Outlook COM (`outlook_handlers.py`, agent):
- 4 handlers mirroring Rust pack: unread_count, latest, send_clipboard,
summarize_inbox. Uses `win32com.client` with lazy import so the module
loads even when pywin32 is missing — handlers degrade gracefully and
speak a "Outlook недоступен" reply.
- `requirements.txt`: pywin32>=305 (optional).
- 7 unit tests mock `_outlook_session` via unittest.mock.
Time tracker (`time_tracker_handlers.py`, agent):
- 5 handlers: start, stop, today, week, reset. JSON store at
`<here>/time_tracker.json`, atomic write-then-rename.
- Same shape as the Rust pack's `jarvis.state` blob: current_session_start
+ sessions[{start,end}]. Local-calendar today boundary.
- Full Russian pluralisation (час/часа/часов, минута/минуты/минут).
- 14 unit tests using the existing `isolated_state` fixture in conftest.py.
`extensions.py` + `main.py`: imports + 9 new proxy do_* functions + 9
new dispatch elifs. `commands.yaml`: 9 new entries.
Tests: 60 → 81 (+14 tracker + 7 outlook).
Python parity for the 5 personality voice commands shipped on the
Rust side:
- New `personality_handlers.py` (~150 lines): do_personality_{greet,
thanks, compliment, how_are_you, tony_quote}. Each picks one of
7-10 phrases per language (RU/EN) at random; greet additionally
buckets by time-of-day (morning/midday/evening/night).
- `extensions.py`: imports + 5 proxy `do_personality_*` functions
pointing at the new module, plus set_speak wiring so handlers can
TTS via the runtime's voice.
- `main.py`: 5 new dispatch elifs feeding the new action types.
- `commands.yaml`: 5 new entries (`personality_*`) mapping voice
phrases to action types.
- `tests/test_personality_handlers.py`: 6 unit tests confirming
each handler returns a non-empty string from the expected pool.
- `.gitignore`: stop tracking runtime state generated by tests
(profiles/, long_term_memory.json, schedule.json, macros.json,
llm_backend.txt, llm_history.json) — these are user data, not
source.
Tests: 54 → 60 (+6).
Plugin loader (mirrors Rust #29):
- plugins_store.py walks %APPDATA%\com.priler.jarvis\plugins\<name>\
for commands.yaml fragments and merges them into VA_CMD_LIST at
startup. Same schema as the project root commands.yaml. Per-pack
`disabled` flag file skips a pack without deletion. Built-in ids
win conflicts so a malicious pack can't silently override
open_browser. 9 unit tests (parity with the Rust tests).
- main.py: plugin merge runs after the root yaml load; logs the
number of extra commands picked up.
Flet GUI (#27 — first parity with the Rust Tauri GUI):
- New gui/ package: __main__.py launches via `python -m gui` (or
gui.bat). app.py routes between 7 pages (Главная / Команды /
Макросы / Расписание / Память / Плагины / Настройки) via
NavigationRail; dark theme matching the Tauri look.
- gui/services.py centralises every store-module call the pages
share — single place to swap implementations later.
- One file per page in gui/pages/, all expose `build(page)` and use
the same card layout idiom. Pages mirror the Tauri ones feature-
for-feature: home (status + counters + launcher), commands
(filterable list of all VA_CMD_LIST entries), macros (list +
delete + recording banner), scheduler (list + remove), memory
(add/forget/search), plugins (toggle/open folder), settings
(profile + LLM hot-swap).
- requirements.txt: flet>=0.85 (optional; assistant runs without).
- 3 smoke tests verify imports + build callables; skip cleanly when
Flet isn't installed.
README: documents both the GUI and the plugin layout in user-facing
terms. Tests: 42 → 54 (+9 plugins +3 GUI smoke).
Closes the Python parity gap with rust. All major imba features now mirrored
EXCEPT live LLM hot-swap (python LLM backend is still env-driven, not voice-
switchable like rust).
dev_handlers.py (new, ~250 lines)
Codebase Q&A:
do_codebase_set "укажи проект C:\path" → memory_store.remember("codebase.root", ...)
do_codebase_where "какой проект сейчас"
do_codebase_ask "что делает функция X" / "найди в коде Y"
Walks the root with same caps as rust:
depth=3, files=30, file_bytes=4096, total=50000
Filters by extension (rs/py/ts/lua/go/...).
Skips .git/target/node_modules/__pycache__/.venv etc.
Feeds digest to Groq LLM with "senior reviewer" prompt.
GitHub PR review (requires `gh` CLI authenticated):
do_github_set_repo "текущий репо owner/repo" → memory_store
do_github_list_prs "какие пиары" → `gh pr list --json number,title,author`,
speaks count + first 3 titles
do_github_summarize_pr "разбери последний пиар" → gh pr view of latest open
PR, feeds title+body+diff stat to LLM, speaks 3-5
sentence review.
extensions.py — 6 new proxy handlers + import dev_handlers + set_speak wires it.
main.py — 6 new action_type dispatches in run_action.
commands.yaml (+6 entries, 142 → 148):
codebase_set, codebase_where, codebase_ask,
github_set_repo, github_list_prs, github_summarize_pr.
Tests: ast.parse passes for all .py files. yaml.safe_load = 148 entries.
PYTHON PARITY VS RUST IS NOW COMPLETE except:
✗ LLM hot-swap voice command (rust calls llm::swap_to + persists DB;
python config.GROQ_TOKEN is read once at startup)
✗ GUI pages (rust has Tauri+Svelte, python is console-only)
Everything else has full parity: memory / profiles / vision / macros /
scheduler / codebase Q&A / GitHub PR review / 14 utility packs.
Mirrors `long_term_memory` and `profiles` modules from rust fork.
memory_store.py (new, ~95 lines)
- JSON store at <here>/long_term_memory.json (atomic write).
- remember(key, value) / recall(key) / forget(key) / search(query, limit) /
all_facts() / build_llm_context(prompt, limit).
- Uses the same JSON shape as the rust side, so the file is
interoperable if you ever symlink the two installations to the same dir.
profiles_store.py (new, ~145 lines)
- Profiles at <here>/profiles/<name>.json. active_profile.txt persists.
- Seeds 5 defaults on first init: default ★, work 💼, game 🎮, sleep 🌙,
driving 🚗 — same as rust fork.
- active() / active_name() / set_active(name) / list_names() /
allows_command(cmd_id).
extensions.py (+6 handlers, +160 lines)
- do_memory_remember — derives key (first 6 words or split on '=')
- do_memory_recall — substring search or top-3 if no query
- do_memory_forget — exact then substring fallback
- do_memory_list — count + first 5 keys
- do_profile_set — target via action.target or voice ("работа" → work)
- do_profile_what — speaks current profile + description
main.py
- +6 dispatch entries for the new action types.
commands.yaml (+10 entries, 119 → 129)
- memory_remember, memory_recall, memory_forget, memory_list
- profile_work, profile_game, profile_sleep, profile_driving, profile_default
- profile_what
Tests: ast.parse passes for all .py files. yaml.safe_load gives 129 entries.
Python parity is now meaningful: memory + profiles are usable by voice
without any rust install. Still missing for full parity: scheduler (needs
background thread), macros (needs phrase-capture state), vision (needs
Groq vision model call), codebase_qa, github_pr.
Mirrors the rust tier-3 batch. 90 → 100 commands (round number — feels
right for an "имба" milestone).
main.py:
+ do_brightness: WMI WmiMonitorBrightness read + WmiSetBrightness write,
±20% step or 100/10 absolute, "no_wmi" sentinel for desktops.
+ do_window_switch: same alias map as process_kill (xrom→chrome/...),
PS finds process by name OR window-title substring, picks newest
StartTime, AppActivate via WScript.Shell.
+ do_spell_out: split chars, join with ". ", _speak.
+ do_network_settings: action.section = wifi|bluetooth|network →
matching ms-settings:* URI via cmd /c start "".
+ do_my_ip: Get-NetIPAddress filtered to non-loopback non-APIPA IPv4,
WAN via Invoke-RestMethod ipify, speaks "Локальный X. Внешний Y".
+ do_summarize_selection: pyautogui.hotkey(ctrl, c) → 250ms wait →
Get-Clipboard → Groq @ temp 0.3 ("3-5 sentence retell") → speak +
clipboard.
commands.yaml +10 entries: brightness_up/down/max/min, switch_to_window,
spell_out, open_wifi_settings, open_bluetooth_settings, my_ip,
summarize_selection.
Backlog: closed all "Tier-3 still TODO" pure-Lua items. Remaining
TODOs (pomodoro, currency-math, stocks, translate-clipboard, daily
briefing, etc.) are even smaller LOC and can roll in the next batch.
Mirrors the rust analog-inspired tier-2 packs. yaml + ast smoke pass.
56 → 82 commands (+26 yaml entries; 14 of them are power/fun/wiki/
news/etc. permutations).
main.py — 7 new action handlers, all built on stdlib urllib.request
(no new dep). Helper _http_get() with User-Agent header so RSS feeds
that 403 default UAs go through.
+ do_news: pulls RSS (default Lenta.ru), extracts <title> via regex,
skips channel title (idx 0), takes 5, sends to Groq for a 2-3
sentence summary at temp 0.4. Falls back to reading first 3
headlines verbatim if no token.
+ do_currency_rate: cbr-xml-daily.ru daily_json.js. Recognises
USD/EUR/CNY/GBP/JPY/KZT in the phrase, computes value/nominal,
diffs against Previous, speaks Russian with direction.
+ do_crypto_price: CoinGecko simple/price for BTC/ETH/SOL/DOGE/TON
with 24h change, picked by phrase keywords.
+ do_fun_ask: joke / fact / quote / compliment via Groq at temp 1.0
with per-kind system prompts, random seed in user message.
+ do_wiki_lookup: ru.wikipedia opensearch → page summary, optional
Groq retell when extract > 350 chars.
+ do_power: dispatch table for shutdown / restart / sleep /
hibernate / logoff / cancel — runs shutdown.exe /s|/r|/h|/l|/a
or rundll32 powrprof,SetSuspendState. 30-second grace on shutdown
and restart with /c message so "отмени выключение" works.
+ do_help_commands: sorts VA_CMD_LIST.keys(), speaks first 12 + total,
full list to clipboard.
commands.yaml +26 entries:
news_headlines, currency_rate, crypto_price, joke / fact / quote /
compliment (action:fun_ask, kind: ...), wiki_lookup, shutdown_pc /
restart_pc / sleep_pc / hibernate_pc / logoff_user / cancel_shutdown
(action:power, op: ...), help.
Inspired by what mainstream voice assistants (Siri/Alexa/Алиса/
Cortana) ship by default + power-user utilities (shutdown,
window/process control).
main.py: prints a 3-line banner at startup with name, version, edition
label ("Python edition") and the command count from the loaded YAML.
Makes the console window self-identifying when both python and rust
forks run simultaneously.
run.bat at repo root: cd /d into repo, sets the cmd title to
"J.A.R.V.I.S. (Python edition)", picks the venv interpreter at
.venv\Scripts\python.exe if present (falls back to system python),
runs main.py, then pauses at exit so the user can read any traceback
instead of the console window vanishing on crash.
Desktop shortcut "Jarvis (Python).lnk" points at this batch file
with the project icon (the rust fork's icon.ico, since python has
no .ico of its own).
Mirrors the rust packs from the same iteration. yaml.safe_load + ast.parse
both pass.
main.py — six new action handlers + a _speak_in_lang helper:
+ do_translate: picks target language from "переведи на X" / "translate to X"
triggers, defaults to English, sends to Groq with strict "translation only"
prompt, drops to clipboard, speaks via SAPI in matching ISO voice
+ do_math_eval: strips trigger, Groq at temp=0.0 max_tokens=64,
short-circuits on "нет" sentinel for non-math
+ do_theme_set: writes AppsUseLightTheme + SystemUsesLightTheme via
Set-ItemProperty under HKCU\...\Themes\Personalize, action param
mode: dark|light
+ do_open_project / do_list_projects: reads/creates
%USERPROFILE%\Documents\jarvis-projects.json (a flat list of
{name, path}), exact-match → substring-match, opens in VS Code if
code.cmd on PATH else Explorer. First call writes a sample config
pointing under USERPROFILE so it works on any host.
+ do_open_sound_panel: rundll32 mmsys.cpl,,1 — Recording tab
commands.yaml — 8 new entries:
+ translate, math_eval
+ theme_dark, theme_light (action params mode:dark/light)
+ open_project, list_projects
+ open_sound_panel
SAPI voice selection iterates GetInstalledVoices() and picks first
matching TwoLetterISOLanguageName — Russian phrases sound Russian,
English/German/etc. sound right too, falls back to default if no
matching voice is installed.
Total commands: 56 (was 49).
Mirrors the new rust packs from the same iteration. Total commands in
commands.yaml is now 49 (was 33). yaml.safe_load + ast.parse both pass.
Window management (8 commands) — uses the existing `keys` action with
pyautogui hotkey strings: winleft+d (show_desktop), winleft+up/down
(maximize/minimize), winleft+left/right (snap left/right), alt+f4
(close_window), winleft+shift+m (restore_all), winleft+tab (task_view).
No new action type needed.
Clipboard read aloud — new clipboard_read action. Shells to
PowerShell Get-Clipboard, caps preview at 400 chars, speaks via the
existing tts.va_speak (which already handles recorder.stop/start).
Notes — new notes_add + notes_open actions. notes_add strips the
trigger phrase, appends "[YYYY-MM-DD HH:MM] body\n" to
%USERPROFILE%\Documents\jarvis-notes.txt, creates the dir if needed.
notes_open shells "cmd /c start" so .txt opens in the user-default
editor.
Process killer — new process_kill action. Strips the trigger ("убей
процесс"/"закрой программу"/etc.), looks up the remainder in the same
alias map as the rust version (хром→chrome, телега→telegram, спотифай
→spotify, едж/эдж/edge→msedge, дискорд→discord, стим→steam, обс→obs64,
проводник/блокнот/калькулятор/диспетчер задач, vs code→code, etc.),
appends .exe if needed, runs taskkill /F /IM, speaks success or
"не нашёл процесс N".
Web search — new websearch action with a `service` param. Bases:
google → https://www.google.com/search?q=
youtube → https://www.youtube.com/results?search_query=
wiki → https://ru.wikipedia.org/wiki/Special:Search?search=
yandex → https://yandex.ru/search/?text=
Strips trigger, urllib.parse.quote-s the rest, opens via webbrowser.open
(respects the user'\''s default browser; config.BROWSER_PATHS only kicks in
for the older url-action commands).
Portability: USERPROFILE used for notes path, no other hardcoded paths.
Brings the python fork in line with the new feature/pc-tools work that
landed in rust. Same UX, same trigger phrases, same env vars.
config.py:
+ LLM_AUTO_FALLBACK = True
+ LLM_AUTO_FALLBACK_MIN_CHARS = 4
Mirrors the rust config knobs. The old "скажи X" trigger still works
(VA_TBR strips it before intent matching, so what remains hits this
fallback anyway) — but now ANY unrecognized query goes to Groq, not
just "скажи" ones.
main.py:
va_respond: replaced the fuzz-on-"скажи" gate with a unified auto-
fallback that fires whenever no command matches, GROQ_TOKEN is set,
and the utterance is at least LLM_AUTO_FALLBACK_MIN_CHARS long.
+ _speak(text): recorder.stop() → tts.va_speak → start, used by the
new action handlers so the mic doesn't pick up the assistant.
+ _set_clipboard(text): subprocess to PowerShell Set-Clipboard, no
new python deps.
+ _strip_trigger(voice, triggers): shared helper for codegen/file_search
where we need the text AFTER the trigger phrase.
New action handlers (registered in run_action dispatch):
+ groq_codegen — extracts query, calls Groq with strict "code only"
system prompt at temp 0.2, strips ```lang ... ``` fences, puts the
result in the clipboard, speaks "Код в буфере, сэр. Строк: N".
+ ocr_screen — PowerShell System.Drawing screenshot to a temp PNG,
shells to tesseract.exe (-l rus+eng), speaks a 120-char preview,
clipboard. Looks for tesseract on PATH first then standard install
dirs; gives a clear "Tesseract не установлен" if missing.
+ sysinfo — CIM-based battery/cpu/ram/disk via PS one-liners, time
via python datetime (gets us localized weekday for free). Topics:
time / battery / cpu / ram / disk / all.
+ file_search — strips trigger, recursive Get-ChildItem in Desktop/
Documents/Downloads (depth 3, top 5), opens explorer /select on
the first hit, speaks count and basename.
+ audio_devices_panel — opens mmsys.cpl. Proper COM-based device
switching (like the rust IPolicyConfig version) is deferred — TODO
in backlog.
commands.yaml:
Total 33 commands (was 19). New:
+ open_notepad / open_calculator / open_explorer / open_terminal
/ open_settings / open_task_manager / lock_screen / screenshot
— all via shell action (no AHK exes needed).
+ volume_up / volume_down — five chained keys=volumeup/volumedown
presses each, like rust's _volume_helper.ps1 with Times=5.
+ media_stop — keys=stop.
+ sysinfo_time / sysinfo_battery / sysinfo_cpu / sysinfo_ram /
sysinfo_disk / sysinfo_all — sysinfo action.
+ file_search, codegen, read_screen, switch_audio_device.
Replaced (were AHK exes that Priler ships and we don't have):
~ music → keys playpause
~ music_off → keys playpause (toggle)
~ music_next → keys nexttrack
~ music_prev → keys prevtrack
Removed (Priler hardware/setup specific, can't sensibly port):
- gaming_mode_on / gaming_mode_off (display config switch)
- music_save (depends on the specific player AHK was driving)
- switch_to_headphones / switch_to_dynamics (replaced with the
generic switch_audio_device that opens mmsys.cpl)
Kept as-is (worked already, no AHK dependency):
open_browser / open_youtube / open_google (url action)
sound_off / sound_on (pycaw via system.volume_mute/unmute)
thanks / stupid / off
Smoke-tested: yaml.safe_load → 33 commands, ast.parse main.py → OK.
Portability: USERPROFILE for screenshot path, $env:SystemDrive for disk,
PATH lookup for tesseract before falling back to standard install dirs.
No hardcoded user-specific paths.
Adds optional 'browser' field to url-action; runtime resolves it via
config.BROWSER_PATHS (subprocess.Popen with the chosen exe). When
omitted, falls back to webbrowser.open (system default). Constructor
GUI gains a browser dropdown on URL action and on multi-step URL.
Existing open_browser/youtube/google updated to use yandex by default
since user prefers Yandex Browser.
shell uses subprocess (with shell=True so user can write a normal
command line); url uses webbrowser.open; keys uses pyautogui.hotkey
for combinations or pyautogui.typewrite for raw text. pyautogui is
imported lazily so the bot still starts on machines without a GUI.
Replaces the 100-line elif chain with a small dispatcher. Each entry in
commands.yaml now carries both 'phrases' (existing trigger list) and
'action' describing how to run it. Action types: exe, play_sound,
system (volume_mute/unmute/exit), sleep, multi.
The data migration lives in the next commit; the dispatcher in this
commit assumes the new schema.
Replace the fixed 10s post-wake-word listen window with a VAD-driven
loop. Once at least 500ms of speech have been seen, 1.2s of silence
closes the window. Hard cap at 15s as a safety net.
Tunable in config.py via VAD_AGGRESSIVENESS, COMMAND_END_SILENCE_MS,
COMMAND_MIN_SPEECH_MS, COMMAND_MIN_LISTEN_MS, COMMAND_MAX_LISTEN_MS.
- Names the assistant as J.A.R.V.I.S. (acronym + MCU pre-Age-of-Ultron)
- British butler tone, addresses user as «сэр»
- Brief replies (1-3 sentences)
- No prudish censorship
uses a second KaldiRecognizer with a JSON grammar of WAKE_WORDS + [unk] for
the wake phase, then hands off to the full-vocab recognizer for the 10s
command window. drops the pvporcupine dependency entirely.
