-
bb60c4db3b
feat(tts): default to offline GPU MeloTTS with brain-role worker
codex/owner/javis_bot
javis-bot
2026-07-22 20:39:54 +09:00
-
8562548571
feat(deploy): add brain/app roles for GPU-split deployment
javis-bot
2026-07-22 15:56:51 +09:00
-
db51bb92e6
feat(compose): allow external Ollama via OLLAMA_BASE_URL override
javis-bot
2026-07-22 13:42:58 +09:00
-
ffc16665e5
fix(controlBrowser): never report moveMouse/search success without a real move
main
v1.10.0
javis-bot
2026-06-24 19:17:46 +09:00
-
5629da7e9f
feat(controlBrowser): add moveMouse (hover) action for the visible cursor
javis-bot
2026-06-24 19:14:04 +09:00
-
83999a5b0b
fix(prompts): classify every sub-8B model (2b/4b/5b/6b) as SMALL
v1.9.1
javis-bot
2026-06-24 18:31:54 +09:00
-
d5fd218c86
fix(reply): stop weak models parroting persona example facts
javis-bot
2026-06-24 18:31:54 +09:00
-
23b1fe692b
docs: warn against setting OLLAMA_INTENT_MODEL larger than the chat model
v1.9.0
javis-bot
2026-06-24 17:57:30 +09:00
-
7bb9718c34
feat(reply): cap spoken replies at a single sentence
javis-bot
2026-06-24 17:55:27 +09:00
-
7da2fcb5e5
feat(stt): beam-search decoding + no prev-text conditioning for accuracy
javis-bot
2026-06-24 17:55:20 +09:00
-
680f5a656a
docs: reflect the separate auxiliary intent/router model in llm_contexts + README
javis-bot
2026-06-23 17:38:58 +09:00
-
b52ffd2b18
perf: run auxiliary LLM calls on a small model, big model only for the answer
v1.8.1
javis-bot
2026-06-23 17:35:40 +09:00
-
140fc56f18
feat: play the Nth YouTube result in browseAndPlay via an index arg
v1.8.0
javis-bot
2026-06-23 15:33:45 +09:00
-
5ee47827f3
perf: cap chat output tokens via ollama_num_predict to bound reply latency
javis-bot
2026-06-23 15:33:45 +09:00
-
c189ce2e65
feat: Korean Chrome locale, agents/llm.md voice instructions, drop emoji rule
v1.7.0
javis-bot
2026-06-23 04:34:17 +09:00
-
086dd5cde7
fix: accept edge as a valid tts_engine and migrate stale persisted engines
v1.6.1
javis-bot
2026-06-23 03:49:53 +09:00
-
f64d76e737
feat: use Edge TTS (Korean Hyunsu voice @ +45%) as the default voice
v1.6.0
javis-bot
2026-06-23 03:44:15 +09:00
-
11c3621093
fix: make container TTS engine env-driven so melo isn't overridden by piper
v1.5.2
javis-bot
2026-06-23 03:27:33 +09:00
-
7ad5d99380
Revert "feat: replace MeloTTS with Coqui XTTS-v2 natural Korean voice"
v1.5.1
javis-bot
2026-06-23 03:15:54 +09:00
-
39a0944105
feat: replace MeloTTS with Coqui XTTS-v2 natural Korean voice
v1.5.0
javis-bot
2026-06-23 03:08:01 +09:00
-
b9f637faa4
fix: stop hardcoding MELO_SPEED so the .env override reaches the worker
v1.4.1
javis-bot
2026-06-23 01:02:43 +09:00
-
2f000ac6c8
feat: load operator instructions from agents/*.md into the reply prompt
v1.4.0
javis-bot
2026-06-23 00:57:54 +09:00
-
677bfcd2a9
feat: log the resolved whisper device on bridge load
v1.3.0
javis-bot
2026-06-23 00:19:20 +09:00
-
e49be6d04e
fix: add
video driver capability so NVENC works in the container
v1.2.3
javis-bot
2026-06-23 00:16:12 +09:00
-
1efabe03b1
fix: strip CR from container shell scripts in Dockerfile build
v1.2.2
javis-bot
2026-06-22 23:35:12 +09:00
-
09cd4c5e31
fix: pin docker shell scripts to LF to stop CRLF breaking the image build
v1.2.1
javis-bot
2026-06-22 23:27:46 +09:00
-
00ce813845
docs: warn that COMPOSE_FILE uses ';' on Windows, ':' on Linux/macOS
javis-bot
2026-06-22 23:13:33 +09:00
-
c56ce1eb30
feat: human-like typing for browser Google and YouTube search
v1.2.0
javis-bot
2026-06-22 21:09:15 +09:00
-
597207dd33
feat: reuse a signed-in Chrome profile for browser web search
javis-bot
2026-06-22 20:57:25 +09:00
-
98a1825d01
feat: headless Chrome fallback for browser web search outside broadcast
javis-bot
2026-06-22 20:52:54 +09:00
-
da27c5a306
docs: warn that personal Google login is blocked on the Gemini CLI path
javis-bot
2026-06-22 19:44:12 +09:00
-
5b6a67963a
feat: make GEMINI_AUTH=oauth authenticate in Docker
v1.1.0
javis-bot
2026-06-22 18:05:22 +09:00
-
53be1567b1
docs: README — OS-specific install/run (Linux CDI vs Windows Docker Desktop)
javis-bot
2026-06-18 04:37:40 +09:00
-
ccddbd6448
test: settings output_language survives save→apply→recreate
v1.0.0
javis-bot
2026-06-16 19:56:13 +09:00
-
7870a76314
fix: persona uses settings output_language, matching the reply directive
javis-bot
2026-06-16 19:47:26 +09:00
-
b3088dd75f
fix: settings output_language overrides the compose env default
javis-bot
2026-06-15 16:57:01 +09:00
-
8868381f13
fix: minimal Chrome flags (drop --test-type/AutomationControlled) + policy infobar suppress
javis-bot
2026-06-15 16:51:11 +09:00
-
247edda3eb
fix: google anti-bot flag + persistent/safe settings apply + TTS engine wiring
javis-bot
2026-06-15 13:13:11 +09:00
-
84e435f916
feat: settings web UI (models / STT / TTS speed / language / LLM instructions)
javis-bot
2026-06-15 13:05:46 +09:00
-
3bdc7d078a
feat: cross-platform compose (Ubuntu CDI + Windows Docker Desktop GPU)
javis-bot
2026-06-15 13:00:04 +09:00
-
8dd6386af8
fix: search like a person — open homepage, type in the site's search box
javis-bot
2026-06-15 12:42:58 +09:00
-
aebf183950
feat: browser-control server on host (real input) + remote-bot routing + ignore env backups
javis-bot
2026-06-15 10:41:57 +09:00
-
1935c1a6bc
feat: split-deployment roles (browser-host on LAN + remote bot)
javis-bot
2026-06-15 10:23:55 +09:00
-
bdb012fc7c
fix: weather passes named city from utterance; clean navigate reply
javis-bot
2026-06-15 01:02:35 +09:00
-
49061d30f0
fix: deterministic browser navigate for 'go back to <site>' (구글로 돌아가)
javis-bot
2026-06-15 00:58:53 +09:00
-
c522e1b285
fix: deterministic weather → one clean Korean sentence (no 'Celsius')
javis-bot
2026-06-14 22:46:03 +09:00
-
54c3ce7d1b
feat: show speaker nickname instead of raw user ID in voice logs
javis-bot
2026-06-14 22:39:06 +09:00
-
d970bf276e
fix: harden Korean-only output lock (front+end, explicit script ban)
javis-bot
2026-06-14 22:25:33 +09:00
-
3d620dc4c7
fix: concise Korean weather reply (current conditions, one sentence)
javis-bot
2026-06-14 21:49:53 +09:00
-
09afc21283
feat: split recording vs STT-processing time in turn logs
javis-bot
2026-06-14 21:45:50 +09:00
-
11a72cb296
fix: deterministic on-screen site search + lock replies to Korean
javis-bot
2026-06-14 21:07:26 +09:00
-
37759f2b2c
perf: conversational fast-path (skip enrichment) + shorter silence wait
javis-bot
2026-06-14 03:21:52 +09:00
-
4bc7f836ae
fix: native tool-calling for qwen2.5 (tools actually fire) + kill Chrome infobars
javis-bot
2026-06-14 03:08:49 +09:00
-
642fa42561
fix: drop webSearch when a site is named in screen-share, forcing controlBrowser
javis-bot
2026-06-14 02:57:51 +09:00
-
f3d34bab4d
fix: always offer controlBrowser in screen-share so on-screen search works
javis-bot
2026-06-14 02:55:33 +09:00
-
c21c5b225f
feat: controlBrowser 'search' action + one-sentence voice replies
javis-bot
2026-06-14 02:52:47 +09:00
-
109dbc7e16
feat: robust controlBrowser — real xdotool input, tabs/back/forward/popups
javis-bot
2026-06-14 02:33:59 +09:00
-
b18217fcdd
fix: let melo-worker honour MELO_DEVICE from env (was hardcoded cpu)
javis-bot
2026-06-14 02:26:24 +09:00
-
3d1e56f60f
feat: controlBrowser tool — human-operable on-screen Chrome
javis-bot
2026-06-14 02:22:36 +09:00
-
927d59f805
perf: run MeloTTS on the GPU (cu128 torch) + warm CUDA at startup
javis-bot
2026-06-14 02:22:36 +09:00
-
44ebfeafa8
feat: per-call LLM timing, speaker ID, cancel captures on leave
javis-bot
2026-06-14 00:38:26 +09:00
-
5c11c5f7e8
perf: pre-warm ollama at the engine's num_ctx (8192) so first turn is hot
javis-bot
2026-06-14 00:25:34 +09:00
-
2c38e7576d
perf: unify Ollama num_ctx so a voice turn keeps one resident model
javis-bot
2026-06-14 00:19:53 +09:00
-
d4e5e7f3f7
perf: pre-warm Whisper + chat model + TTS at bridge startup
javis-bot
2026-06-14 00:04:03 +09:00
-
de5384d166
feat: show per-stage timing (듣기/LLM/TTS) in the transcript channel
javis-bot
2026-06-13 23:58:49 +09:00
-
ddebdd7542
feat(stream): single-account Go-Live — broadcast on the conversation's session
javis-bot
2026-06-13 22:37:26 +09:00
-
e8234b7fb1
feat(stt-log): log the WHOLE turn pipeline to the transcript channel
javis-bot
2026-06-13 22:13:39 +09:00
-
62bb0ab87e
feat(bot): mirror heard voice turns to a text channel
javis-bot
2026-06-13 22:06:38 +09:00
-
4e446c1d7c
perf(stt): default Whisper model small -> medium for better Korean accuracy
javis-bot
2026-06-13 22:03:18 +09:00
-
3a4776c709
perf(brain): env to disable pre-loop planner; cut voice latency
javis-bot
2026-06-13 21:59:35 +09:00
-
f12e6b28c2
perf(bridge): lock STT to Korean + add per-stage turn timing
javis-bot
2026-06-13 21:53:47 +09:00
-
f2b43cb310
fix(bridge): stop dropping real speech — disable avg_logprob confidence floor
javis-bot
2026-06-13 21:42:36 +09:00
-
2e21185dc0
fix(bot): never broadcast on the conversation's account (it deafened STT)
javis-bot
2026-06-13 21:33:50 +09:00
-
863337c6eb
feat(selfbot): dedicated DISCORD_STREAM_TOKEN for the broadcast account
javis-bot
2026-06-13 15:30:33 +09:00
-
5961faed38
fix(selfbot): gate broadcast 'live' on real Go-Live WebRTC connect
javis-bot
2026-06-13 15:22:52 +09:00
-
6f4464b2c4
chore(bot): log an explicit line when the broadcast auto-starts on join
javis-bot
2026-06-13 15:00:07 +09:00
-
6d72e10f9c
feat(bridge): gate Whisper behind Silero VAD; harden broadcast auto-start
javis-bot
2026-06-13 14:53:54 +09:00
-
39c7a22a12
fix(bridge): gate STT on real speech so noise doesn't trigger replies
javis-bot
2026-06-13 14:45:22 +09:00
-
568a1ae50b
fix(bot): auto-start broadcast on voice join in userbot mode
javis-bot
2026-06-13 14:39:29 +09:00
-
d6c029d7d5
fix(bridge): keep decimals, versions and URLs whole in TTS sentence split
javis-bot
2026-06-13 01:28:37 +09:00
-
5c295420ea
perf(bridge): stream TTS per sentence to cut voice reply latency
javis-bot
2026-06-13 01:09:39 +09:00
-
989a4f3e98
perf(memory): keep embed model warm across turns (keep_alive 0 -> 5m)
javis-bot
2026-06-12 23:45:16 +09:00
-
67d0ae711c
docs(readme): make Discord token setup userbot-first
javis-bot
2026-06-12 21:59:58 +09:00
-
932aacef6e
feat(bot): delay login until MeloTTS voice is warm
javis-bot
2026-06-12 21:59:50 +09:00
-
ccbaed9030
feat(weather): romanise non-Latin place names before geocoding
javis-bot
2026-06-12 21:59:42 +09:00
-
f3a1d92620
fix(brain): route auxiliary small-model calls to an available model
javis-bot
2026-06-12 21:59:35 +09:00
-
c8a04a110f
fix(brain): recover colon-JSON and single tool_call object forms
javis-bot
2026-06-12 21:59:26 +09:00
-
f89246a14d
docs(docker): clarify userbot mode in compose/run-bot, bot token optional
javis-bot
2026-06-12 21:26:00 +09:00
-
8a2a109d5e
feat(brain): make OUTPUT_LANGUAGE lock robust on small models
javis-bot
2026-06-12 21:18:47 +09:00
-
006a32276a
feat(brain): add OUTPUT_LANGUAGE reply-language lock
javis-bot
2026-06-12 21:08:44 +09:00
-
40877b65b3
docs(env): mark DISCORD_BOT_TOKEN optional — blank runs userbot mode
javis-bot
2026-06-12 21:01:08 +09:00
-
b91c05a355
perf(brain): pin chat model per-request, unload embeddings; default qwen2.5:3b
javis-bot
2026-06-12 20:36:19 +09:00
-
7792be254a
perf(brain): keep ollama models resident to kill voice cold-start latency
javis-bot
2026-06-12 19:12:23 +09:00
-
b17961e9e3
feat(tts): add MeloTTS Korean voice via warm worker with offline-baked cache
javis-bot
2026-06-12 19:01:54 +09:00
-
3e333763fb
fix(stream): enable captions only for real Korean tracks, skip auto-generated
javis-bot
2026-06-11 18:41:25 +09:00
-
835459111e
fix(bot): userbot voice via @discordjs/voice 0.19 + DAVE (selfbot voice handshake fix)
javis-bot
2026-06-11 18:19:10 +09:00
-
921a757371
feat(bot): userbot (selfbot) mode — voice conversation on one user account (stage 1)
javis-bot
2026-06-11 11:08:39 +09:00
-
03375a77d0
fix(docker+brain): make container Chrome search work, harden, improve toggle
javis-bot
2026-06-11 10:44:56 +09:00
-
ca86390407
feat: couple broadcast to voice + voice-controlled broadcast toggle
javis-bot
2026-06-11 10:08:30 +09:00
-
35e754d6ee
fix(docker): in-container Gemini CLI OAuth, broadcast audio, ports + hardening
javis-bot
2026-06-11 01:36:30 +09:00