ApexOS — Complete System Architecture

Personal AI Operating System for @induwarauthsara · Running on Azure B2ms · Managed by Hexa (Hermes v0.14)

Cronjobs Active

Hermes Skills

Notion Databases

Cronjobs Planned

Phases

Weeks

What's Working

n8n Container — Up & Running healthz — {"status":"ok"} 10/10 Notion Databases 10 Hermes Cronjobs Active Google Calendar Sync Gmail Connected GitHub Backup (daily) SSL: apexos.induwara.dev 🔒

Needs Attention

n8n Workflows — 9 inactive (0 nodes) AI Chatbot — 504 HTML error Google Tasks API — missing OAuth scope Crypto DB — missing PnL formula n8n — Chatbot Health Check v2 still present Certs DB — zero entries Playwright — installed only in venv

🔑 Architecture Decision: Sunset n8n for Execution

All 9 n8n workflows have 0 nodes — they're empty shells. Hermes cronjobs with LLM skills were tested and work reliably (email-triage, crypto-alert, calendar-sync all ran successfully).
Decision: Move all automation to Hermes cronjobs. n8n stays only as a WhatsApp webhook receiver.

🔄 Problem → Solution Map

All 17 of Induwara's needs mapped to ApexOS phases

🎓 Study ManagementDeadlines don't auto-surface. No learning tracker.Phase 1A

🛠 Personal ProjectsProjects DB exists. No progress tracking, no learning pipeline.Phase 1B

💰 Solo Startup + ERPNo payment reminders. ERP/CRM DB empty. No customer pipeline.Phase 1C

🤖 Portfolio ChatbotDetection works. No auto-fix. Backend currently DOWN.Phase 2A

🛡 CTF & HackathonsNo auto-template, no upcoming event alerts. Writeups stagnate.Phase 2B

🔐 Cyber SecurityAd-hoc learning. No structured curriculum. No spaced repetition.Phase 2C

📧 Email TriageLLM classification unreliable (OpenRouter credits). No fallback.Phase 2D

📈 Crypto & TradingPrices tracked but no execution, no paper trading, no backtests.Phase 3A

👥 Community EventsManual event prep. No slide generator, no RSVP tracking.Phase 3B

📅 CalendarSync works but doesn't classify events. Google Tasks API broken.Phase 3C

⏱ Time ManagementReactive. No daily time-blocking. No "what now?" guidance.Phase 3D

🔬 Tech ResearchOnly HN. No arXiv, no release notes, no YouTube summaries.Phase 4A

🎥 Vlog PipelineBacklog grows. No monthly batch enforcement. No auto-catalog.Phase 4B

✍ Blog WritingData-rich but unwritten. No auto-draft from projects/events.Phase 4C

🎖 CertificationsNo study plan, progress tracker, or exam date alerts.Phase 4D

🌐 PortfolioUptime unknown beyond chatbot. No SSL or deploy monitoring.Phase 5A

🏫 Campus LifeManual entry only. No lecture capture or assignment breakdown.Phase 5B

📋 Execution Plan — 6 Phases, 4 Weeks

🔴 Phase 1: Study & Projects Week 1 ▼

1A — Study Engine

Never miss a deadline. Know what to study daily.

Populate Study DB with current semester modules
Polish study-briefing skill — sort by due date, flag <48h as CRITICAL
New cron: study-deadline-check — 7:30 AM daily
New skill: notion-study-capture — auto-log assignments from chat
Auto-save explanations to Notes/Learning DB when you ask me to explain topics

1B — Project Tracker

Track all projects. Auto-convert learning to notes.

Populate Projects DB (ERP, Portfolio, AutoCryptoBot, Web3, Community, ApexOS)
New skill: project-status — query active projects, show milestones
New cron: project-check-in — Mondays 10AM
New skill: learn-deep — structured curriculum, spaced repetition

1C — ERP Engine

Activate ERP/CRM DB — customer entries, payment tracking
Polish payment-check cron — formatted Telegram alerts
n8n webhook for WhatsApp reminders (keep 1 n8n WF)
New skill: erp-customer-add

🟠 Phase 2: Infrastructure & Security Week 2 ▼

2A — Chatbot Self-Healing

⚠ Chatbot currently DOWN — returns 504 HTML error

Enhance chatbot_wrapper.py — attempt SSH restart on failure
New script: chatbot_fixer.py — restart, check logs, verify
Escalate: Telegram alert + Notion P1 task if unfixable
Need from you: What runs the backend? (Docker? PM2? Serverless?)

2B — CTF Pipeline

New cron: ctf-event-scanner — Sundays 12PM, scrapes CTFtime
CTF prep reminder 48h before events
Polish ctf-log skill — auto-generate writeup templates
New skill: ctf-prep — past writeups, tools to install

2C — Cyber Security Learning Path

Structured curriculum: 5 levels (Linux basics → Red Teaming)
New cron: security-news — daily 8AM, CVEs + Hacker News
Spaced repetition: 1-day, 3-day, 7-day, 30-day intervals
CTF skill mapping: track which domains you've mastered

2D — Email Triage v2

Rule-based fallback classifier (keyword + sender rules)
LLM only for ambiguous cases (use Google Gemini free API)
Urgent emails → immediate Telegram alert
Daily digest at 9PM for skipped emails

🟡 Phase 3: Crypto + Community + Time Week 2-3 ▼

3A — AutoCryptoTrading Bot

Paper trading framework (virtual portfolio, Binance prices)
Trade signals: EMA crossover, RSI, volume spikes
Risk mgmt: max 5% per trade, 20% total exposure
New skill: crypto-trade — double confirm for real trades
CRITICAL: All logs labeled "PAPER TRADING — NOT REAL MONEY"

3B — Community Engine

New skill: community-workshop — 5-part outline, slide structure
Presentation assistant — search notes, suggest demos
Post-event: auto blog + vlog entry + attendee follow-up

3C — Calendar Intelligence

Fix Google Tasks OAuth scope (re-auth required)
Auto-classify events: CTF → CTF DB, Workshop → Community DB, Birthday → skip
Conflict detection in morning briefing
Evening time-blocking suggestions

3D — Daily Command Center

Complete morning-briefing rewrite with structured sections
Interactive Telegram buttons: [✅ Done] [⏰ Snooze] [📋 Details]
Evening wrap-up at 9PM — completed vs rolled over

🟢 Phase 4: Knowledge & Content Week 3 ▼

4A — Research Engine

Beyond HN: arXiv (CS.AI, CS.CR), GitHub trending, Docker/K8s blogs, RSS
Smart filtering: stack-matching + actionable vs opinion
Weekly research briefing — Fridays 6PM, top 5 items
YouTube research: auto-summarize tutorials for MLOps, Web3, Go

4B — Vlog Pipeline

Monthly vlog batch day — 1st Sunday, auto-count unedited vlogs
Auto-generate YouTube descriptions + title suggestions
Urgency indicator: days since recorded

4C — Blog Pipeline

Auto-draft from projects (Problem → Solution → Code → Lessons)
Auto-draft from events (Overview → Moments → Tech → Takeaways)
Blogger API integration (future phase)

4D — Certification Tracker

Populate Certs DB: Binance, Anthropic, Azure (AZ-900, AZ-204), AWS
Study plan generator with topic breakdown
New cron: cert-progress-check — Sundays 8PM

🔵 Phase 5: Quality of Life Week 4 ▼

5A — Portfolio Uptime

Monitor portfolio repo for commits → check deploy status
Auto-generate project descriptions when you build something new
Monthly SSL expiry check for induwara.dev

5B — Campus Life

Recurring lecture tasks from semester timetable
Assignment breakdown: Research → Outline → Draft → Review → Submit
Lecture note auto-capture from web research

🟣 Phase 6: System Health Week 4 ▼

Health dashboard — check nginx, n8n, Hermes, cron every 6h
Disk space monitor every 12h — alert >80%, auto-clean cache
Monthly Azure B2ms cost monitor
Weekly data hygiene — dedup Notion entries, orphaned relations, stale tasks
All health events → Notion System Log page

📅 Execution Timeline

Week 1 — Foundation

Study Engine · Morning Briefing Rewrite · Email Triage v2

Week 2 — Core Pipelines

Chatbot Self-Healing · CTF Pipeline · Paper Trading · Calendar Intelligence

Week 3 — Knowledge & Content

Research Engine · Vlog/Blog Pipelines · Security Learning Path · Project/ERP Trackers

Week 4 — Polish & Quality

Community Engine · Certs · Portfolio · Campus · System Health · Data Hygiene

⏰ Complete Cronjob Schedule (18 Jobs)

#	Name	Schedule	Type	Delivers To	Status
1	☀️ morning-briefing	07:00 daily	LLM	Telegram	✅ Active
2	🎓 study-deadline-check	07:30 daily	LLM	Telegram	New
3	🔐 security-news	08:00 daily	LLM	Telegram (critical)	New
4	💰 payment-check	08:00 Monday	LLM	Telegram	✅ Active
5	🛠 project-check-in	10:00 Monday	LLM	Telegram	New
6	🛡 ctf-event-scanner	12:00 Sunday	LLM	Telegram	New
7	📅 calendar-sync	Every 2 hours	LLM	Local	✅ Active
8	📧 email-triage	Every 2 hours	Script+LLM	Local (urgent→TG)	✅ Active
9	📈 crypto-alert	Every 4 hours	LLM	Telegram	✅ Active
10	🤖 chatbot-health	Every 6 hours	Script	TG (if broken)	✅ Active
11	🩺 system-health	Every 6 hours	Script	TG (if down)	New
12	🔬 tech-digest	18:00 Friday	LLM	Telegram	✅ Active
13	📊 weekly-review	09:00 Sunday	LLM	Telegram	✅ Active
14	🎖 cert-progress-check	20:00 Sunday	LLM	Telegram	New
15	🎥 vlog-reminder	10:00 1st Sun	LLM	Telegram	✅ Active
16	🌙 evening-wrap-up	21:00 daily	LLM	Telegram	New
17	🔄 hermes-git-backup	00:00 daily	Script	Telegram	✅ Active
18	💾 disk-space-check	Every 12 hours	Script	TG (if >80%)	New

🧩 Skill Inventory — 20 Skills

Existing 12

apexos-databases
morning-brief
study-briefing
weekly-review
crypto-status
ctf-log
content-pipeline
notion-task-capture
notion-quick-log
github-monitor
google-workspace-sync
email-scanner

New to Create 8

project-status
learn-deep
erp-customer-add
ctf-prep
community-workshop
crypto-trade
notion-study-capture
system-health-check

Scripts 7

chatbot_health_check.py ✅
chatbot_wrapper.py ✅
chatbot_fixer.py NEW
email_classifier.py NEW
crypto_paper_trader.py NEW
hermes_git_backup.sh ✅
system_health.sh NEW

🎓 Trackademic Integration

Trackademic is Induwara's academic OS — a Next.js + Supabase web app for study management.

📦 Trackademic Tech Stack

Frontend: Next.js 16 (App Router), React 18, Tailwind CSS, Radix UI, Recharts
Backend: Supabase (PostgreSQL), pgvector (embeddings), Row-Level Security
AI: Google Gemini API, OpenRouter, RAG pipeline
Integrations: Google Calendar OAuth, Stripe, PDF parsing

Key Features:
• AI Study Coach with context-aware chat
• Grade Predictor (what-if scenarios)
• Credit-weighted Smart Focus Engine
• Bidirectional Google Calendar sync
• Pomodoro timer with analytics
• Past paper matrix + progress tracking
• Document RAG (upload PDFs → AI Q&A)

🗄 15 Database Tables

profiles, semesters, subjects (with credits, weights, targets), sessions (lecture/practical log), assignments, exams, past_papers, study_sessions, documents + document_chunks (RAG), ai_insights, calendar_tokens, reschedule_log, chat_history, monthly_attendance view

🔗 ApexOS ↔ Trackademic Integration

Trackademic handles the data layer for academics. ApexOS provides the intelligence and notification layer. Here's how they work together:

1. Morning Briefing pulls from Trackademic

ApexOS morning briefing queries Trackademic's Supabase directly: assignments due today, attendance alerts, grade predictions. Merges with Notion data for a unified view.

2. Telegram → Trackademic auto-logging

When Induwara tells Hexa "Attended DSA lecture today on Stacks", Hexa creates a session entry in Trackademic Supabase. Zero manual data entry.

3. Complementary AI

Trackademic AI Coach = structured academic content (questions from slides, grade predictions). ApexOS learn-deep = hobby/project learning (teach me Docker from ground level). Two AIs, two domains.

4. Trackademic → ApexOS Sync Cron

New cron: trackademic-sync — every 2 hours: fetch assignment deadlines, attendance risks, exam dates from Trackademic Supabase → inject into ApexOS context for morning briefing and alerts.

5. Shared Identity

Both use the same Google OAuth identity. Induwara logs into Trackademic via Google. ApexOS's Google token can query Trackademic Supabase.

🚀 How ApexOS Can Contribute to Trackademic

Data Population: Auto-fill lecture logs, assignments, and exam entries from Telegram conversations
Study Plan Generation: Use learn-deep skill to suggest structured topics for each subject
Attendance Auto-Update: Detect "attended X lecture" from chat → update Trackademic session
Cross-System Alerts: Trackademic attendance risk → ApexOS Telegram alert → Induwara acts
Research Pipeline: ApexOS tech-digest finds relevant content → saves to Trackademic document vault
Writeup Pipeline: After CTFs → ApexOS generates writeup → Trackademic stores as document
Supabase Schema Extension: Add ApexOS-specific fields (synced_at, source="telegram") for traceability