The spare-key note scored 0.832 to 0.867 against a spare passport, a blue
shirt, a blue document box and a car key. Score and margin cannot separate
those: the right note runs 0.817 to 0.892 and the silent cases 0.787 to
0.874, so the ranges overlap and structure has to decide.
RecallAllowed now takes two structural facts from the router. A locative
question must corroborate every identity term against the candidate's
subject, read up to its first dictionary-proven verb, so a location object
in the note cannot answer for the thing being located. A turn that is not
question-shaped needs a named shared topic even when it ends in '?', which
is what "я отменил напоминание про молоко" lacked when it recalled an
unrelated note at 0.825 with no runner-up to fail the margin.
query_min_score moves 0.55 to 0.80 for tokenizer rev 2. The held-out
fixture answers 14/27 real recalls and 0/14 false ones.
LocativeAnswerVerifier is the resident-model second opinion, kept behind
the deterministic gate and wired into nothing. The measurement that says
why is docs/evals/2026-08-15-locative-answerability-verifier.md.
--no-verify: master is the working branch this session by the owner's call.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
CLAUDE.md, internal/config/voice.go and docs/routing.md all say the routing
heads graph is a fine-tuned copy of the embedder, never the embedder's own file.
Nothing enforced it. The daemon loaded whatever the key pointed at, so pointing
both keys at one file cost recall with no error and no log line, which reads as
ordinary drift rather than as a misconfiguration.
validateVoice now refuses it at load. Both paths are cleaned and made absolute
first, so "./m.onnx" and "$PWD/m.onnx" are one path, and then compared with
os.SameFile, which catches a copy that is a symlink or a hard link. A path that
does not stat is left to the loader, whose error message is better than this
check can give.
Refusing to start is deliberate and it differs from the loader's treatment of a
broken weights file, which logs and leaves the heads nil on purpose. That case
is a missing accelerator. This one is a working file in the wrong role, and a
daemon that cannot route well should say so rather than answer worse.
deploy/mavend.json points the two keys at different files, so the live config
still starts.
embedder.heads_path is empty by default and deploy/mavend.json sets
it. A missing or broken weights file logs and leaves the heads nil,
because refusing to start over a routing accelerator would trade a
working box for a better one.
TestONNXRoutingHeads is the same cascade TestONNXBaseline scores with
one arm added, so the two are directly comparable. It also checks the
Go tokenizer against the Python one, since the heads were trained
through transformers and are read through a hand-written tokenizer: a
mismatch shows up here as a score below what Python measured on the
same weights, and nowhere else. That is how the reversed word pieces
were found.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013ptwopxyo3Z2kwFckHkLvN
Three tangled defects, fixed together because each one hid the others.
DefaultActMatcher matched an exact English prefix and internal/tool.Matcher
delegated straight to it, so no Russian utterance could reach a tool: 55 of the
69 lines in models/seeds/act.txt routed to IntentAct and fell to proposeGap.
Tools now carry spoken aliases from deploy/mavend.json, matched as exact leading
tokens, longest phrase first. Config data, not a stem pattern in code. The
comment claiming "the production matcher is fuzzy" was false and is gone.
Seven lines were exact duplicates inside models/seeds/query.txt, each one a
second identical vector double-weighting its region.
"как дела у сервера" carried both a query and a system label. It leaves
system.txt, because replySystem's stats arm answers "системная статистика пока
не подключена." and always did. The mode inventory records that shape as
act.tool.hoststats rather than a system mode.
Fixture unchanged at 69/91, and it cannot see any of this: no host-stat case and
no Russian act in it. TestActMatcherAliases is the coverage.
docs/evals/2026-08-06-russian-acts-reach-tools.md has the numbers.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0117tgnmbgZpHVV3XSNw8Qua
WorkerConfig, TtsConfig, EmbedderConfig, WeatherConfig and ToolConfig. None of
them is reachable except through a voice block, so they read better beside it.
The VoiceConfig struct only. Its defaults, normalise/validate arms and
UseLLMRouter follow in the next commit, because the pre-commit cap counts a
pure move at twice the block size.
The RouterThreshold doc drifted: it said the default was 0.35 and that 0.0
meant permissive. DefaultRouterThreshold is 0.55 and applyDefaults replaces
anything <= 0, so permissive was never reachable. Fixed the comment, not the
code.
claude