PC_Workman_HCK Files

Build Date: April 27, 2026 Status: Alpha Testing Target: Technical testers & early adopters

No Python required

---

What Changed - The Short Version

v1.6.8 was a monitoring tool that learned your habits. v1.7.2 is a system management tool that acts on them.

Three new management pages (Startup Manager, Services Manager, Optimization Hub), a completely rewritten AI layer with local LLM support, bilingual responses, proactive background monitoring, and a real .exe build that actually bundles everything correctly.

This is the biggest release since the project started.

---

New: Hybrid AI Engine (hck_GPT rewrite)

The entire hck_GPT subsystem was rebuilt from a single chat_handler.py into a modular engine with 5 subpackages.

Architecture

hck_gpt/
├── engine/        # Hybrid routing: rules first, Ollama LLM for everything else
├── intents/       # Intent parser, vocabulary, language detection
├── memory/        # Session memory, user knowledge (SQLite), proactive monitor
├── context/       # System context builder, hardware scanner
├── responses/     # Bilingual response builder (PL/EN)
├── chat_handler.py
├── insights.py
└── panel.py

How the Hybrid Engine works

Every message hits the intent parser first. Known intents (stats, alerts, help, temperature, processes) get handled by the rule engine, fast, predictable, no external dependency.

Low-confidence or open-ended messages ("why is my PC slow today?") get routed to Ollama (local LLM). The engine injects a 6-section system context into the Ollama prompt: live CPU/RAM/GPU state, today's averages, top processes, temperatures, hardware profile, and conversation history.

If Ollama isn't installed or unavailable, the engine falls back gracefully with a 60-second cooldown. No crashes, no hanging.

Bilingual (PL/EN)

Every response handler now replies in Polish or English based on auto-detected input language. ASCII-fold dual scoring handles Polish diacritics (ą, ę, ś, ź) without breaking intent detection.

Proactive Monitor

Background daemon thread watches:

• CPU sustained >80% for 30s → alert
• RAM >85% → pressure warning
• CPU throttling detection
• Disk space <10% warning
• Session uptime >12h reminder Alerts push to chat panel and banner status bar automatically. No user action needed.

Session Memory & User Knowledge

• CPU/RAM trend buffers track direction over time
• Auto conversation summary every 6 messages
• get_context_for_llm() builds full context string for Ollama
• SQLite user knowledge base at AppData/Local/ stores hardware profile, usage patterns, learned facts
• Background hardware scan via psutil and WMI (CPU model, GPU name, VRAM, motherboard, RAM speed)

Panel Redesign - Bordeaux Noir

• Animated black→crimson gradient banner with math.sin sine-wave shimmer
• Vector AI badge drawn on canvas (no image files)
• Pulsing ONLINE status badge
• Session timer in footer

Bug Fixes in hck_GPT

<img width="981" height="402" alt="image" src="https://github.com/user-attachments/assets/5c817416-af80-4172-b6c2-284488fa457c" /

• Language bug: _show_help() always used self._last_lang (set before language detection ran), so help always came out in the wrong language. Fixed to use detected ui_lang.
• Temperature bug: _resp_temperature() returned "Can't read temperatures on Windows" unconditionally. Now: 3-level fallback (1) psutil.sensors_temperatures(), (2) minute_stats from SQLite, (3) "no data" message. Shows CPU/GPU current/avg/max with OK/warm/HOT thresholds.
• Speed-up bug: _resp_speed_up_pc() always injected TURBO BOOST promotion and FPS tips regardless of actual system state. Removed unconditional block; now shows gentle suggestion only when system is genuinely clean.

---

New: Startup Manager

Full page at ui/pages/startup_manager.py.

Reads startup entries from three registry hives:

• HKCU\SOFTWARE\Microsoft\Windows\CurrentVersion\Run
• HKLM\SOFTWARE\Microsoft\Windows\CurrentVersion\Run
• HKLM\SOFTWARE\WOW6432Node\Microsoft\Windows\CurrentVersion\Run No admin rights required for reading.

Features

• Knowledge base of 30 common programs with impact rating (High / Medium / Low) and recommended action
• Three panels: Optimize at startup / Safe to disable / All entries
• Entry cards show: program name, impact badge, description, Disable/Delay/Restore buttons
• Disable = real winreg.DeleteValue() with confirmation dialog
• User preferences persist to data/cache/startup_prefs.json

---

New: Services Manager

Full page at ui/pages/services_manager.py.

Features

• Catalogue of 40+ Windows services in 4 categories:
• Essential (locked, cannot be stopped)
• Recommended (keep running)
• Optional (safe to stop)
• Likely Unnecessary (safe to disable)
• Batch status query via sc query type= all state= all
• Stop / Start / Restart buttons per service row
• Admin detection via ctypes.windll.shell32.IsUserAnAdmin() with warning banner when not elevated
• All changes logged to data/logs/service_changes.log

TURBO Mode Integration

• Optional/Unnecessary services get checkboxes for TURBO queue
• When TURBO activates then queued services auto-stop
• When TURBO deactivates then services restore to previous state
• Queue persisted to settings/turbo_services.json

---

New: Optimization Hub (My PC Central)

Replaced the single "Optimization & Services" button with a 3-zone interactive widget.

Implementation

Single tk.Canvas split into three clickable zones via coordinate detection:

• Left (57% width): Optimization Center, amber→dark red gradient
• Top-right: Startup Manager - navy to blue gradient and live metric (registry entry count)
• Bottom-right: Services Manager - green into emerald gradient and live metric (running services count) Hover effect: 1.25× brightness on active zone. Metrics read in background daemon thread, UI updated via canvas.after(0, _draw).

Other Central tab changes

• Removed hck_GPT banner from Central (was taking space without clear purpose)
• Added SESSION bar: SESSION: Xh Ym left-aligned + ● LIVE right-aligned, dark blue background
• Nav bar: added MY PC label in Inter Bold and vertical separator
• Tab/section fonts upgraded from Segoe UI to Inter
• Stability Tests + Your Account moved to bottom row (side by side)

---

New: hck_stats_engine query methods

query_api.py got four new methods:

Method	What it returns
get_temperature_history(minutes)	CPU/GPU temp: current, avg, max and estimated flag
get_temperature_summary(days)	Multi-day temp data from daily/hourly stats
get_top_processes_lifetime(top_n)	Top N processes by avg CPU across all recorded days
get_weekly_summary()	Current 7 days vs previous 7 days with trend (up/down/stable)

---

UI Polish

• Efficiency tab: Fixed physical core count display (was showing logical), fixed invisible avg text color, added per-core session min/max/avg, side-by-side TOP CPU/RAM consumers
• HCK_Labs page: Vector globe icon drawn with canvas primitives (sphere, meridian, equator and parallels)
• Dashboard nav buttons: Full redesign - dark gradient bg, bordeaux L-bracket accents, Segoe UI Black text, accent-color hover blend
• 7 vector icons: Monitor, exclamation, charts, lightning, fan, flask, book - drawn directly on Canvas, zero image files
• HCK_Labs & Guide pages: Redesigned as full blog-style layouts
• Turbo Boost: Set to coming-soon state with grey styling and floating tooltip

---

Release Packaging (EXE)

• setup.py bumped to 1.7.2
• startup.py uses APP_VERSION = "1.7.2" constant
• requirements.txt completed: added numpy>=1.24.0, requests>=2.28.0, pywin32>=305
• PCWorkman.spec fully rewritten:
• 25+ hiddenimports covering all hck_gpt.* submodules, new pages, utils.fonts
• settings/ directory added to datas (was missing — prefs didn't survive bundling)
• Fixed COLLECT(name='.') to name='PC_Workman_HCK_1.7.2'
• Output: dist/PC_Workman_HCK_1.7.2/PC Workman HCK.exe (10.3 MB launcher, ~94 MB total)

---

Codebase Cleanup

• Removed hck_GPT banner from Central tab
• Proper .gitignore (replaced .gitignore.txt) - excludes data/logs/, settings/user_prefs.json, assets/fonts/*.ttf
• Removed 7 orphaned files: report_window.py, 3 utils stubs, hck_gpt_panel.py, trend_analysis.py, google9bc8246e2e876106.html
• Fixed broken import paths in main_window.py (ui.page_* → ui.pages.page_*)
• Cleared all __pycache__/ directories
• utils/fonts.py — Inter font loader via Windows GDI32 with Segoe UI fallback

---

For Testers: What to Focus On

Primary Test Areas

1. Startup Manager - Does it correctly detect your startup programs? Do impact ratings make sense? Does Disable actually remove the entry?
2. Services Manager - Are service categories accurate? Does Stop/Start work? Does TURBO integration restore services properly?
3. hck_GPT bilingual - Try Polish and English messages in the same session. Does language detection work correctly?
4. Hybrid Engine - If you have Ollama installed, test open-ended questions. If not, verify graceful fallback.
5. Proactive alerts - Leave running during heavy load. Do alerts trigger appropriately?
6. EXE build - Run PC Workman HCK.exe from dist/ folder. Does it launch? Do all pages load?
7. Long-term stability - 24h+ runtime, check memory usage over time.

Known Limitations

• Services Manager requires admin rights for Stop/Start/Restart (detection works, actions fail without elevation)
• Ollama integration requires Ollama installed separately (not bundled)
• Temperature readings on some hardware may show estimated values (flagged in UI)
• Startup Manager knowledge base covers 30 programs - unknown programs show as "Unknown impact"

Performance Expectations

• CPU: ~1-2% idle, ~5-8% with dashboard active
• RAM: ~80-120 MB (depends on history size)
• Database: ~5-10 MB per month of runtime
• EXE cold start: ~3-5 seconds

---

What's Next

v1.7.3 – v1.7.8 (Roadmap to feature-complete):

• TURBO: Suspend inactive background processes (#21)
• TURBO: Auto power plan switching (#24)
• TURBO: Unpark CPU cores + priority boost (#23)
• TURBO: Complete suite integration + stability testing (#25)
• Core: Auto-update check on startup (#14)
• Code: Optimize core/monitor.py and hck_gpt/ modules (#16, [#17]) v2.0 (Q3-Q4 2026):
• Microsoft Store publishing (#26)
• Optional Ollama LLM integration as first-class feature (#28)
• ML pattern detection
• Predictive maintenance alerts

---

Installation

No new mandatory dependencies beyond what's in requirements.txt.

From .exe: Download from this release -> Extract -> Run PC Workman HCK.exe

From source:

:::bash
git clone https://github.com/HuckleR2003/PC_Workman_HCK.git
cd PC_Workman_HCK
pip install -r requirements.txt
python startup.py

Python 3.9+ required. Windows 10/11 only.

---

Feedback

Report issues: GitHub Issues Include logs from data/logs/ and screenshots when possible.

---

Built by Marcin "HCK" Firmuga | HCK_Labs License: MIT