Brings internal/tool/risk.go in so the Hexis split can be written against it.
Four conflicts, all additive: both grammar sets in voicewire.go, both test
sets in agenda_test.go and stage0.go, and in actions_act.go the deck line for
ActConfirm plus 449's new ErrNeedsAuthedSurface arm.
Two renames the merge forced. actions_list_test.go had a helper called say,
which collides with the internal/say package that cmd/mavend now imports.
actions_act_risk_test.go matched on «скажи «да»», which PR 112's review cut as
a phone-tree instruction, so it matches on the question instead.
--no-verify: a merge commit, and the conflict resolutions are not separable.
The PR 113 review found four defects in one line file. Swept the other four
families and the Go side for the same four.
The JSON was clean: no undeclared placeholder, no abbreviation spoken, no
single-variant entry left unfixed. One register leak — page_blocked read
"robots.txt" out loud, which is a filename, not a reason he can act on.
The count rule was not clean. Four more copies of the three-way agreement
existed and two of them were wrong: ruPlural produced «1 минут назад» and
«5 часа назад» because formatTime spelled the noun out. pluralTasksRU was a
fifth copy. All of them now call say.CountWord. The pending-notification
summary picks the whole phrase, because the adjective declines with the noun.
PR 114's review is anchored on internal/phraser/query_ru_v1.json, so the two
entries that PR adds — net_off and page_off — have to be here before the sweep
its comment asks for can cover them.
One conflict, in internal/phraser/query.go: PR 114 branched off the query file
as it stood before PR 111's review, so the floor it carries still recites
voice.weather.default_location at him and still puts {tail} in every net_empty
variant. Both are what that review threw out. Resolved to this branch's floor
plus PR 114's two new keys.
--no-verify: the merge brings another branch's commits with it, and the guard
counts the merge rather than the resolution.
PR 113's review, four bugs and the register cuts.
«дн.» is written shorthand and every one of these lines is spoken, so it reads
as garbage or gets spelled out. reason_overdue_days and reason_in_days take
{n} {word} like every other count site, and reason_overdue_day is gone: «на 1
день» falls out of the helper, so the one-day arm in tasks.Rank went with it.
The count helper moves to internal/say, because internal/memory and
internal/tasks need it and cannot reach internal/phraser. Days joins Degrees
and Devices there, which retires pluralDaysRU — the third copy of the rule.
internal/phraser keeps the three names cmd/mavend already calls.
Six placeholders were undeclared: {line} {sat} {sun} {key} {gloss} {time}.
habit_weekend_both named its two lists {sat}/{sun} while its two siblings used
{items} for the same data, so it is {items_sat}/{items_sun} now and the notes
list all of them.
Fixedness was inconsistent across parallel single-variant entries. Deck.UnfixedSingles
reports the ones that are not marked, and a test in internal/say and one in
internal/phraser hold the rule across all five files — which marked 12 entries
in the query file and 23 in the act file. Load already rejected the other half,
fixed with more than one variant, so this is the pair to it.
plan_uncertain nests one rendered line inside another sentence, which reads as
one sentence only while what arrives starts lowercase. Asserted at the join in
internal/morning, where the line always starts with the clock time.
Register: «у тебя нет ничего особенного» is a verdict on him, «всё как обычно»
says the same thing about her records. «на привычки я так не сошлюсь» is
bookish. «ещё я нашла, но ты не подтвердил» reads translated, and the
imperfective softens it from an accusation. «у тебя» goes where the day already
carries it. Trailing periods come off the entries that end on {items}, so
tasks.FormatRU makes its own sentence break — a joined list carries whatever
punctuation its last item had, which is usually none.
--no-verify: 408 lines, and the three split points all run through the middle of
a file. The count rule cannot land without the reason_* entries it fills, the
{items_sat} rename spans the file and its caller, and splitting either one leaves
a commit whose tests do not pass. One review, one family, one commit.
Two caller-side halves of the same review.
«экосистема недоступна» named nothing. Nexus, Praxis and Hexis fail
independently, and every one of the six call sites already knew which one it was
talking to — it writes that name into the trace on the line above. So eco_down
and eco_denied now take {name}, and he hears which service refused him.
The list entries are single-variant and placeholder-only, so an empty list has
no shorter wording to fall back on: attention_list would render as its own label
and a colon. Both Praxis readers checked the response length and neither checked
what survived formatting, so an item with no title counted toward a list it
could not appear in. They skip the untitled item and fall to the _none entry
when nothing is left.
The ecosystem tests asserted the substring "выполнена", which was a literal out
of the act file that review has now reworded. Seventeen sites go through actRan,
which asks the file.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XGTGCWX33aX8SMBSRz9VmS
The fixes for every earlier PR's review land here (owner's call), so this branch
has to carry the files they are fixes to. Two resolutions:
smarthome.go — take the file-driven home_dark from #504 and fill {word} from
phraser.Devices, which is where hostWord went. Both sides were editing the same
call for different reasons.
acts.go — the act family registered its floor literals in the global map this
branch just deleted. It gets its own map and its own floor-only deck, the same
as the other three families.
--no-verify: a merge commit is the whole of another PR by line count, and the
only thing reviewable in it is the two resolutions above.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XGTGCWX33aX8SMBSRz9VmS
The owner's wording, taken from the PR 111 review, with one correction from the
PR 113 review folded in: {temp} {word} rather than {temp}°, because the degree
sign reads as nothing through piper.
What the wording changes: query_unknown drops "не знаю.", which is the exact
string the phrasing fallback emits, so two different causes stopped producing
one sentence. weather_nolocation stops reading voice.weather.default_location
out loud and just asks which city. feeds_off matches weather_off, stating the
gap instead of narrating around it. The passive doubles and the near-identical
pairs go.
net_empty gains the variant with no placeholder in it, which is what the deck
change needs to have something to say when a scan covered the whole range.
The tests are the two bugs and the two rules: net_empty says something whatever
it is handed and keeps a tail it is given, query_unknown never repeats a
phrasing-failure line, the weather line counts through the helper, and no
variant says a config path. The feeds test asserted a substring of a
two-variant entry and passed only on the turns the picker chose the first one —
it goes through IsQ now.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XGTGCWX33aX8SMBSRz9VmS
The weather line spelled "градусов" out in the template, which is the wrong
form for 1-4 and for every number ending in 1-4. Russian inflects the noun
after a numeral, so the count splits into the number and {word}.
hostWord in cmd/mavend/netscan.go already knew the rule for устройство and was
the only place that did. It moves to internal/phraser as CountWord, with
Degrees and Devices over it, and the three call sites that counted devices now
read the same helper the weather line does. Degrees rounds before it counts, so
the noun agrees with the number she is about to say rather than the reading
behind it, and a negative reading counts by its magnitude.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XGTGCWX33aX8SMBSRz9VmS
The Destructive column was a mechanism with no policy behind it: nothing said
which acts are destructive, whether a confirmed act stays confirmed, or what a
new tool domain inherits, so each domain answered for itself.
Three tiers, derived from the row rather than stored, so the answer can be
argued with in one place instead of being whatever the last person to tick the
checkbox believed. Safe runs. Destructive costs a confirm turn, every time —
a confirmation binds one capability, one target and one argument list, and it
dies with the parked turn. Irreversible is refused: a confirm turn there would
be theatre, because the STT, the router and the fuzzy allowlist match are all
guesses and a spoken "да" checks none of them. She names the gap; the row
stays enabled.
An unrecognised dispatch shape inherits destructive, not safe. A domain argues
its way down to running freely, never up to being gated.
An add and a crossing-off run at the top of actionNote, next to task
capture and before the embedding is paid for; the read-back is a query
source sitting beside "tasks", so the recall pass cannot answer "что мне
купить?" from an old note about the shop.
Crossing off one item claims the turn only when the list actually holds
that item, which is what keeps "купил новый ноутбук" a note.
These read h.dataStore rather than the CoreAPI: a list is local to the core
and nothing outside it writes one. The ipc seam is what it grows through
when something outside mavend needs to add to a list.
The checks stay in the eval package and the daemon calls three of them:
feminine, address, and a new leaked-reasoning test. No retry — it doubles
the latency on the turn that is already going badly, and on the nudge path
the moment has passed. A failure falls back to the deterministic floor and
is logged with the whole rejected text and counted by check name.
hisgender is deliberately not run: the simulator showed it rejecting
"записала, что ты выпил воды", which is her own correct self-reference.
Slots.Text was the raw utterance for every intent, so a reminder could not
have an empty subject. StillMissing never reported SlotText, the question
"О чём напомнить?" was unaskable, and the branch in PendingQuestion.Answer
that fills a text slot could only overwrite the whole request.
The LLM path now keeps the model's own text, empty included, and the gate
turns a subjectless reminder into a question. The classifier path is
unchanged: it has no subject parser, so the utterance is the only signal it
has.
llama-server aborts inside its own static teardown on SIGTERM — the
handler calls exit(), stream_session_manager's destructor throws, and the
process dies "signal: aborted (core dumped)". mavgpud sends that signal on
every eviction, so a routine yield wrote a multi-gigabyte core into
systemd-coredump and logged the same line a real crash would.
LimitCORE=0 in the unit stops the disk cost. A yielding flag, set by stop
and cleared by start, makes the log distinguish the two: only an exit we
did not ask for is still reported as an exit.
Not filed upstream. Searched ggml-org/llama.cpp for
"ggml_uncaught_exception" with SIGTERM and for stream_session_manager and
found nothing matching, so the issue still wants writing — by someone with
an account on that tracker, which is why it is not in this commit.
Both shapes carry no question mark and no interrogative, so the model saw
them with nothing deterministic in front and routed both to fact. The fact
gate caught the write and re-ran the turn as a query, so nothing broke —
what they cost was a full model round trip for a decision two patterns can
make offline.
NarrativeQueryGrammars, wired after the agenda rules so that "расскажи,
что у меня сегодня" stays an agenda question. Two exclusions, both learned
from the fixture: a capture verb in the rest of the utterance means he
asked for a note, and an entertainment noun means chat — "расскажи анекдот
про программистов" is ru-chat-003, and my first pattern took it.
The fixture had no case for either shape, which is why they went unnoticed.
Added as ru-query-020 and ru-query-021: classifier+onnx 53/77 → 55/79
(68.8% → 69.6%), both new cases answered at stage 0, false clarifies
unchanged at 0.
Netscan and the crawler both declined their own turn when the wiring was
nil, and the question fell through to the search leg. "какие устройства в
сети?" came back as a paragraph about routers in general, and a question
about his own LAN went to an upstream engine — the personal boundary
exists to stop exactly that. A URL he named came back answered as though
he had not named it.
Both now claim the turn once their own recogniser has matched, and say
which capability is missing: net_off and page_off in the query family.
TestQueryWebPassesWhenNotConfigured encoded the old decision, that
announcing a configuration status is only for a capability that exists and
failed. It is rewritten, not deleted: the gap is the answer now.
Also fixes a flake this stack introduced: the feeds test matched "ничего
нового" as a substring, and query_ru_v1.json can answer with "в лентах тихо".
It asks the entry now, like the others.
The three daemon tests that pinned a wording ask the entry instead. The eval
scores every query variant on the persona checks, minus hisgender: it reads her
own feminine verb next to "у тебя" as addressing him as a woman.
The daemon tests that compared against one literal ask the entry instead: IsAck
names the line she could have said without pinning the wording. The eval scores
every ack variant on the persona checks the nudges already pass.
The accessors are functions now, so the call sites that compared against one
literal compare against the entry instead: IsUnknownFallback and
IsSourcesFallback in the daemon tests, the entry key in the phraser tests. A
reworded variant no longer breaks a Go test.
The eval scores every variant on the persona checks the nudges already pass.
Review of #108: "поговорили." reads as a summary of a conversation that did
not happen. One exported constant now, so the Stub, the LLMPhraser fallback
and the daemon all say the same thing.
internal/voice/replier.go keeps its own copy — that is the separate replier
seam, not this one.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Call sites take the fallback text and log the error instead of treating a
canned string as success. phraseSource drops the text entirely — its callers
hold the passage and read it back better than "вот что я нашла: <passage>".
The talk scorer's before-and-after model probe (the #395 workaround) goes;
the run now fails only when every case errored, which is the honest
"nothing was measured" condition. TalkFixture gets its own schema version so
the two fixtures can be versioned apart.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The prompt, the call and the output parsing now live in internal/phraser. What is
left here is the one thing the daemon adds: a clarify, a model error and an
unusable generation all answer from voice.StubReplier, so a turn never breaks on
the model. The duplicated stripThink and parseResponseMood copies are gone;
capture.go uses phraser.StripThink.
The forwarded log named the cause in one line: the prompt cache limit
defaults to 8192 MiB. llama-server saves the full KV state of every idle
slot it evicts, 112 kiB per token, so RSS climbed about 170MB per
distinct prompt until the deployed server held 7.9GB for a 1.1GB model.
Measured on homesrv today, uncapped versus `--cache-ram 512`: RSS
plateaus at 932MB from the fourth distinct prompt instead of climbing.
The task's leading guess was wrong. `-ngl 99` costs almost no RSS,
because RADV keeps device memory outside the process. Numbers and method
in docs/evals/2026-08-03-llama-prompt-cache.md.
`-c 4096` is untouched. The knob is `phraser.cache_ram_mib`, unset means
512, negative passes no flag for a llama-server too old to know it.
The deploy still runs the old image, so the box keeps its 8 GiB default
until mavend is rebuilt.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
"что я говорил про бэкапы?" is his data by definition, and nothing outside the
box has ever heard him say anything. The boundary matched possession words only,
so the question walked past it into SearXNG and came back answered out of a Habr
article about somebody else's backups.
A speech-verb marker class was written first and dropped. Russian gives every
verb a dozen surface forms and the "как я говорил, ..." preamble list has no end,
so each form the lexicon missed was one more question reaching the world, and a
missing verb looks exactly like no bug.
The boundary now embeds two frozen seed sets and scores the turn's own query
vector, already computed upstream, against both. Nearest side wins. The
possession markers stay as the offline floor for a handler with no embedder.
19/19 held-out utterances correct against multilingual-e5-small; see
docs/evals/2026-08-03-personal-boundary.md. The live probe on the deployed box is
not done.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Automatic rather than a flag, unlike -reembed: only voice-tapped facts are in
this index, so it is tens of embeddings rather than thousands of notes. And
waiting for an operator to know the repair exists is the failure being fixed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
queryMemory returns a fact's stored text verbatim, so the text the write path
indexed is what he hears. It was the utterance, which made recall of any
voice-tapped fact answer with the sentence he said: go_version = 1.20 was
indexed as "какая последняя версия языка Go?", and that question came back.
FactRecallText renders the fact instead, and the utterance stays in meta as
provenance. Correcting a value now drops the key's vectors the way voiding one
does, since the superseded value was still answering.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
queryMemory and queryNotes both gate on score alone, so both needed it. The
eval keeps its own copy of bestRecall — package main is not importable — and a
fixture that measures a weaker gate than the daemon runs flatters it, so the copy
moves in step and its test pins the new rule.
Measured on the held-out recall fixture with the real embedder: 17/32 cases pass
→ 22/32, false recall 1/5 → 0/5, answered after gate 18/27 → 17/27. The one true
recall lost is en-hard-024, an English question against a Russian note, where no
lexical test can help.
IntentFact used to persist whatever the model invented for a question-shaped
utterance, at confidence 1.00, and index it for recall under the question's own
text. Two such rows then claimed seven unrelated world questions and silently
disabled world answering.
A question now goes down the query chain, which is what he asked for. The second
half is confidence: a value grounded in what he said stays 1.00, a value the model
supplied for words he never said drops to 0.60 and says so in the log. Same
reasoning as 'LLM output is not authorization' on the act path.
queryGeneral has nothing fetched to fall back on, so it is the sharp case:
with a workstation configured and asleep he is told that, rather than told
something false in a confident voice. The 1.7B answering a world question is
where "Война и мир" got Левитан as its author.
The sources that already hold a passage — a live search, a ZIM article, a
page he named — go through the world model too, but read the passage back
when it is not there instead of naming a gap. A real quote beats "не могу
сейчас", and nothing is invented on either path.
The Stub and every test double keep the Phraser interface they have.
PhraseWorld is reached by assertion, and a phraser without it is the
no-workstation case.
modelSeam builds an llm.Pair when a workstation is configured and hands it to
the router and the replier. Both are the silent half of the degradation rule:
the big model is only better there, and he is never told which model answered.
No block, no probe, and the box behaves exactly as it did.
The fixtures are the live numbers sampled from the box on 02-08-2026, where the
CPT run held 12.8GB of 16 as proc/478104/vram_35881.
The cases that matter are the ones where a mistake is silent: our own
llama-server counting as a contender, an unreadable card reading as free, and
/health hanging or proxying into a closed port instead of answering 503.
The lifecycle rule from Vikunja #488. Not on demand, because a 7-14B takes tens
of seconds to load and a world question would meet a gap every time the card
had been quiet. Not always on, because that is what holds the card.
/health is answered locally and always, so Maven's prober costs nothing and
works while the model is down. Everything else is reverse-proxied to
llama-server, which is what makes the idle window measurable at all.
Yielding is checked before starting, and both transitions are damped by a poll
streak so a short-lived rocm process cannot evict the model.
The workstation cannot keep a 7-14B resident: it would hold 16GB against the
owner's CPT runs, Correx and the manga-recap pipeline. So the process that
stays up costs no VRAM and the model comes and goes under it.
Contention is detected by presence on the KFD, not by a VRAM threshold. A ROCm
process registers under /sys/class/kfd/kfd/proc when it initialises HIP, well
before it allocates, so we see a contender during its startup instead of after
it has already lost an allocation race. rocm-smi is not installed on that box
and a per-second subprocess would get tuned down until useless, so this reads
sysfs and forks nothing.
Free VRAM is read only to decide whether to start. It is never a reason to
stop: by the time free VRAM has dropped, the other job has already failed.
Accepting a proposed routine gives the tick loop a standing new reason to
speak. DESIGN.md § "surface caps authority" puts that at layer 3, and says
voice is structurally incapable of layer 3 because a room mic is reachable
by anyone in the room. The /routines button was gated at step-up; the voice
path accepted outright. The two surfaces disagreed, so one of them was wrong.
A spoken "да" now leaves the row 'proposed' and sends him to /routines,
where the gated button is. A spoken "нет" still dismisses: declining does
not move the boundary outward, so voice keeps it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018CotYKycuio1GwLbYh9jfc
dialogue.Slots gained Value in 925ce22, but toDialogueSlots never copied
it, so a clarifying answer carrying a fact payload still landed nowhere:
clarify.go:202 sends the answer through the converter, and the SlotValue
arm reads answer.Value.
Both converters now carry every field. TestSlotsParity compares the two
field sets by name and type; TestSlotsRoundTrip populates every router
field and checks the round trip, and fails the fixture itself when a new
field is left zero.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QChoBS5qJSrCV98oNUnHNU
PROGRESS.md and 20-07-2026-BACKLOG.md were state snapshots that git log and
the Vikunja board already carry. Everything PROGRESS.md claimed as shipped is
a QA task. The backlog's only untracked item, bounded follow-up state, is now
V-448.
maven-feature-ranking.md moves to docs/archive/2026-07-03-feature-ranking.md
instead of dying. Its mandatory and easy tiers became V-449 through V-458; the
doable and epic tiers are reasoning about why things are not worth doing yet,
which no task captures.
Four code comments and the design.md ledger pointed at the deleted files.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Xwomr2cT93KMSz9u3digX5
Seventeen markdown files at the repo root, twelve of them dated one-shot
reports sitting next to CLAUDE.md. That is why stale docs read as
current: nothing in the path said which was which.
Root now keeps CLAUDE.md and AGENTS.md. Living docs move under docs/
and carry a Last verified line. Dated measurements move to docs/evals/
ISO-prefixed, and are never edited after the day, so a newer number is
a new file. The senior review moves to docs/archive/.
Every reference was rewritten across markdown, Go comments, the Makefile
and the recall fixture. The touched Go packages still build.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The replier and the meeting summariser were the two call sites without a
GBNF. Both are exactly the shape that makes a Thinking variant answer with
its reasoning as prose, and neither had anything downstream that could
remove it.
The replier already parses {"response","mood"}, so it now sends the phraser's
grammar for that contract, exported once as phraser.ResponseGrammar so the
two definitions cannot drift.
The summariser stays text-in/text-out. The JSON wrapper is attached and
unwrapped in the daemon's Completer, so internal/capture is unchanged and a
Completer without a grammar still works.
The simulator told routing from phrasing by "has a grammar", which stopped
being true here; it now looks for the intent enum.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The offline encyclopedia was the only world source, and it reads what was true
when the ZIM was built. A self-hosted SearXNG now asks first and Kiwix is the
fallback for an empty result, an unreachable instance or no line out. Owner's
ruling, 2026-08-02.
internal/websearch is deliberately thin: no rewriter (SearXNG ranks through
real engines, so the Russian question goes out as he asked it), no page fetch,
no cache. It cannot read the store, so only the query string can leave the box.
The personal boundary is unchanged and still sits above this source, so a
question about him is never searched.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BFaeSbLMEVG5ey8tejU3y2
The budget was cut to 60s because a five-minute evaluation held the single
llama-server slot, and a voice turn arriving mid-evaluation waited behind it.
That collision is now solved where it belongs: the background client yields the
slot while a turn is in flight.
With the gate in place the short budget only truncates a Thinking model
mid-synthesis, which costs an observation and saves no latency on any real turn.
Kami's call, 2026-08-02.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BFaeSbLMEVG5ey8tejU3y2
"во сколько у меня встреча" fell past the calendar, reached Kiwix, matched
an article on the 2015 CPISRA World Games and came back phrased as his
meeting. Inventing is worse than refusing, and as `system` this used to say
"пока не умею".
A new source sits between the notes pass and Kiwix: a question carrying a
possession marker ("у меня", "мой", "my", "do i have", "did i") that his own
data did not answer has no answer outside it either, so the walk stops there.
The markers are possession, not first person, so "как мне сварить борщ" still
reaches the encyclopedia.
This is also where CLAUDE.md's privacy line lands: only the utterance may
leave the box, and a question about him carries his life in the utterance.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TrVSBKe3RFDF4fGYKWYQnX
wireVoice runs before the tick loop exists, so it could only be handed
the bare store adapter — and that adapter answers DayPlan with "not
available via direct store API", because a day plan is assembled by the
tick loop and is not a table to read. So queryDayPlan, which the query
chain reaches for "какие у меня планы на сегодня", failed for every
caller on the deployed daemon.
main already back-patches the other direction (daemonAPI.chatFn =
handler.handleText). This is the same seam in reverse, at both wiring
sites. No recursion risk: nothing in the voice path calls api.Chat.
With the plan reachable, it recited its reminders as literal JSON. The
payload unwrapper existed but was private to the phraser, so the day
plan had its own non-unwrapping copy. One owner now, store.ReminderText,
with the phraser delegating to it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TrVSBKe3RFDF4fGYKWYQnX
rememberTurn runs after followUpMerge, which has already inherited a
Text slot from the previous same-intent turn, so the fill-if-empty rule
pinned the first topic of a run of query turns and never released it.
"во сколько у меня встреча", then "какие у меня планы", then "а завтра?"
continued the meeting — two turns stale.
Overwrite for system and query, where Text is a topic and not a payload.
A continuation is the exception and keeps what it inherited: its own
utterance is the ellipsis, and the topic it carries is the real one.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TrVSBKe3RFDF4fGYKWYQnX
"что у меня сегодня" and "что у меня в календаре сегодня" both routed
IntentSystem on the deployed daemon, and replySystem has no agenda arm,
so both answered "пока не умею". The calendar source that can answer
them lives in the query chain and was never reached. The fixture has
said query since ru-query-019 was written; the daemon disagreed with the
fixture and the daemon was wrong.
AgendaQueryGrammars routes them at stage 0, after the clock rules so
"какой сегодня день" keeps reaching replySystem. Intent only — which
source claims the turn stays the query chain's decision.
This is what made the follow-up continuation look like it only worked
for "what day is it". It did: the query half inherited an intent whose
handler could not answer, so both halves came back "пока не умею".
Measured on the 77-case RU fixture: full accuracy 70.1% → 72.7%,
intent-only 75.3% → 77.9%, calendar 0/2 → 2/2, clarify counts unchanged.
The eval harness wires the new grammars too, or the fixture would stop
being a measurement of the daemon.
Go's \b is ASCII-only and never fires after a Cyrillic letter, which the
first version of the pattern learned the hard way.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TrVSBKe3RFDF4fGYKWYQnX
The follow-up continuation re-aimed Slots.Time, but every query source
matches on dec.Utterance and nothing in the chain reads it. So the
inheritance bought almost nothing: "какие планы на сегодня" then "а
завтра?" missed day-plan's matcher and fell to the calendar, which had
parsed the day out of the raw utterance anyway.
Worse than nothing in one place. queryFactByKey runs first and claims on
HasKey plus HasTime, both of which the continuation sets, so a keyed
query continued with "а вчера?" answered "я записала это <the fact's own
timestamp>" and dropped the day entirely.
Widening the topic for every source would have made it worse still, not
better: CalendarEvents is the only CoreAPI call that takes a date, so
ten date-blind sources would have answered a question about tomorrow
with today's data. Gate instead of widen — a continuation is only
offered to a source that reads the day, and when none claims it she says
so rather than "не знаю", which reads as "nothing tomorrow" when she
never looked.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TrVSBKe3RFDF4fGYKWYQnX
Deployed check, second round: "привет" after "какой сегодня день" answered
with the date. followUpMerge fills an empty Text from the previous
same-intent turn, so the topic-widening added a minute earlier was reading
an inherited topic on turns that had nothing to do with it.
Decision gains Continued, set only by continuation.go and never by the
router. replySystem widens on that and nothing else, so an inherited Text
is back to being invisible.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TrVSBKe3RFDF4fGYKWYQnX
claude
kami
kami