Single biggest user-facing change: you can now switch local↔cloud LLM by voice
without restarting. Plus four new packs and a perf nudge.
Shared LLM global (crates/jarvis-core/src/llm/mod.rs)
- GLOBAL: Lazy<RwLock<Option<Arc<LlmClient>>>>. Single source of truth.
- init_global() / current() / swap_to(LlmBackend) / current_backend_name() / parse_backend(str).
- llm_fallback + llm_router migrated: they no longer own a separate LlmClient,
they read from the global on each request. Hot-swap is now LIVE — change
backend, next chat call uses the new one.
- parse_backend accepts both english and russian aliases:
"groq"/"cloud"/"облако"/"клауд" → Groq
"ollama"/"local"/"локал"/"локальный" → Ollama
Voice + Lua LLM switcher (resources/commands/llm_switch/, 3 commands)
- "переключись на локальный" / "перейди на оллама" → swap to Ollama
- "переключись на облако" / "используй грок" → swap to Groq
- "какой у тебя мозг" / "облако или локально" → status query
Underlying Lua API (registered globally):
jarvis.llm({messages}, opts) -- as before, now uses global client
jarvis.llm_status() -> "groq" | "ollama" | "none"
jarvis.llm_switch(name) -> name on success, nil on failure
Media keys pack (resources/commands/media_keys/, 4 commands)
- "пауза" / "плей" → VK_MEDIA_PLAY_PAUSE
- "следующий трек" → VK_MEDIA_NEXT_TRACK
- "предыдущий трек" → VK_MEDIA_PREV_TRACK
- "стоп музыка" → VK_MEDIA_STOP
- Helper _media.ps1 uses user32.keybd_event (P/Invoke through Add-Type).
- Works with anything that listens to global media keys: Spotify, Yandex Music,
YouTube (focused tab), Foobar2000, Winamp, etc. No OAuth, no API keys.
Codebase Q&A pack (resources/commands/codebase_qa/, 3 commands)
- "укажи проект <path>" → stores path in jarvis.memory under codebase.root
- "какой проект сейчас" → speaks current path
- "что делает функция X" / "найди в коде Y" / "объясни код"
→ walks the folder (depth 3, 30 files cap, 4KB per file, 50KB total),
filters by source extensions (rs/py/ts/lua/go/...), feeds digest to LLM
with a "senior reviewer" system prompt, speaks 3-5 sentence answer.
Scheduler cancel-by-text (crates/jarvis-core/src/scheduler.rs)
- new pub fn find_by_text(query, limit) -> Vec<ScheduledTask>
- new pub fn remove_by_text(query) -> usize (count removed)
- new pure helper find_by_text_in(tasks, query, limit) for tests
- Lua: jarvis.scheduler.{find_by_text, remove_by_text}
- Voice: "отмени напоминание про воду" / "удали задачу про разминку"
- 3 new unit tests (52 total in jarvis-core).
TTS pre-warm (crates/jarvis-app/src/main.rs)
- Call jarvis_core::tts::backend() once on startup so the first speak()
doesn't pay Piper binary discovery + voice loading cost. Cuts first-speak
latency by ~150-300ms depending on backend.
Build: cargo build --release -p jarvis-app -p jarvis-gui both green.
Tests: 52/52 jarvis-core unit tests pass.
User asked the code stays easy to fix/extend as the feature surface grows.
This commit refactors what was duplicated, centralises what was scattered,
adds tests where there were none, and writes the architecture doc that
contributors will need.
Deduplication
- tts::play_wav extracted to tts/mod.rs as pub(crate). Was identical in
piper.rs and silero.rs (~25 lines × 2). Both backends now call super::play_wav.
Centralised env config
- new crates/jarvis-core/src/runtime_config.rs — single doc-file for every
JARVIS_* / GROQ_* / OLLAMA_* env var. Includes:
- ENV_* constants with doc comments
- get(name) / get_bool(name, default) / get_parse(name, default) helpers
- feature-flag wrappers: llm_tts_enabled(), llm_router_enabled(),
llm_router_threshold()
- log_effective_config() prints active values on startup
- migrated llm_fallback (JARVIS_LLM_TTS), llm_router (JARVIS_LLM_ROUTER,
JARVIS_LLM_ROUTER_THRESHOLD) to use the new helpers. Pattern set for
future migrations.
Lua boilerplate killer
- new crates/jarvis-core/src/lua/api/cmd.rs exposing:
jarvis.cmd.ok(msg?) — play_ok + speak + {chain=false}
jarvis.cmd.chain_ok(msg?) — same but chain=true
jarvis.cmd.error(msg?) — play_error + speak + {chain=false}
jarvis.cmd.not_found(msg?) — play_not_found + speak + {chain=false}
jarvis.cmd.silent_ok / silent_error
- refactored 5 packs (daily_briefing/off, memory_pack/list, pomodoro/stop,
habit_nudge/stop_all, scheduler/clear) — each lost 3-4 lines of repetitive
play_*/speak/return boilerplate. Pattern for future packs documented in
ARCHITECTURE.md.
Tests (was 32, now 49)
- long_term_memory: 8 new tests for normalize_key, search_in (rank, limit,
empty), build_context_from (empty + populated), serde round-trips for
MemoryRecord and Store. Extracted pure logic (search_in, build_context_from)
into pub(crate) functions to enable testing without global state.
- profiles: 6 new tests for Profile::allows_command (empty/whitelist/blacklist/
deny-wins-over-allow), serde round-trip with all fields + minimal-fields
tolerance via #[serde(default)].
- runtime_config: 2 tests for get_bool / get_parse defaults.
- All 49 tests pass.
New mood/energy log pack
- resources/commands/mood_log/ with 2 commands:
mood.record "запиши настроение 7" / "сегодня мне грустно" → stores
timestamped entry via jarvis.memory.remember
mood.recap "как прошла неделя" → LLM summarises last 30 entries
- Showcase: composes memory + llm + cmd helpers in <30 lines per script.
ARCHITECTURE.md (new, 250 lines)
- Crate layout, data flow diagram (mic→action 10 steps), per-module
responsibility table, configuration layers, TTS pipeline diagram,
Lua sandbox details with API quick-ref, background-services overview,
"how to add a pack/feature/TTS backend" recipes, test coverage map,
build instructions with MSVC env, git workflow with Forgejo NO_PROXY trick.
- Aimed at someone who just cloned the repo and needs to fix a bug fast.
Build: cargo build --release -p jarvis-app -p jarvis-gui both green.
Tests: 49/49 jarvis-core unit tests pass.
Big push driven by user feedback ("делай имбу") and web research on what
voice assistants need to be the ideal:
TTS backend abstraction (P0.1)
- new module crates/jarvis-core/src/tts/{mod,sapi,piper,silero}.rs
- TtsBackend trait with SapiBackend (current PowerShell), PiperBackend
(rhasspy/piper, neural quality), SileroBackend (python subprocess)
- JARVIS_TTS env var picks (sapi|piper|silero). Auto-detect Piper if
binary + voice present in tools/piper/. Falls back to SAPI on missing.
- SpeakOpts {lang, detached, raw} replaces ad-hoc args. text_utils
sanitiser applied unless raw=true.
- llm_fallback + lua/api/tts both routed through tts::backend().
- tools/piper/install.ps1 downloads piper.exe + ru_RU-irina-medium.onnx
from rhasspy releases + huggingface. Smoke-test included.
- tools/silero/silero_tts.py helper (PyTorch); rust spawns it as subprocess.
IMBA-1 Agentic LLM router
- crates/jarvis-app/src/llm_router.rs
- When fuzzy/intent matcher fails, LLM picks the closest command from the
full registry. Returns JSON {command_id, confidence, reason}.
- Threshold-gated re-dispatch via substitute phrase. JARVIS_LLM_ROUTER=1
enables; JARVIS_LLM_ROUTER_THRESHOLD overrides 0.55 default.
- Inserted in app.rs::execute_command between "no match" and existing
llm_fallback chat fallback.
IMBA-2 Long-term memory
- crates/jarvis-core/src/long_term_memory.rs — JSON store at
APP_CONFIG_DIR/long_term_memory.json. Atomic write-through.
- remember/recall/search/forget/all/build_context API.
- Lua bindings: jarvis.memory.* (5 functions).
- llm_fallback auto-injects relevant facts (substring search of prompt)
into system message before LLM call.
- Pack resources/commands/memory_pack/ with 4 commands: remember, recall,
forget, list.
IMBA-3 Profile switching (work/game/sleep/driving/default)
- crates/jarvis-core/src/profiles.rs — JSON profiles at APP_CONFIG_DIR/profiles/
Auto-seeds 5 defaults on first run with personality + allow/deny lists +
greetings + emoji icons.
- active_profile.txt persists choice across restart.
- Lua bindings: jarvis.profile.{active,set,list,allows,active_name}.
- llm_fallback prepends profile personality to system prompt.
- Pack resources/commands/profile_switch/ with 6 voice triggers.
IMBA-4 Multimodal screenshot + vision LLM
- crates/jarvis-core/src/lua/api/vision.rs — gated on HTTP sandbox.
- jarvis.vision.screenshot() captures via PowerShell System.Drawing.
- jarvis.vision.describe(prompt?) sends base64 PNG to Groq vision model
(default llama-3.2-11b-vision-preview, override via GROQ_VISION_MODEL).
- Pack resources/commands/vision/ with 2 commands: describe + read_error.
P0.2 Continuous conversation grace window
- config::CONVERSATION_GRACE_MS = 30_000.
- app.rs: after command result, if grace_ms > 0 keep listening WITHOUT
re-wake for the grace duration. Existing CMS_WAIT_DELAY back-dated so
the existing timeout fires at start + grace_ms.
Tests: 24/24 jarvis-core unit tests pass (including 5 text_utils).
Build: cargo build --release -p jarvis-app and -p jarvis-gui both succeed
on Windows MSVC (VS 2026 Enterprise vcvars64).
Notes for setup:
- Piper voice install: pwsh tools/piper/install.ps1 (downloads ~90 MB).
- GROQ_TOKEN needed for IMBA-1 (router) and IMBA-4 (vision).
- All features are opt-in via env vars or auto-detect; existing SAPI +
fuzzy match path remains the default.
User complaint: SAPI reads every dotted acronym character-by-character with a
"точка" between letters. Unbearable on every speak. No real TTS analog
shipping yet (Silero/Coqui/Inworld in roadmap), so the immediate fix is text
preprocessing.
New module `jarvis-core::text_utils` with `sanitize_for_speech()`:
- Specific brand mapping: J.A.R.V.I.S. → Джарвис, U.S.A. → США, U.K. →
Британия, U.S. → США, S.O.S. → сос.
- Generic dotted-acronym collapse: any `[letter].[letter].[letter].(...)`
run of 3+ letters gets its dots stripped. "T.O.N." → "TON".
- URL stripping: anything starting with http/https gets replaced with the
word "ссылка". SAPI reading URLs char-by-char is unlistenable.
- Em/en-dash normalisation: — → -, « » " stripped.
- Whitespace collapse.
Wired in two places:
- `jarvis.speak(text, opts?)` Lua API runs sanitiser unless opts.raw=true.
- `jarvis-app/llm_fallback::speak_via_sapi` runs sanitiser before the SAPI
PowerShell shell-out for LLM auto-fallback replies.
5 unit tests in `text_utils::tests` covering jarvis-name collapse, generic
acronym pattern, URL stripping, dash normalisation, and "leave clean
russian sentences alone" — all green.
Real TTS upgrade (Silero v4 first cut) tracked as P0.1 in roadmap.
GUI front-end does not subscribe to IpcEvent::LlmReply, so the LLM
auto-fallback was effectively silent — user got the OK chime but no
spoken answer. Backlog item resolved.
llm_fallback::handle now also invokes speak_via_sapi() after firing
IpcEvent::LlmReply (both for successful replies and the fallback
error message). The text goes through System.Speech.Synthesis.
SpeechSynthesizer, with a preference pass that picks the first
installed ru-RU voice if present (Microsoft Irina / Pavel etc.) so
Russian replies sound right out of the box, and falls back to the
default English voice otherwise.
PowerShell is invoked fire-and-forget (no Wait, stdout/stderr null)
so the voice loop never blocks. A long answer keeps speaking while
jarvis returns to wake-word listening; the mic may pick up some of
the speech which is the same compromise voices::play_* already makes.
A cleaner version would pause pv_recorder for the duration of the
utterance — that goes in the backlog as part of the eventual
unified IpcEvent::Speak refactor.
Toggle: set JARVIS_LLM_TTS=false to disable (e.g. on a server where
the GUI is the speaker). Non-Windows builds get a no-op stub.
new module crates/jarvis-app/src/llm_fallback.rs holds an optional
LlmClient + ConversationHistory built once at startup. if GROQ_TOKEN is
unset the module logs a warning and stays disabled — voice commands keep
working as before.
both the wake-word voice path (recognize_command in app.rs) and the
gui-side text command path (process_text_command) now check whether the
recognized phrase starts with one of the configured trigger phrases
(ru: 'скажи' / 'ответь' / 'произнеси'). when it does, the remainder of
the phrase is sent to Groq and the reply is published as a new
IpcEvent::LlmReply { text } so the gui can speak it.
on api error the trailing user turn is popped from history, the russian
fallback line is sent over ipc and a 'error' voice cue plays. the loop
itself never panics.
Bossiara13