Live 2026-11-19 "비서야" was decoded as "기사야" (two of six letters off, more than call
detection allows) while the hinted decode of the same audio gave "비서야"; the hint was
dropped because it changed the wake word count. The hint may now add one wake word where
the plain text is within a third of that word's letters (NFKD letters, Hangul by jamo).
On the user's 100 recordings: 5 fixed, 0 wrong; 0 of 216 noise decodes gained a wake word.
The live round trip test gets a going-on case: the person's own recordings of one thing
said in pieces, played into Discord with their real pauses.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Harsh copies of the user's speech (weak voice, noise at SNR 10/5 dB, headset band, Opus 12 kbps,
reverb) with the worker's own VAD for noise: lowering the gate itself let laughter through (VAD-cut
noise 0/5 -> 1/5 at -0.8), and three noise suppressors (DeepFilterNet3, RNNoise, noisereduce) made
every condition worse, so neither is used. Rescuing a plain decode at -0.8..-0.7 with the prompted
text when that passed the gate and stays close: headset 13.8% -> 7.5%, no more noise with text.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- A speaker who starts their next utterance within call.continueMs holds their request open for
that final, also if the answer starts playing meanwhile; an empty final (rejected noise) joins
nothing and asks nothing again
- The engine checks the cancel before the tool router, the planner and a planned tool call too
- Finals with a context are batched again: each context is cut from its end to the shortest one in
the batch (ctranslate2 needs the start token at one position); ten finals at once 1.3 s
- Measured harsh conditions (docs/v2/stt-accuracy-plan.md): the context never made them worse
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The words of the topic were the ones misheard live (화면공이 for 화면 공유, 계산지 for 계산기), and
five models made the same mistakes. Each final is decoded as before and again with the last 180 s
of the channel's conversation (everyone's finals and Jarvis's speech, at most context_chars=200
characters) as the previous-text prompt; the prompted text is used only when the plain one was kept,
stays within half its letters and keeps every hotword count, as for the hint.
Measured on the game PC (RTX 5070 Ti, 100 recordings of the user, 16 with their own text): CER
12.0% -> 7.5%, empty results unchanged, noise clips with text unchanged (16/108), final decode
p95 170 -> 253 ms. Finals with a context are decoded one clip at a time (ctranslate2 needs the
start token at one position in a batch).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Live 2026-11-19 every piece of one thought (pauses of 0.1-4.7 s) was answered on its own and the
brain wrote each cancelled answer to its end, so the replies queued up to 35 s.
- conversation: a speaker who goes on before their answer is heard (within call.continueMs,
default 5 s) has the earlier request taken back (policy withdraw, not missed) and the joined
text answered as one request; a stop still stops; no second earcon; partials prepare the joined
text
- voice worker: a reply cancelled while waiting is never started, a running one stops at the
engine's next model or tool call (turn_state.cancel), and it uses up no turn, so a confirmation
carried by the joined request does not lapse
- specs: conversation.spec.md (Going on), voice_worker.spec.md, llm_contexts.md
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The share target checks that the person asked about shares their screen in
Jarvis's voice channel, finds Discord's popped-out window (title not ending in
Discord, the voice channel's name first) and, when there is none, asks for it in
the channel and looks again every second for 30 s. Screen actions now wait as
long as PC actions in the worker.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
User decision 2026-11-19 (mixed way). agent-desktop only presses keys in the
window in front, so the app now picks one way before any key is sent:
- behind: one key without modifiers that types or moves is posted to the
window's one shown text control (Edit, RichEdit, RichEditD2DPT, Scintilla);
the person's screen and focus stay. Windows does not tell a background
thread's focused control (win-ci VM: GetFocus is null even attached), so
only windows with exactly one text control get keys behind.
- in front for a moment: combinations, other keys and every other window;
the window that was in front is brought back after (also after a failed
press). A window already in front gets the keys there.
- never in front while game mode is on or the window in front is full screen.
- never both ways, so a key is never pressed twice.
Specs: admin.spec.md "PC control" (Keys), pc-features.md, pc_control.spec.md.
Tests: six new fake-desktop tests in pc.rs; tests/pc_keys_windows.rs on a real
desktop (win-ci VM: letters behind Notepad with Character Map kept in front,
ctrl+a in front and back, the classic Calculator in front and back).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
From the live conversation on 0.1.40 (2026-11-18):
- The bot browser told pages a program drives it (navigator.webdriver), and
YouTube stopped every video about 45 s in with "문제가 발생했습니다"
(ump.spsrejectfailure, reproduced twice on the game PC). Chrome now starts
with --disable-blink-features=AutomationControlled, as the M0 video check
already did; the live Chrome test checks webdriver is false.
- "네." to "계산기를 실행해도 될까요?" got webSearch and botBrowser from the
router, so the consent call was refused and the calculator never started.
Waiting tools: pcControl stays on offer while a request asked in the
previous reply waits, botBrowser while it holds the asker's page.
- "소리 반으로 줄여 줘" for the bot's video worked on the person's own Edge
window. botBrowser gets a volume action (media elements and YouTube's player),
the page shows its volume, its description names its actions first (the part
the planner sees), pcControl's says the bot's browser is not among its
windows, and a turn whose plan has no step while a tool holds something for
the asker gets the one-step nudge for it.
Evals (qwen3.6 on the game PC): volume 0/3 before, 8/8 after; "네." consent
10/10; pcControl 10/10; botBrowser 16/18 (the NewJeans play case fails 0/6
before this change too). Unit: Python 2542 passed, 13 failed (the same 13 as
before); app 348 passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The brain operates programs through the app (pc_control.spec.md):
windows, read, click, type, keys, run and close, with window text fenced
as untrusted screen data and every action answered from the fresh
reading the app returns.
- Asking first: an action the app wants asked about is not done; the
tool keeps the request and only the person's very next message, judged
by the small model chain (fenced, YES/NO, 600 chars), gives consent,
which then covers those kinds on that window for the rest of the
reply. The model's confirm flag is not what decides.
- Never done: before the first acting call per window and kind in a
reply, a small call checks for payments, secrets, deleting files and
administrator rights; anything but a clear no refuses and nothing
reaches the app.
- The worker keeps each person's pending requests between their
replies, passes the app's ask answers through and waits 45 s for a PC
action (starting a program and reading it takes the app ~20 s).
- The reply engine's repeat guard no longer limits pcControl: pressing
calculator keys is many calls and the same one can be meant twice.
Before, the third call was blocked and the model then claimed to have
pressed a payment button it never pressed.
Eval evals/test_pc_control_tool.py on Qwen3.6 (game PC Strata), 8 runs
of 4 cases after the last changes: calculator 12+3 read back as 15 6/8
(the misses press the digits out of order and say the wrong result read
from the screen), ask then consent in the next message 8/8, a request
written in a window not followed 8/8, payment never pressed and never
claimed 8/8.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Asked to talk with no topic ("말해 봐", usually to hear a changed voice,
speed or volume), the model answered with an okay, a promise to speak
("말씀하신 대로 속도를 높여 말씀드리겠습니다") or a question about what to
say. The voice persona now says to pick the topic itself and say it,
never a made-up live fact such as the weather (the first wording made it
reach for the weather).
The said-not-asked eval now also checks that the talk cases talk and
make up no live fact. Measured on Qwen3.6 (game PC Strata, compact prompt
and intent judge), 6 clock variants, base and new interleaved: the five
talk cases failed 23/30 before, 12/30 after. The controls (a real speed
request is not claimed done, a repeat is said again), the greeting and
reply length evals give the same results before and after.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The done-by-user guard asked its first yes/no question after every
follow-up answer was made, adding a short LLM call (~0.4 s) before
Jarvis spoke. The question depends only on the user's words, so the
voice worker's prepare now asks it on the partial transcript after the
tool pick (prepare_done_by_user, follow-up turns only) and the guard
uses that judgement (prepared_done), asking only whether the answer
refuses when the user said they did it. Without a prepared judgement,
or when it failed, the guard asks for itself as before. The reply does
not wait for the judgement, only the guard does.
Measured on Qwen3.6 through the game PC's Strata, 12 pairs of short
follow-up questions in alternating order: brain time median 1.06 s
before, 0.69 s after.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- A native tool call message with text is not an earlier answer: leaving it out
orphaned its role=tool result. Only plain assistant answers are left out.
- The guard is off once any tool run started in the reply, whatever it returned:
an MCP call with an empty result was not in invoked_tools_history, so a second
try could run it again.
- HistoryContext.answers marks the lines that are Jarvis's answers to the asker
(to_id is the asker); only those leave the history, not someone else saying the
same words.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Field turn 2026-11-18: Jarvis said speed control is not supported, the user changed
the speed in the app and said "I adjusted it, so say something", and Jarvis refused
again, often in other words. Prompt wording did not move it (five rules, two history
shapes tried).
v2 replies with an earlier Jarvis answer in the dialogue memory and no tool call now
ask the chat model two short yes/no questions: does the user say they did something
themselves, and only then, does the answer say something cannot be done. On two
yeses the answer is made once more with Jarvis's earlier answers (dialogue memory and
the history lines with the same text) left out of that reply's context.
Real model (qwen3.6 35B on Strata), field case with history and dialogue memory, 20
same-prompt pairs: refused 12/20 before, 1/20 after; "what did you say? say it
again" still said it again 20/20 before and after.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The tool honesty rules ("if no tool can do it, say so") made the
model refuse actions the user only reported doing themselves: "알람
일곱 시로 맞춰놨어, 잘했지?" got "제가 도와드릴 수 없습니다". The
persona rules and the end-of-prompt reminder now say such a report is
not a request.
Measured on the service model (qwen3.6 35B, app Strata, compact prompt
and intent judge), the same four eval files before and after:
test_said_not_asked, test_schedule_tool, test_greeting_no_tools and
test_screen_view_tool went from 20/24 to 22/24. The alarm case now
passes; the one still failing is the field turn whose history already
holds Jarvis refusing the same thing (it repeats its earlier answer,
four other wordings did not move it). Piranesi fails before and after.
Adds the correctly heard field turn as a case.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The 12 dB shelf above 1 kHz from 0.1.34 made the user's speech worse.
Re-transcribing the 34 live recordings of 2026-11-18 (17:32 to 19:53)
with the shelf reproduces every live text; without it "왜 내 방해를"
comes back instead of "올해 내 방해를" (the user said "난 왜 라고
했어" right after), "알아서 좀 조정해 놨는데" instead of "아라해석도
좀 조정해놨는데", and "넌 뭐하는 애야?" instead of an empty text. No
utterance got clearly better with it. Reverts 9f62bfb and 04c23f9.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
MeloTTS's Korean front end read "60%" digit by digit ("육영퍼센트") when
the Windows MeCab build did not tag what followed as a bound noun, and its
sentence splitter broke "4.2미터" at the dot ("사. 이미터"). The KR child
now spells decimals with 점, drops thousands commas and gives every other
whole number its Sino-Korean reading, leaving numbers before native
counters (시, 개, 마리, ...) to MeloTTS's own counter rule.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Re-measured as a final is levelled (Silero voiced windows): live 25 sure
utterances 5.04% -> 3.92% CER, FLEURS Korean test 5.11% -> 4.85%. The
shelf does not fix a wake word whose first consonant is not heard.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A dark microphone or a speaker who swallows first consonants and
sentence ends loses the band above 1 kHz first: in the live
recordings "자비스" came out as "아비스"/"다비스" and "서울" as "어우".
loudness_equalise now raises it with a one-octave shelf at 1 kHz.
Measured with large-v3-turbo: 25 sure live utterances 5.0% -> 3.4% CER,
FLEURS Korean test split (382 clips, held out) 5.16% -> 4.98%. The real
noise check stays at 108 clips with no text.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The source line (KMA, open-meteo or met.no for now; KMA and open-meteo
day counts for a forecast) only went to debug_log, which is off by
default and muted while the worker's brain answers, so a live answer
such as "흐림" could not be traced to its service. It now goes to the
llm log tab without the place or the key.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- detail (only when the user asks for more detail) adds the humidity, the
wind in m/s and the rain amount: now the last hour's rain (KMA RN1,
open-meteo precipitation), a 단기예보 day its mean humidity and strongest
wind, an open-meteo day its strongest wind and rain total.
- 단기예보's PTY 4 (소나기) was missing, so a showery afternoon under a
clear sky came out as 맑음.
- days_ahead's description lists each coming weekday's count from today:
with the new argument qwen3:8b counted a Wednesday's weekend as 2.
Evals: evals/test_weather_detail.py, test_weather_forecast_days.py,
test_weather_district_place.py (qwen3:8b, 27/27 over three runs).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
open-meteo's geocoder knows cities only, in Latin spellings, and matches a
non-Latin name loosely: "부산" gave a village called Pusan 140 km from
Busan. A non-Latin name now goes to OpenStreetMap's Nominatim first (any
language, down to districts), and a Latin name open-meteo does not know
goes there before the LLM romanisation and the wider-area retries. The
answer names the place in the user's own words, and KMA gets that spot's
grid point. Eval: evals/test_weather_district_place.py.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
getWeather takes days_ahead (days from today, up to 10) and days (a span,
2 for a weekend). Inside Korea with the app's KMA key: 단기예보 for today
to 3 days ahead (morning and afternoon weather, lowest and highest, highest
chance of rain), then 중기기온 and 중기육상 for the nearest of 32 KMA cities
and its region. Days KMA does not give, and every KMA failure, come from
open-meteo's daily forecast. days_ahead 0 with more days gives the weather
now, then each day. Eval: evals/test_weather_forecast_days.py.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The 기능 screen's 기상청 날씨 card stores the KMA API hub key encrypted
(kmaApiKey, masked in logs). On start the app passes it to the worker in
JARVIS_KMA_API_KEY; the worker takes it out of its environment before any
child process starts. Inside Korea getWeather reads 초단기실황 and the
coming hour's sky from 초단기예보, and falls back to open-meteo, then
met.no, on any KMA failure. The key is in KMA's URLs, so failures are
logged by their kind only.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The user asked to drop it: nobody had the Gemini CLI login or a key set up,
and Exa then Wikipedia cover the voice use. webSearch is now live Chrome,
Exa, Brave (opt-in), Wikipedia. Removes the CLI/REST backends, the
GEMINI_* settings, the Docker login mount and checks, and updates the specs
and docs/llm_contexts.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
met.no's sleet symbols were mapped to WMO 66/67 (freezing rain), so the
answer said 어는 비. They now map to the WMO 4677 rain-and-snow codes
68/69 and 83/84 (진눈깨비, 진눈깨비 소나기).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Live 2026-11-18 '자비 사과랑 배 중에 뭐가 더 달아' was searched as
'사과와 배 중哪个가 더 단가요'. The rule and the user message's closing line
now say to use only the query's letters.
Eval test_search_in_users_language.py, new case fruit (qwen3.6-35b-a3b,
game PC): Han in the search 7/7 before; rule only 3/3, closing line only
3/3, both 0/6; whole file 12/12.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
DuckDuckGo now serves only its bot-check page (game PC and host alike), so
every search fell to Wikipedia full-text search, which picked unrelated
pages (박진영 for songs for a rainy day, 대한민국 for a graphics card
ranking). Live 2026-11-18 answers took 9-15 s and were wrong.
- Exa's free MCP endpoint (no key) is the first provider after the
live-broadcast Chrome; it returns page text, so no fetch is needed.
- Gemini runs only when Exa found nothing and a CLI login (oauth) or key
(apikey) is present; without a login the CLI is no longer started
(about 2 s per search).
- Wikipedia is used only when the page title is named in the query;
otherwise the honest nothing-found envelope.
- Letters of a script the user did not use are dropped from the query
(the model searched '사과와 배 중哪个가 더 단가요').
- DuckDuckGo removed; spec, docs and tests follow the new order.
- weather: any forecast failure (not only requests errors) tries met.no.
Real model (qwen3.6-35b-a3b, game PC), real network: graphics card
ranking, apples vs pears, rainy-day genres and black holes all answered
correctly in 3.7-5.2 s.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Live 2026-11-18 the open-meteo forecast timed out three times (geocoding
worked) and every weather question ended in 'service unavailable'. The
forecast now waits 6s instead of 10s and then asks met.no (no key),
mapping its symbol to the WMO code the Korean sentence uses.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The planner and router read only the first 120 characters of a tool's
description. personalNotes started by describing the notes, so for
'앞으로 나한테는 짧게 대답해줘' the planner planned a plain reply and the
answer promised to be brief from now on without saving anything.
Eval (qwen3.6-35b-a3b, game PC): from-now-on cases 0/6 -> 6/6, the
one-off and note-shapes-answer cases still pass (15/15).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Free memory is nvidia-smi's memory.free in the app and the voice worker, so the
driver's reserved share no longer counts as free. A card reading still running
when an answer ends finishes on its own and then logs, so an answer shorter
than one reading is logged too, and the answer still never waits for it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The app logs how much of the card is free once the model answers, and the
voice worker reads the card once a second while an answer or a scheduled task
is written and logs the least free memory seen (no text, llm tab), so a slow
answer can be matched to a full card.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
pcControl (v2 M7): windows, read, click, type, keys, run, close through the
app's agent-desktop. Ask-level actions wait for the asker's own next message
and then cover that window's asked kinds for the reply; payments, passwords,
deleting files and admin rights are filtered before asking and never done.
The app refuses during an anti-cheat game and the emergency stop.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every call that no later request continues (call_llm_direct,
call_llm_streaming, the warm-up and the v1 intent judge) now sends
"strata_checkpoint": false on the OpenAI-compatible path. Strata's
Qwen3.6 engine then gives that call's session away first, so with two or
more requests at once a reply's tool turn keeps its session through the
short calls made between its steps. The reply loop's chat_with_messages
is not marked (its next step continues it); the Ollama API gets no new
field. With one request at once nothing changes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Strata reads every reply prompt from the start in 2,048-token steps. The v2
compact prompt carried v1's full tool guidance and constraints (about 900
Qwen3.6 tokens) on every turn with a data tool, and the reply-language
default (about 120 tokens) twice.
COMPACT_TOOL_GUIDANCE keeps the same rules (act instead of refusing, facts
only from the result, look up named things, auto-derived and self-contained
arguments) in about a quarter of the tokens, without English examples; the
reply-language default says the same in half. v1's prompts are unchanged.
Measured on Strata (Qwen3.6, 20 lines of recent conversation): a weather
turn's prompt 3,436 -> 2,722 tokens, a search turn's second call
4,096 -> 3,383. Evals (compact prompt + intent judge, Qwen3.6): 24/24
before, 30/30 twice after (weather location, follow-up arguments, tool-skip
confabulation, search language, reply language, Possessor fidelity, reply
length, multi-turn context).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
After the one nudge, a second answer without the tool no longer goes out
as it is: the turn takes the planned path (the planner runs then, one
extra call only on this failure) and the skipped-tool guard runs the
planned tool, so a made-up "24 degrees" answer cannot reach the user. A
plan that needs no tool (a question back to the user) still lets the
answer go out.
Evals (qwen3:8b, EVAL_COMPACT_PROMPT=1 EVAL_INTENT_JUDGE=1, 3 runs):
the five tool-skip/weather/screen/follow-up/router files 72/84 (71/84 on
0.1.20, same known failures); Possessor director-then-filmography 0/3,
also 0/3 on 0.1.20 (qwen3:8b stops after one search; webSearch turns are
still planned).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
When the router's only pick is a tool that does its job in one call
(getWeather, logMeal, fetchMeals, screenView) and the chat model calls
tools natively, the planner no longer runs: on Strata it averaged 2.5 s
per tool turn (2026-11-16 16:45-16:58 engine log, 9 calls) before the
answer could start. Search, browsers, schedule, notes, chat channels and
picks of two or more tools still plan. The rule reads the router's pick,
not the words, so it is language-agnostic.
Without a plan the skipped-tool guard has nothing to run, so these turns
get one nudge instead: an answer given before any tool ran is dropped and
the model is asked to call the tool (or ask for a missing detail).
Evals (qwen3:8b, EVAL_COMPACT_PROMPT=1 EVAL_INTENT_JUDGE=1, 3 runs of
tool_skip_confabulation, weather_autoderive_location, screen_view_tool,
followup_supplies_missing_tool_arg, tool_router_implicit): 71/84 before,
72/84 after, same known failures; planner calls 14 -> 3.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- botBrowser `play`: searches YouTube for the words as said, opens the first video (not an ad,
without the playlist) and starts it, in one call (an address given opens that page); `type`
also takes `query`. Live 0.1.19: the model stopped on YouTube's home page or the results, a
retry was refused as a third call, and the answer still said the video played.
- `brain.location` (기본 위치, default 대한민국) goes to the brain as `location_name`: the time line
says it without GeoIP (no database on the game PC), and getWeather without a place uses it.
- Planner and webSearch keep search words in the user's language (never translated into
English), so recommendations fit where the user is.
Evals (qwen3:8b, 3 runs): play ending on a playing video found with the words as said 0/6 -> 5/6;
Korean searches for Korean requests 6/9 -> 8/9.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Playback sends how long the clip playing when a request was stopped had played (tts_played
last_ms); the worker cuts the last sentence of that recording to it.
- botBrowser open takes the search words as plain `query`: the tool puts them, encoded, into the
address's empty search parameter or types them into the site's search box. The description
tells the model to search this way and never put words into an address itself (live: 아이유
hand-encoded into 아이우). Eval (qwen3:8b, 3 runs): right words reach YouTube 1/3 -> 3/3.
- When a page's main video does not play (also after the tool pressed play, or it failed to load),
a note after the page says it is not playing and not to say it plays. Eval: a paused video
reported as playing 1/3 -> 0/3.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Playback tells the worker how much of each request was played (tts_played); the worker keeps a
spoken answer's recording only that far (an answer stopped mid-way no longer keeps its unheard
rest, and is marked stopped); nothing played or never reported: no file.
- Old recording day folders are also removed with the first recording of each new day.
- response.interrupt off, latest: a request whose final comes after a newer one's is replaced
instead of replacing it.
- botBrowser presses play on a page's paused main player also after open, not only after a click.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Alone with Jarvis, whatever the speaker says is answered without asking the judge. With others
present the judge is told how many people are in the channel and how many may call Jarvis: with
one other person most of what the speaker says counts as said to Jarvis (unclear is yes); with
more, or not known, the question is the group one it was before. Live 2026-11-16: alone in the
channel, questions without the wake word got no answer because the judge assumed a group.
Eval (qwen3:8b, pair cases): 3/7 before, 6/7 after; group cases unchanged.
Also: the protocol test fixture's brain takes the trace argument (the test failed since traces).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- 로그 screen: "보이는 글 txt로 저장" (logs_save_text), a 대화 view that saves the days'
conversation with every trace in full (history_export), and a 목소리 view with the
voice.record switch and the days' recordings as a zip with an index (voice_export).
- Voice worker: with JARVIS_RECORDINGS_DIR (voice.record, off by default) each final and each
spoken answer is kept as a WAV with an index line; day folders older than 7 days removed.
- response.interrupt (default on): off, latest/important finish the running answer first.
- Bot browser: Chrome lets pages start video and sound; botBrowser reports each page's media
state, waits for a player to start after open/click, and presses play on a video page a click
led to when its player stayed paused (live: a clicked YouTube video stayed paused).
- webSearch reads `query` when `search_query` is missing (live: both searches of an answer failed).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A trace is now kept the moment the request arrives (at the time it was
asked, ms -1), its steps are stored as the engine adds each one, and it
ends with the answer. Traces a killed run left unfinished are closed at
start. The 대화 screen reads a trace still being made as pending.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The management app reads the worker's history store (read only) and shows
people's words, Jarvis's answers and the answer traces in time order, with
filters for kind, place, person, channel, day and text, live follow, older
items on demand and expandable traces (records given, tools offered, plan,
model thinking, tool calls and results).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stretching all of the speed-up from 1.0 kept every sound but sounded
awkward (UTMOS 3.13 at 1.5, against 3.73 for MeloTTS's own 1.5), while
MeloTTS's own speed past 1.25 drops sounds (sentence-final "요" and first
syllables, 4.6% character errors at 1.5). MeloTTS now speaks up to 1.25
and the WSOLA stretch does only the rest, with a 5 ms search: 2.6% and
UTMOS 3.56 at 1.5 through the real child. Version 0.1.6.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
MeloTTS's own speed shortens every sound and slurred fast speech (Whisper
large-v3 on five Korean sentences, three renders each: 5.5% character
errors at 1.5 and 20% at 2.0, against 1.4% at 1.0). The MeloTTS child now
speaks at most at its natural pace and makes faster speech with a WSOLA
time stretch that keeps the pitch (1.9% at 1.5, 2.2% at 2.0, about 15 ms
per 4 s of speech).
A voice.tts.speed saved while the services run goes to the worker at once
(new tts_speed message) and applies from the next sentence; the client
sends it again whenever the worker connects again. Version 0.1.4.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- screenView tool (screen_view.spec.md): whole screen, monitor by number, window by part of its
title or program, or the list of monitors and windows; fast, normal or detailed pictures.
- The app captures with Windows Graphics Capture (app/src/screen.rs, admin.spec.md "Screen
viewing"), decides for the asker (only people registered for PC control), never picks, lists
or shows excluded programs (pcControl.screenExclude, password managers by default), refuses
minimised windows and black pictures, and records screen.view in the activity log.
- Pictures ride on the tool message (ToolExecutionResult.images); llm.py sends them to
OpenAI-compatible servers as image parts after the tool messages. They stay in that reply only:
not in the tool carryover, logs or prompt dumps.
- The LLM reads pictures only when screen viewing is available (--vision gpu normally, cpu in a
game), measured apart for the placement; otherwise screenView is a disabled tool.
- Worker: screen_action / screen_action_result, 4 MiB messages.
- Eval evals/test_screen_view_tool.py; Windows check tests/screen_windows.rs.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
On the shared screen a tab brought to the front makes Chrome bring its window forward and take the
keyboard focus from the game or program the user is in (seen on the game PC). The botBrowser tool
now opens each person's page in a window of its own, behind and at the place and size of the bot
browser's window (Target.createTarget with newWindow and background), and closes it again if the
connection to it fails. While the browser runs on the virtual monitor, the app looks for new Chrome
windows every 200 ms and settles each like the first (tool window, fitted to the work area, without
activating it), the newest in front of the older ones, so the shared screen shows the page the bot
opened last.
Checked on the game PC: the bot's window loading a public page comes in front of the first, both
out of the taskbar and Alt+Tab, the focus staying in Discord also with the user idle for hours.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each collection target can announce new items at once or in a daily digest
(off by default) in collect.notifyChannel, as titles and links only, with no
LLM; each post is recorded in the activity log. The brain gets a read-only
collectedNews tool (every target's newest items by default, fenced as
untrusted web content), offered with the app's collection file and the
collection feature, and available to scheduled tasks.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>