeout error message now states the deadline and that
the gateway will retry with a fresh adapter (loud failure, not a
silent wait).
- Regression tests: current-generation progress connects; a polling
error during strict cold start fails fast without scheduling
background recovery (the #67498 idle-threads shape); stale-generation
progress is rejected.
Progresses #67498
Use wall deadlines for deleteWebhook and start_polling, then fail cold startup unless getUpdates proves progress. This lets the gateway discard partial PTB state and retry with a fresh adapter.\n\nRefs #67498
The cherry-pick auto-resolution + AST sweep applied plain utf-8 at three
.env reader sites where the salvaged PRs (#62617, #62123) deliberately
use utf-8-sig — a Notepad BOM must not hide/duplicate the first key.
Restore the contract (tests pin it).
AST-driven pass over every Path.read_text()/write_text() without an
explicit encoding= across non-test code: 71 sites in 34 files
(skills_hub, hermes_cli/main+profiles+service_manager+container_boot,
mem0/hindsight/honcho plugins, achievements dashboard, release/CI
scripts, productivity+comfyui skill helpers, agent/*). Verified zero
positional-encoding collisions before insertion; per-file compile()
check after.
Adds a check-windows-footguns rule flagging bare single-line
read_text/write_text (multi-line forms stay covered by the AST guard
test from #38985). Together with the salvaged contributor commits this
retires the ~169-site bare file-I/O class (#37423's long tail).
Follow-up to review feedback:
- mem0 _prompt_api_key read .env with the locale default, so a Notepad
BOM hid the first key from the masked current-value lookup; read it
with utf-8-sig + errors=replace like the canonical readers in
hermes_cli/config.py.
- hindsight _load_simple_env used plain utf-8; it also parses the Hermes
.env during post_setup, where a BOM stuck to the first key. Switch to
utf-8-sig + errors=replace.
- Add hindsight regressions: BOM key matching in _load_simple_env and in
the cloud post_setup writer, plus non-ASCII round-trip preservation,
and a mem0 regression for the BOM'd masked-key lookup. The BOM tests
fail without the fix on any platform.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
