qwen2.5:7b leaked Chinese/Cyrillic mid-reply despite the OUTPUT_LANGUAGE
lock, which was buried mid-prompt. Repeat the lock at the END of the system
prompt (recency) and ban specific foreign scripts explicitly.
getWeather returned a verbose multi-section English forecast that the 3B
re-synthesised into long, CJK/°F-leaking answers. Hand it a ready-to-speak
Korean one-liner (지금 <곳> 날씨는 <상태>, 기온 N도(체감 M도)입니다) and drop the
hourly/7-day firehose from the default voice reply.
- Site-specified search ("네이버에서 X 검색해줘") now runs controlBrowser.search
directly in the engine when broadcasting, instead of relying on the 3B model
to emit the tool call (it kept narrating "검색하겠습니다" without acting).
- Set OUTPUT_LANGUAGE=ko so replies are Korean-only — stops the small model
leaking CJK/Hanja and English fragments (每, 朗, "feels like") into weather
and other answers, and keeps them concise.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Greetings/small-talk routed no data tool yet still ran the episodic memory
enrichment (LLM keyword extract + diary/graph search, ~1s) every turn. Skip it
when the router picked no external-data tool — the always-injected warm profile
still personalises the reply. Also drop the voice silence-detection wait
800ms -> 600ms for snappier turn-taking. Warm "안녕" now lands well under the
3-4s target.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Root cause of "weather/search do nothing": the engine forced TEXT tool-calling
for all <=7B models, but qwen2.5:3b emits clean NATIVE tool calls and fails at
the text format — so it just narrated ("부산 날씨는 맑습니다") and never called
getWeather/webSearch/controlBrowser. Use native tool-calling for tool-capable
small families (qwen2.5/qwen3/llama3.x/mistral); native still auto-falls back
to text on HTTP 400, so non-tool models (gemma) are unaffected.
Also launch Chrome with --test-type (removes the "--no-sandbox unsupported
flag" infobar) and disable the Translate feature/popup.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The 3B model kept choosing webSearch over controlBrowser even when offered, so
'네이버에서 X 검색' still used the invisible web path. When broadcasting and the
user explicitly names a site, remove webSearch from the allow-list so the
on-screen browser is the only search route.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The small router reflexively routed every "search/open" intent to webSearch
and never surfaced controlBrowser, so "네이버에서 X 검색해줘" did nothing on the
broadcast. Union controlBrowser (+browseAndPlay) into the allow-list every
turn in screen-share mode (like setBroadcast), and steer the model in the
system prompt to prefer the on-screen browser over webSearch when available.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- Add a one-shot `search` action (site=naver/google/daum/youtube/bing) that
navigates the on-screen browser straight to the results page, so a small
model can satisfy "search X on Naver" in a single tool call instead of a
fragile navigate->type->enter chain.
- Sharpen the tool description to steer the router to controlBrowser (not
webSearch) for anything that should happen IN the visible browser.
- System prompt: answer in one short sentence (voice assistant) — also cuts
TTS time.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- Install xdotool + wmctrl (late Docker layer, preserves the melo cache) so
the on-screen Chrome gets real X input (visible cursor, char-by-char typing)
instead of synthetic events; falls back to the Playwright API if absent.
- Fix active-tab detection (probe document.visibilityState instead of assuming
tab 0) so sequential ops target the right tab.
- Add back / forward / refresh; new/switch/close tabs via real keyboard
(Ctrl+T / Ctrl+<n> / Ctrl+W) when xdotool is present.
- Auto-dismiss native JS dialogs; closePopups clears blank/popup tabs.
- Report broadcast (Go-Live) state in status from the turn context.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds a general browser-control tool (navigate to any site, list/open/close/
switch tabs, close popups, click, type, scroll, screenshot) for the Go-Live
Chrome, on top of the existing CDP + xdotool human-input layer (visible
cursor, char-by-char typing). Closes the gap where "open Naver" had no tool
and the model confabulated success. Also adds a system-prompt rule against
claiming actions no tool actually performed.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- llm.py: log each Ollama call's caller + total/load/prompt/gen durations
so a slow voice turn is attributable to a specific internal call
(router/enrichment/digest/main); a RELOAD marker flags cold reloads.
- voice.ts: track in-flight Opus captures and abort them on session
destroy(); drop any utterance that finishes after the user left, so no
trailing post-leave VAD turns are reported.
- userbot.ts: show the speaker's Discord user ID on each transcript line
(answered and dropped) so it's clear whose audio produced the turn.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Ollama keeps a separate loaded model instance per (model, num_ctx). The
main agentic chat used num_ctx=8192 while the router/enrichment/digest
passes used 4096, so every voice turn forced at least one cold reload
(~3.4s) when switching context sizes — the dominant per-turn latency
(measured: resident chat call 0.27s vs cold 3.4s).
Introduce a single OLLAMA_NUM_CTX (default 8192, env-tunable for tight
VRAM) used by call_llm_direct, chat_with_messages, call_llm_streaming and
the planner, collapsing a turn to one resident instance.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Warm per-turn timing showed STT 0.1s, TTS ~1-3s, but the reply engine (LLM) was
8-17s — even a simple "고마워" took 16.7s — because it makes multiple model calls
per turn. Add a PLANNER_ENABLED env override (config.py) and default it to 0 in
the userbot compose so the pre-loop planner's extra LLM round-trip is dropped on
this latency-sensitive voice deployment. Also pins STT_LANGUAGE=ko in compose.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Empirical A/B/C measurement against the live RTX 5050 Ollama stack
(qwen2.5:3b + nomic-embed-text) showed keep_alive=0 unloads the embed
model ~2s after every call, so each turn after a brief idle gap pays a
cold reload. VRAM is not the constraint (~4.4-4.7 GB free with both
models resident) and keep_alive=0 never evicted the chat model, so CPU
embedding (num_gpu=0) gave no benefit. A short positive keep_alive is
the fastest of the three: it keeps the ~0.3 GB embed model resident
across consecutive turns at negligible VRAM cost.
Add tests/test_embeddings.py covering the warm-across-turns behaviour.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Open-Meteo's geocoder only matches Latin/English spellings, so a Korean city
name like "서울" returns zero results even though the place exists. With
OUTPUT_LANGUAGE locked to Korean the tool-calling model naturally fills
`location` with the Korean name, which dead-ended on every weather request
("could not find location").
When the first geocode is empty and the name is non-ASCII, ask the warm small
model for the common English exonym ("서울" -> "Seoul") and retry once. ASCII
names skip the round-trip entirely.
Live-verified: "서울 날씨 알려줘" now returns real Seoul weather. Tests cover the
romanise-and-retry path and the ASCII short-circuit.
qwen2.5:3b emits tool calls in text shapes the parser dropped, breaking
two reviewer-reported behaviours:
- `getWeather: {"location": "Seoul"}` (a JSON object after the colon) was
dumped wholesale into {"query": "{...}"}, so `location` never reached the
tool. getWeather then ran with empty args, returned the auto-detected
location's weather, the model noticed the mismatch and retried — looping up
to 8 times before giving up with an English error. Now the JSON object after
the colon is parsed directly as the argument dict.
- `call_stop: {"id":..., "function": {"name": "setBroadcast",
"arguments": "{\"action\": \"stop\"}"}}` — a single tool_call object without
the `tool_calls: [...]` array wrapper, behind a `call_xxx:` label — matched
no form, so the raw JSON leaked to the user AND setBroadcast never ran
("방송 꺼줘" did nothing). Now name + arguments are pulled from the embedded
`function` object when the name is in the allow-list.
Field-captured from the live qwen2.5:3b brain (2026-06-12). Tests cover both
shapes, non-ASCII args, dict/string arguments, and unknown-tool rejection.
Harden the reply-language lock so qwen2.5:3b reliably stays in the locked
language instead of leaking the query language back in:
- reply_language_directive(): single resolver with clear precedence —
explicit OUTPUT_LANGUAGE lock wins over the Piper/Chatterbox English-only
fallback (this deployment's actual TTS is Korean MeloTTS, so the legacy
English lock was both wrong and contradicting the Korean lock).
- Stronger, override-explicit directive wording, inserted near the FRONT of
the system prompt so a small model gives it primacy over the persona.
- build_system_prompt(output_language=...): rewrite the persona's "in the
user's language" clause to the locked language so the persona stops
fighting the lock.
- docs/llm_contexts.md: document the resolver, precedence, and placement.
Live-verified on the running brain (qwen2.5:3b): Korean voice-style input
and a cold English query both return fully Korean replies with no CJK/Hanja
leak. Tests cover unset/set/agnostic/whitespace + precedence + persona rewrite.
Add an optional OUTPUT_LANGUAGE env var that forces every reply into a
single language. When set, output_language_directive() injects a "respond
only in <language>" instruction (also forbidding other scripts) into the
chat loop's system prompt, next to the existing TTS English-only lock.
Empty (default) keeps the multilingual "reply in the user's language"
behaviour, so upstream is unaffected.
For the Korean-only deployment this also suppresses the occasional trailing
CJK/Hanja fragment qwen2.5:3b leaks on free-form chit-chat.
- system_prompt.py: language-agnostic output_language_directive() helper
- engine.py: read OUTPUT_LANGUAGE, append directive in _build_initial_system_message
- docker-compose.yml + .env.example: document/pass the new var
- docs/llm_contexts.md: note the new gating on the main reply context
- tests: cover unset/set/agnostic/whitespace cases
Replace the blunt global OLLAMA_KEEP_ALIVE=-1 (which kept every model,
including nomic-embed, resident in VRAM forever) with per-request residency:
- llm.py: all three /api/chat payloads send keep_alive=30m so the actively
used chat model stays resident and voice turns never pay a cold reload.
- embeddings.py: /api/embeddings sends keep_alive=0 so nomic-embed unloads
right after each call instead of squatting in VRAM next to the chat model.
- docker-compose.yml: drop the global OLLAMA_KEEP_ALIVE=-1; document the
per-request scheme on the ollama service.
Switch the default chat model qwen3:8b -> qwen2.5:3b. Verified live on the
RTX 5050 (8GB):
- ollama ps: qwen2.5:3b 2.4GB, 100% GPU (8B was 92% GPU / 8% CPU), UNTIL ~30m
(the 30m pin, not "Forever"); nomic-embed absent after several enriched turns.
- nvidia-smi: ~3.2GB VRAM used total (qwen 2.4GB + whisper 0.7GB) vs ~6.6GB.
- Korean /text turns: warm 1.7-4s (cold first load ~52s), vs ~5-7s on 8B;
time/weather/places tool calls fire and reply in Korean.
Known limitation: qwen2.5:3b can occasionally leak a trailing CJK phrase on
free-form chit-chat (factual/tool replies stay clean).
Addresses review findings on the dockerized stack:
- Container Chrome search was dead: add --remote-debugging-port + a non-default
--user-data-dir (Chrome 136+ refuses CDP on the default profile), add the
playwright dep (browse-search.mjs connectOverCDP) with browser download
skipped, and connect to 127.0.0.1 not "localhost" (container localhost -> ::1
while Chrome binds IPv4). Verified: browse-search returns real results.
- Broadcast toggle reliability: always offer setBroadcast in screen-share mode
(the embedding/keyword router dropped it for non-English utterances) and make
its description force a tool call. "방송 꺼줘"->stop now 5/5; no false triggers.
- Stop the broadcast on voice leave (no orphaned stream).
- Security: bind VNC/noVNC to loopback by default (VNC_BIND override) and the
bridge to the container loopback (BRIDGE_HOST=127.0.0.1), not published.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Completes the STREAM_BROWSER=true behaviour:
- handleJoin auto-starts the broadcast on voice join and wires the session to
the guild streamer; each turn reports the live state to the brain so search
routes Chrome (live) vs Gemini (off).
- New setBroadcast tool lets the user toggle the broadcast by voice ("방송
켜줘/꺼줘") via the LLM (no hardcoded phrases); it refuses when
STREAM_BROWSER=false. The directive flows brain -> bridge (broadcast_action)
-> bot streamer.start/stop, guarded by isActive() so it's idempotent.
- Per-turn IPC uses a thread-local (reply/turn_state.py) instead of threading
params through the whole engine chain: bridge sets broadcasting in, tool
records the directive out; Tool.execute exposes broadcasting on ToolContext.
Bot typecheck clean; brain covered by tests/test_set_broadcast.py (+ existing
routing tests). Specs + docs updated.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
STREAM_BROWSER becomes the broadcast *capability* master flag; the live
screen-share state (new ToolContext.broadcasting, passed per turn by the bot)
decides the backend:
- master off -> broadcast disabled, always Gemini
- master on + live on -> on-screen Chrome (visible on the stream)
- master on + live off -> Gemini
context.broadcasting is None outside the voice path (evals, text entry, older
bot) and falls back to the master flag, so current behaviour is unchanged.
This is the brain-side foundation; bot-side wiring (bridge passes broadcast
state, auto-broadcast on voice join, voice on/off toggle) follows.
Specs + docs/llm_contexts.md updated. Covered by
tests/test_web_search_broadcast_routing.py.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
yolo auto-approves every tool call, so a real-time search query could in
principle trigger write/shell tools. Default approval mode still auto-runs the
CLI's read-only web search in headless mode but never silently approves
destructive tools. Verified end-to-end: a grounded query returns a current
answer in ~23s with the account OAuth login. Test asserts yolo is absent;
specs and docs/llm_contexts.md updated.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds a GEMINI_AUTH=oauth (default) sub-mode that shells out to the Gemini CLI
using the user's Google-account login instead of an API key. gemini_cli_search()
runs `gemini -p <query> -o json --skip-trust --approval-mode yolo`, strips
GEMINI_API_KEY/GOOGLE_API_KEY and sets GOOGLE_GENAI_USE_GCA=true so the CLI
selects the account OAuth method and fails fast when no login exists. Bounded by
a 30s timeout and fail-open to the DDG/Brave/Wikipedia cascade on any failure
(CLI missing, login expired, quota 429, timeout). GEMINI_AUTH=apikey keeps the
legacy REST path. Specs and docs/llm_contexts.md updated; behaviour covered by
tests/test_realtime_gemini_cli.py.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Completes the two info modes in the Python brain:
- config.py: read STREAM_BROWSER / GEMINI_API_KEY / GEMINI_MODEL from env into
Settings (stream_browser, gemini_api_key, gemini_model). Verified load_settings
reads both modes.
- realtime_search.py: two fail-open backends returning the same fenced
UNTRUSTED-WEB-EXTRACT envelope: browser_search() shells the Node CDP helper to
drive the on-screen Chrome (visible on the broadcast); gemini_search() calls
the Gemini REST API with google_search grounding.
- web_search.run(): routes by mode before the DDG cascade (true->browser,
false->Gemini), falling through to DDG/Brave/Wikipedia on any miss.
- browse_and_play tool: plays a YouTube video on the shared screen (true mode
only); registered in the tool registry.
- specs + docs/llm_contexts.md updated (new Gemini LLM context); CLAUDE.md spec
registry updated.
Verified live against the running Chrome: true-mode webSearch returned real
Google results for "오늘 서울 날씨", browseAndPlay played the IU 밤편지 MV, and
false-mode degrades gracefully on a bad/absent key. A valid GEMINI_API_KEY is
still needed to confirm the real Gemini grounding output.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
