feat: habit tracker with gamification + KB updates

Merge feature/habit-tracker into master (squashed): ✨ Habit Tracker Features: - Bead chain visualization (30-day history) - Weekly lives recovery system (+1 life/week) - Lucide icons (zap, shield) replacing emoji - Responsive layout (mobile-optimized) - Navigation links added to all dashboard pages 📚 Knowledge Base: - 40+ trading basics articles with metadata - Daily notes (2026-02-10, 2026-02-11) - Health & insights content - KB index restructuring 🧪 Tests: - Comprehensive test suite (4 test files) - Integration tests for lives recovery - 28/29 tests passing Commits squashed: - feat(habits): bead chain visualization + weekly lives recovery + nav integration - docs(memory): update KB content + daily notes - chore(data): update habits and status data Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
2026-02-11 11:03:54 +00:00
parent a2eae25fe1
commit 719cd5f743
97 changed files with 19631 additions and 1233 deletions
--- a/memory/kb/youtube/2026-02-10-claude-multi-agent-orchestration.md
+++ b/memory/kb/youtube/2026-02-10-claude-multi-agent-orchestration.md
@@ -0,0 +1,128 @@
+# Claude Code Multi-Agent Orchestration with Opus 4.6, Tmux and Agent Sandboxes
+
+**Sursă:** <https://youtu.be/RpUTF_U4kiw>  
+**Autor:** Andy Devdan  
+**Durată:** 24:02  
+**Data salvare:** 2026-02-10  
+
+---
+
+## TL;DR
+
+Andy demonstrează noua funcționalitate de **multi-agent orchestration** din Claude Code (Opus 4.6), combinată cu Tmux și agent sandboxes (E2B). Arată cum un agent principal poate crea echipe de agenți specializați care lucrează în paralel pe task-uri separate, fiecare în propriul context și sandbox izolat. Demonstrația include 8 aplicații fullstack one-shotted de Opus 4.6, apoi re-deployate automat de două echipe de câte 4 agenți.
+
+**Mesaj cheie:** Limitarea nu mai e modelul (Sonnet 4.5+ sunt suficient de capabile) — limitarea suntem NOI: abilitatea de prompt engineering, context engineering și de a construi sisteme agentic reusabile.
+
+---
+
+## Puncte cheie
+
+### 1. Multi-Agent Orchestration = Next Level
+- **Noul workflow:** Create team → Create tasks → Spawn agents → Work in parallel → Shutdown → Delete team
+- **Tooling nou:** `team_create`, `team_delete`, `task_create`, `task_list`, `task_get`, `task_update`, `send_message` (comunicare inter-agent)
+- **Tmux = vizualizare:** Fiecare sub-agent rulează într-un pane separat; poți vedea în timp real 4-8 agenți lucrând simultan
+
+### 2. Agent Sandboxes (E2B) = Secure Playground
+- Izolare completă: agenții lucrează off-device, zero risc pentru local machine
+- Reusable: `/reboot` command pentru re-deploy aplicații în sandboxuri noi
+- Scalabil: Andy avea 24 de sandboxuri active simultan
+
+### 3. Multi-Agent Observability = Critic
+- Sistem custom de logging care captează toate evenimentele: session start/end, tool calls, task updates, messages
+- Poți urmări fiecare agent individual: ce tools a rulat, câte token-uri, status
+- **Fără observability = vibe coding** (nu știi ce se întâmplă → nu poți îmbunătăți)
+
+### 4. Real Constraint = You & I
+- Modelele pot deja mult mai mult decât știm noi să extragem
+- **Limitările reale:**
+  1. Tool-urile disponibile (agent sandboxes, tmux, observability)
+  2. **Noi înșine:** prompt engineering + context engineering
+- **Fiecare engineer e limitat de:** knowledge of tools + ability to engineer prompts/context
+
+### 5. Practical Workflow Demo
+- **Step 1:** Opus 4.6 creează 8 aplicații fullstack (one-shot, fără intervenție)
+- **Step 2:** Agent principal primește task: "Build agent team for each codebase, summarize setup"
+- **Step 3:** Primary agent deschide 8 tmux panes, spawns 8 Haiku agents (cheaper, specialized)
+- **Step 4:** Fiecare agent explorează 1 codebase, face setup, returnează summary
+- **Step 5:** Primary agent compilează tot într-un raport final (context = doar 31% din limit!)
+- **Step 6:** Repeat pentru re-deploy în E2B sandboxes (2 teams × 4 agents)
+
+### 6. Best Practices
+- **Specialized agents > generalist:** Un agent = un task specific → finish → shutdown
+- **Reset context:** După fiecare team workflow, ștergi team-ul și task-urile → fresh context
+- **Iterative:** Dacă ceva nu merge (ex: 2 apps lipsă date), spin up ad-hoc team să repare
+- **Observability first:** Nu poți scala dacă nu știi ce fac agenții
+
+---
+
+## Quote-uri importante
+
+> "The game on the field is changing. It's no longer about what the models allow us to do. As of Sonnet 4.5, these models can do much more than you and I give them credit for. The true constraint of agentic engineering now is twofold: **it's the tools we have available and it's you and I**."
+
+> "**The true limitation is you and I.** It is our capabilities. It's our ability to prompt engineer and context engineer the outcomes we're looking for and build them into reusable systems."
+
+> "How can you prompt engineer and context engineer with great powerful models to get more engineering work done than ever **with confidence**? We want to be building **systems of trust** with our agents."
+
+> "You want to spin up **specialized agents that do one thing extraordinarily well**. They focus on one task and then they finish."
+
+> "This whole idea that engineers are going to be replaced by this technology to me is absurd. **Engineers are the best positioned to use agentic technology**."
+
+> "Every engineer is limited by their tools and **their knowledge of their tool**. That's why multi-agent observability is super key."
+
+> "It all comes back to things we talk about on the channel all the time: **Core Four — Context, Model, Prompt, Tools**. Everything boils down to that."
+
+> "Scale our compute to scale our impact." _(tema recurentă)_
+
+---
+
+## Takeaways pentru Marius
+
+### 1. Multi-agent orchestration = aplicabil pentru ROA
+- **Scenarii potențiale:**
+  - Un agent principal coordonează 3-4 agenți: unul pt. migrare DB, altul pt. testing, altul pt. documentație
+  - Feature complex (ex: raport nou ROA) → agent team: unul design UI, altul backend logic, altul validări ANAF
+  - Maintenance tasks: un agent verifică logs, altul rulează backup-uri, altul face health check
+
+### 2. Observability = critical pentru încredere
+- Dacă implementezi agenți în producție pentru clienți → ai nevoie să știi EXACT ce au făcut
+- Simplu: log toate tool calls + results într-un DB/fișier → audit trail complet
+
+### 3. Context engineering > model size
+- Nu trebuie să folosești Opus pentru totul
+- **Pattern:** Primary agent (Opus/Sonnet) coordonează, sub-agenți (Haiku) execută task-uri simple
+- **Economie:** Haiku e mult mai ieftin, perfect pentru task-uri repetitive
+
+### 4. Sandboxing = safety net
+- Dacă ai agenți care modifică cod/baze de date, rulează-i în containere izolate (Docker/LXC)
+- Zero risc pentru sistem principal
+
+### 5. Fresh context after each workflow
+- Delete team + tasks după finalizare → evită "context pollution"
+- Fiecare task mare = fresh start
+
+---
+
+## Tools menționate
+
+- **Claude Code** (Opus 4.6, Sonnet 4.5, Haiku)
+- **Tmux** (vizualizare multi-pane pentru agenți paraleli)
+- **E2B** (agent sandboxes cloud-based)
+- **Multi-agent observability system** (custom logging, link în descriere)
+- **Agent sandbox skill** (custom skill pentru management E2B)
+- **Tactical Agentic Coding** (cursul lui Andy, promovat în video)
+
+---
+
+## Tags
+
+@work @scout #ai #claude #multi-agent #orchestration #automation #prompt-engineering #context-engineering #observability #sandboxes
+
+---
+
+## Acțiuni posibile
+
+- [ ] Testează multi-agent orchestration în Claude Code (experimental flag)
+- [ ] Construiește observability simplu pentru agenți ROA (log tool calls → SQLite)
+- [ ] Explorează Docker/LXC sandboxing pentru agenți care modifică cod
+- [ ] Studiază pattern: Primary (Opus/Sonnet) + Workers (Haiku) pentru economie
+- [ ] Implementează "fresh context" workflow: șterge history după task finalizat
--- a/memory/kb/youtube/2026-02-10-openclaw-10x-powerful.md
+++ b/memory/kb/youtube/2026-02-10-openclaw-10x-powerful.md
@@ -0,0 +1,161 @@
+# I made my OpenClaw 10x more powerful (seriously)
+
+**URL:** https://youtu.be/cod50CWlZeU  
+**Durată:** 43:45  
+**Data salvare:** 2026-02-10  
+**Tags:** @work @tools #openclaw #automation #ai-agents
+
+---
+
+## TL;DR
+
+Tutorial complet pentru configurarea avansată OpenClaw pe VPS (Hostinger), acoperind: upgrade web search la Perplexity Pro, configurare multi-model (Opus/Sonnet switch), security audit, context engineering cu 9 fișiere core (agents.md, soul.md, user.md, etc.), living files theory, heartbeat proactiv, cron jobs automatizate, și structură agentic company (team OpenClaw + personal OpenClaw per angajat).
+
+---
+
+## Puncte Cheie
+
+### 1. **Setup inițial (VPS vs Mac Mini)**
+- VPS = mai ieftin, accesibil remote (~$10-15/lună Hostinger)
+- Mac Mini = $600+ upfront, local
+- **Hostinger one-click deploy** - simplifică setup-ul dramatis (API keys via UI)
+
+### 2. **Upgrade Web Search → Perplexity Pro**
+- Default Brave Search = slab
+- **Perplexity Sonar Pro Search** (via OpenRouter) - web search rapid
+- **Perplexity Sonar Pro Deep Research** - research profund (trigger: "deep research")
+- OpenClaw se self-upgrade-uiește: schimbi config prin plain English
+
+### 3. **Multi-Model Setup (Opus/Sonnet/Codex)**
+- Opus 4.6 = scump, pentru taskuri creative/importante
+- Sonnet 4.5 = cheaper Opus, general use
+- GPT-5.3 Codex = coding/debugging + FREE cu ChatGPT subscription ($20 sau $200/lună)
+- Comenzi: `/model opus`, `/model sonnet`, `/model codex`
+
+### 4. **Context Engineering - 9 Core Files**
+- **agents.md**: Operating rules, workflows, permissions
+- **soul.md**: Personalitate, ton, valori (verbose/concis, Christian/secular, etc.)
+- **user.md**: Info despre tine (identity, timezone, location, goals)
+- **memory.md**: Long-term memory, lessons learned, IDs
+- **tools.md**: API docs, gotchas, how to use tools
+- **identity.md**: Rol agent (marketing/coding/recruiting), team org
+- **heartbeat.md**: Proactive checks (email, calendar, files)
+- **bootstrap.md**: Startup instructions
+- **user.md folders**: /personal și /business cu goals.md, problems.md, research.md
+
+### 5. **Living Files Theory** 🔥
+- **Dead files**: Google Drive, MacBook local → inaccessibile AI agents
+- **Living files**: VPS markdown files → AI poate citi/modifica/îmbunătăți/folosi ca context
+- **Advantage**: Orice research/deep search salvat → reutilizabil forever (nu pierdut în ChatGPT history)
+- Exemplu: "save web search results as markdown in /docs"
+
+### 6. **Security Best Practices**
+- **NU expune gateway public** (use Tailscale/VPN)
+- **Strong SSH passwords**, disable port 22
+- **NU leak API keys** (never share publicly)
+- **NU conecta la Moldbook/public forums** → prompt injection risk
+- Weekly 30-60min security audit → "analyze VPS and tell me how exposed we are"
+
+### 7. **Heartbeat System (Proactive AI)**
+- Default 30 min, recomandat 15 min
+- Citește heartbeat.md: check email, calendar, files
+- Dacă nimic → "HEARTBEAT_OK" (silent)
+- Dacă ceva important → trimite notificare (WhatsApp/Discord)
+- **Optimization**: Use Haiku 4.5 pentru heartbeat (cheap, fast) vs Opus (scump)
+
+### 8. **Cron Jobs (Automation Nirvana)**
+- Plain English: "create cron job every day 11pm Dubai time reminding me to sleep"
+- Replacement pentru n8n/Zapier cu AI decision-making
+- Exemple:
+  - Daily morning priorities + meeting prep (citește calendar + email)
+  - Weekly "check best AI model and upgrade config"
+  - Track wife's period (lol)
+  - Industry trends monitoring
+- **Benefit**: Connect once (Google Calendar, Gmail, etc.) → forever automations
+
+### 9. **Agentic Company Structure** 🏢
+- **Personal OpenClaw** per angajat (context personal + acces team)
+- **Team OpenClaw** central (company docs, financials, goals, team structure)
+- Inter-agent communication: John's agent ↔ Michael's agent ↔ Team agent
+- **Timeline**: Companies adopt this by mid-2026 (not 12-18 months, 2-4 months!)
+- **Competitive advantage**: 10-20 hours/week savings per person
+
+### 10. **Model Selection Strategy**
+- **Opus**: Creative/important tasks (priorities, strategic decisions)
+- **Sonnet**: Day-to-day execution (cheaper Opus)
+- **Codex**: Coding, debugging, technical work (+ FREE with ChatGPT sub)
+- **Haiku/Gemini Flash**: Heartbeat checks (cheap, fast)
+
+---
+
+## Quote-uri Memorabile
+
+> "The difference between a beginner with a free ChatGPT subscription and an advanced user running multiple OpenClaws has never been greater. We're going to see people who as a single person can build multi-million dollar companies, if not billion dollar companies."
+
+> "Living files theory: Every single file becomes exponentially more valuable if it's on a VPS accessible with a powerful AI agent that can modify it, use it as context, share it, build upon it, make decisions from it."
+
+> "Learning how to verbalize your thoughts, your goals, your preferences, your judgment and turn them into clear markdown files that we give to agents — this is going to be the most valuable skill in 2026."
+
+> "You might say 'Okay David, but you just gave it a prompt, a bit more info about you. What's the big deal?' The big deal is that it has it stored forever on the VPS. Every hour you put into improving your OpenClaw will pay off dividends forever."
+
+> "If you cannot verbalize your preferences, you cannot start automating tasks and workflows in your life."
+
+> "OpenClaw is in its own VPS. I haven't approved a single thing. It just does whatever it wants to do. This really gets more powerful as AI models get faster and more competent than humans."
+
+---
+
+## Acțiuni Practice (pentru Marius)
+
+### ✅ Deja făcut:
+- [x] OpenClaw pe VPS (moltbot)
+- [x] Anthropic API (Opus/Sonnet)
+- [x] Heartbeat configurat
+- [x] Context engineering started (agents.md, soul.md, user.md)
+
+### 🔄 De făcut (prioritate):
+1. **Upgrade web search la Perplexity Pro** (Sonar Pro Search + Deep Research via OpenRouter)
+2. **Security audit** - "analyze VPS and tell me how exposed we are"
+3. **Living files structure**:
+   - Creează /personal și /business folders
+   - Populate goals.md (business: $X/month, personal: health/learning)
+   - problems.md (current roadblocks)
+   - research.md (industry trends, notes)
+4. **Multi-model aliases**: `/model opus`, `/model sonnet`, `/model codex`
+5. **Heartbeat optimization**: Switch la Haiku 4.5 pentru heartbeat (cost savings)
+6. **Cron jobs utile**:
+   - Morning report (8:30) - priorities + meeting prep
+   - Evening report (20:00) - summary + tomorrow prep
+   - Weekly industry trends
+   - ANAF monitor (existing, deja rulează)
+
+### 🚀 Viitor (Q2-Q3 2026):
+- [ ] Team OpenClaw pentru Romfast (company docs, financial data, team structure)
+- [ ] OpenClaw per coleg (angajat nou, colegă 70 ani)
+- [ ] Inter-agent communication (personal ↔ team)
+
+---
+
+## Insights pentru Echo
+
+**Lecție cheie:** "Living files" = game changer pentru cum organizăm knowledge. Orice research/deep search salvat în markdown pe VPS → reutilizabil forever vs pierdut în ChatGPT history.
+
+**Aplicație imediată:**
+- memory/kb/ = living files system (youtube, coaching, insights, projects)
+- Dashboard files.html = access la living files
+- Orice YouTube processing, email insights, calendar research → salvat în kb/
+
+**Security mindset:**
+- Weekly 30-60min security audits
+- NEVER leak API keys/gateway tokens
+- Use Tailscale (deja configurat) + strong SSH
+
+**Team vision (Romfast):**
+- Marius = personal OpenClaw (clawd VPS)
+- Romfast = team OpenClaw (separate VPS, company context)
+- Angajat nou + colegă = personal OpenClaw (teach them setup)
+- Goal: Q3 2026 - inter-agent communication functional
+
+---
+
+**Status:** ✅ Processed  
+**Next:** Update notes index