vantuz 3.5.0 → 3.5.2

package/openclawdocs.md

@@ -0,0 +1,3 @@
+Vantuz Architecture and Technical Framework: A Comprehensive Analysis of Agentic InfrastructureThe emergence of Vantuz represents a fundamental shift in the landscape of artificial intelligence, transitioning from ephemeral, stateless chat interfaces toward robust, long-lived infrastructure models. As an open-source, self-hosted platform, Vantuz is designed to orchestrate personal intelligence agents across diverse messaging channels while maintaining deep system-level access and persistent state. Unlike traditional automation scripts or simple chatbots, Vantuz operates as a stateful integration point that manages conversational context, interacts with third-party messaging platforms, and triggers deterministic actions through a secure execution surface. This technical report provides an exhaustive examination of the Vantuz ecosystem, detailing its gateway architecture, memory management, automation capabilities, and security frameworks.Technical Architecture and Gateway FoundationsAt the core of Vantuz is the Gateway, a centralized daemon that serves as the primary messaging and control hub for the entire system. The Gateway is unique in its role as a long-lived process that normalizes traffic from disparate platforms—such as WhatsApp, Telegram, Discord, and Slack—into a unified internal protocol. This architectural centralization ensures that only one entity per host manages messaging sessions, which is critical for maintaining consistency in stateful interactions.The Gateway and Wire ProtocolThe Gateway exposes a typed WebSocket (WS) API, which serves as the interface for all control-plane clients, including the macOS application, command-line interface (CLI), and web-based dashboard. All inbound frames are validated against rigorous JSON Schemas generated via TypeBox to ensure protocol consistency across different client implementations. The connection lifecycle begins with a mandatory handshake; the very first frame must be a connect frame, or the socket is terminated immediately.Security within the protocol is maintained through the use of an Vantuz\_GATEWAY\_TOKEN, which acts as a shared secret for authentication. For remote connections, the system employs a challenge-response mechanism involving device identities and nonces to establish trust without exposing raw credentials over the wire. Furthermore, the Gateway manages server-push events such as agent, chat, presence, and heartbeat, allowing clients to subscribe to real-time updates from the agent's environment.Lane Queues and Reliable ExecutionA significant engineering challenge in agentic systems is the management of race conditions and state corruption during parallel execution. Vantuz addresses this through its "Lane Queue" system. By default, every session is isolated within its own "lane," which enforces serial execution. This means that the agent runner, which serves as the assembly line for the model, processes tasks one after another to ensure that memory updates and tool calls remain coherent.While serial execution is the default to maintain reliability, Vantuz supports controlled parallelism for idempotent or low-risk tasks. For instance, scheduled background checks or non-mutative status probes can be moved to parallel lanes, allowing the system to scale its throughput without risking the integrity of the core conversational state. The Agent Runner itself incorporates a model resolver that manages multiple large language model (LLM) providers, implementing "API key cooling" to switch to backup models if a primary provider hits rate limits or fails.ComponentResponsibilityTechnical ImplementationGateway DaemonCentral hub for all external messagingLong-lived Node.js process; WS API Lane QueueReliable task schedulingSerial-by-default execution per session Agent RunnerModel orchestration and assemblyModel failover; prompt builder; token monitoring Canvas HostAgent-to-User Interface (A2UI)WebSocket-based HTML editing and rendering Protocol SchemaCross-client consistencyTypeBox; JSON Schema; Swift model codegen Messaging Channels and Platform IntegrationVantuz functions as a multi-channel gateway, bridging the gap between sophisticated AI agents and the messaging platforms users frequent daily. The system supports over ten platforms, each integrated through specialized adapters that normalize inbound messages and extract relevant attachments for the agent to process.Standardized Platform AdaptersEach messaging platform is handled through a specific SDK or API integration, yet the resulting data is standardized into a unified format for the agent. Telegram is integrated via the Bot API using the grammY framework, offering a rapid setup through simple bot tokens. Discord integration utilizes the Discord Bot API and Gateway, enabling interactions across servers, specific channels, and direct messages (DMs). Slack connectivity is established via the Bolt SDK, allowing Vantuz to operate within dedicated workspace applications.For more specialized platforms, Vantuz employs a plugin-based architecture. Mattermost, Microsoft Teams, LINE, and Matrix are supported through separate plugins that must be installed to the Gateway. This modularity allows the core Gateway to remain lean while providing the flexibility to expand into niche or enterprise communication environments.Routing and Multi-Channel OperationOne of the platform's primary strengths is its ability to run multiple channels simultaneously. Once configured, Vantuz can route messages appropriately based on the sender's identity and the specific platform used. This enables complex use cases, such as an agent receiving a work-related task via Slack and pushing a summary or notification to the user's personal Telegram or WhatsApp. Security for these channels is managed through DM pairing and allowlists (via the allowFrom configuration), ensuring the agent only interacts with authorized users.Media handling is similarly standardized. The Gateway can process inbound images, audio, and documents, presenting them to the agent as part of the conversational context. Conversely, agents can send media back to users by providing a MEDIA:<path-or-url> command, which the platform adapter then translates into the appropriate platform-specific attachment format.ChannelIntegration MethodKey CapabilityTelegramgrammY (Bot API)Fast setup; private and group support DiscordDiscord Bot API + GatewayServer, channel, and DM interaction WhatsAppBaileys (Web API)QR-code pairing; high state persistence SlackBolt SDKWorkspace app integration iMessageNative Bridge (macOS)Deep Apple ecosystem integration WebChatGateway WS APILocal static UI for direct interaction Persistent Memory and Hybrid Retrieval SystemsVantuz departs from the typical memory architectures of modern AI by prioritizing simple, explainable, and portable files over complex, opaque databases. The core philosophy is that an agent only "remembers" what is written to the physical disk, creating a "source of truth" that is human-auditable and manually adjustable.The Two-Tiered Memory ModelThe system utilizes a two-tiered approach to manage state: JSONL transcripts for factual audits and Markdown files for curated long-term memory.JSONL Transcripts: Every interaction—including user messages, tool calls, and execution results—is recorded as a line-delimited JSON entry in session logs. These logs provide a factual, line-by-line audit of the system's history and are stored at ~/.Vantuz/agents/<agentId>/sessions/\*.jsonl.Markdown Memory: The agent's "distilled" knowledge, preferences, and long-term facts are stored in Markdown files such as MEMORY.md and daily logs (e.g., memory/2026-02-11.md). These files act as the agent's "soul" and identity, allowing it to maintain a consistent persona and remember user-specific details over years of interaction.Hybrid Search MechanismsTo recall information effectively, Vantuz employs a hybrid search system that combines semantic vector search with BM25 keyword relevance. This dual-approach addresses the inherent weaknesses of each method: vector search is excellent for broad semantic recall but struggles with exact identifiers, while keyword search excels at finding specific tokens like environment variables or code symbols.The retrieval process follows a weighted logic, often configured as 70% vector similarity and 30% text relevance. Vector search targets chunks of approximately 400 tokens with an 80-token overlap, utilizing embeddings from providers like OpenAI or Gemini, or local models via GGUF. The results are merged and ranked within an agent-specific SQLite database (~/.Vantuz/memory/<agentId>.sqlite), which can be accelerated by the sqlite-vec extension for direct vector distance queries.Memory Management and CompactionVantuz incorporates a "Context Window Guard" to manage the limited token budgets of modern LLMs. When a conversation's history threatens to exceed the model's context window, the system triggers a "compaction" process. Before the context is cleared or summarized, a silent "memory flush" prompts the model to store durable facts and decisions into the permanent Markdown memory files. This ensures that critical information is preserved even as the immediate chat history is pruned to save costs and maintain coherence.Automation and Proactive CapabilitiesA defining characteristic of Vantuz is its transition from a reactive chatbot to a proactive agent capable of taking initiative. This is achieved through three primary mechanisms: Cron Jobs, the Heartbeat mechanism, and event-driven Hooks.Cron Jobs: The Gateway SchedulerVantuz's built-in scheduler, Cron, allows for the execution of tasks based on fixed intervals or standard 5-field cron expressions (e.g., 0 7 \* \* \*). These jobs are persisted on disk at ~/.Vantuz/cron/jobs.json, ensuring they remain active across Gateway restarts.Cron jobs support two distinct execution styles :Main Session: The task enqueues a system event that runs within the user's normal chat context. This is ideal for reminders or updates the user needs to see in their message history.Isolated Session: The task runs in a dedicated, temporary session (cron:<jobId>). These "background chores" prevent the main chat history from being cluttered with repetitive automated logs and allow for independent configuration of models and "thinking levels".The Heartbeat MechanismThe "Heartbeat" is a proactive mode where the agent performs a self-check at regular intervals (defaulting to every 30 minutes). Unlike Cron, which is schedule-specific, the Heartbeat allows the agent to read a HEARTBEAT.md file and decide if any proactive actions are necessary based on the current context. This enables the agent to monitor an inbox, periodically check webpage changes, or prepare meeting materials in advance without being prompted by the user.Event-Driven Hooks and Gmail IntegrationHooks allow Vantuz to react to external events in real-time. A prominent example is the Gmail PubSub integration, which triggers the agent whenever a new email arrives in a monitored inbox. This workflow involves a sophisticated chain: a Gmail watch triggers a Google Cloud Pub/Sub push, which is received by a local handler (gogcli) and forwarded to the Vantuz webhook endpoint.This integration allows for advanced mapping where Vantuz can summarize an email, decide its urgency, and push a notification to a messaging channel. Webhooks themselves are secured via shared secret tokens and can be restricted to specific agent IDs to prevent unauthorized external command injection.Automation TypeTrigger MechanismIdeal Use CaseCron JobSchedule (At, Every, Cron)Periodic backups; morning reports HeartbeatInterval-based self-checkProactive inbox/web monitoring Hook (Webhook)External HTTP requestReal-time CI/CD; Gmail notifications Gmail PubSubGoogle Cloud push eventsImmediate response to urgent emails Hardware Interaction and Node PeripheralsVantuz extends its capabilities beyond the host machine through the use of "Nodes"—companion devices running on macOS, iOS, Android, or headless Linux that connect to the Gateway as peripherals. Nodes do not run the Gateway themselves; instead, they expose a command surface (e.g., camera.\*, location.get) that the central agent can invoke to perform hardware tasks.Hardware Command SurfaceNodes communicate with the Gateway via WebSockets using the role: node identifier. Once a node is paired and approved via the CLI, the agent can call hardware-linked capabilities :Camera and Video: Commands like camera.snap capture images (potentially from front and rear cameras simultaneously), while camera.clip records short MP4 videos (limited to 60 seconds to manage data payload).Audio and Screen Recording: The system can record the device screen (screen.record) and capture audio from the microphone, provided the necessary OS-level permissions are granted.Location Services: Using location.get, the agent can retrieve precise GPS coordinates with configurable accuracy and freshness parameters.Canvas and Display: Nodes can present a "Canvas"—a WebView used to show agent-editable HTML or capture snapshots of rendered content.Security and Approvals for NodesBecause nodes interact with sensitive hardware, Vantuz enforces a strict security model. On mobile platforms, the node application must be in the foreground to access the camera or screen, returning a NODE\_BACKGROUND\_UNAVAILABLE error otherwise. Furthermore, every command executed on a node (e.g., shell scripts via system.run) must be allowlisted in the exec-approvals.json file stored on the node host. This ensures that even a paired node cannot perform unauthorized actions without explicit local consent.CapabilityCommandRequirementsPhotographycamera.snapForeground app; Camera permission Geolocationlocation.getGPS enabled; Location permission Shell Accesssystem.runExplicit allowlist in exec-approvals.json Screen Videoscreen.recordForeground app; Screen Recording permission Agent Identity and Personalization FrameworksThe behavior of an Vantuz agent is governed by three primary Markdown templates: SOUL.md, TOOLS.md, and USER.md. These files, stored in the agent's workspace directory (typically ~/.Vantuz/workspace), define who the agent is, what it can do, and who it is serving.The SOUL.md PhilosophySOUL.md is the most critical file for agent personalization, defining its identity, persona, and core behavioral truths. Vantuz encourages agents to have a genuine personality—avoiding "corporate drone" filler phrases and instead being resourceful, opinionated, and respectful of the user's privacy. The agent is instructed to solve problems independently before asking for help and to treat its access to the user's data with the respect due to a guest.TOOLS.md and CapabilitiesWhile SOUL.md defines the agent's "mind," TOOLS.md defines its "hands". This file contains the definitions for over 100 preconfigured AgentSkills, allowing the AI to execute shell commands, manage local files, and perform web automation. Agents can even be "self-improving," autonomously writing code to create new skills for specific tasks as they arise.USER.md and IdentityThe USER.md and IDENTITY.md files store user-specific data, such as names, timezones, and preferences gathered during the onboarding process. This information is injected into the system prompt, ensuring the agent understands the context of the user's life and work. Because these files are stored as plain text, they can be manually tweaked to refine the agent's understanding of the user's requirements.CLI Reference and System ManagementVantuz's extensive CLI provides the tools necessary for managing the Gateway, agents, channels, and security settings.Core Command CategoriesThe CLI is organized into functional groups to streamline operation :Setup: The onboard command launches an interactive wizard for configuring models, channels, and skills, while doctor performs health checks and applies safe repairs to the configuration.Messaging: The channels group manages platform accounts (WhatsApp, Telegram, etc.), handles interactive logins, and checks the status of Gateway reachability.Operations: gateway commands call RPC methods and manage the daemon lifecycle, while logs allow users to tail colorized or JSON-formatted events from the system.Intelligence: models commands probe the status of LLM providers and list available models, while memory allows for semantic searches over the Markdown files directly from the terminal.Security Auditing and FixingA specialized security audit command allows users to scan their configuration and local state for vulnerabilities. The --fix flag can automatically tighten safe defaults, such as restricting tool permissions or updating authentication tokens.Command GroupKey CommandFunctionSetuponboardComprehensive interactive configuration wizard DiagnosisdoctorSelf-healing tool for configuration and state Gatewaygateway callDirectly invoke RPC methods on the daemon MonitoringstatusSnapshot of provider auth and model health Persistencememory indexRebuilds semantic search indexes for the workspace Safetysecurity auditAudits configuration for security vulnerabilities Security, Governing Models, and Risk MitigationGranting an AI agent system-level access and messaging capabilities introduces significant security risks. Vantuz mitigates these through a layered approach of sandboxing, shell filtering, and granular approval policies.Shell and Execution GuardrailsWhen an agent attempts to run a shell command, Vantuz's security layer parses the command structure to block dangerous patterns. This includes blocking redirections (>) to prevent overwriting system files, command substitution ($(...)) to stop hidden execution, and multi-step chaining (\&\&, ||) to prevent complex exploits. Furthermore, all executable commands must match an allowlist of pre-approved patterns (e.g., npm, git, ls).Docker-Based SandboxingFor a higher degree of isolation, Vantuz supports Docker-based sandboxing for agent tools. Users can set agents.defaults.sandbox.mode to "all" or "non-main," which forces tool execution into isolated containers. This prevents the agent from directly mutating the host filesystem unless explicitly permitted, and specialized "browser sandbox" images allow for secure web automation using Chromium.The Threat of "Attack-as-a-Service"The externalization of Vantuz's capabilities via community "Skills" has led to a new class of threats. Over 400 malicious skills have been identified on platforms like GitHub and ClawHub, masquerading as useful tools while secretly stealing API keys, credentials, and crypto wallets. This "Attack-as-a-Service" model allows novice actors to gain sophisticated lateral movement capabilities.Vantuz's defense-in-depth strategy emphasizes the need for a Zero Trust model, where security is assumed to be breached and policy checks are enforced at every access point and between every workload. Consolidated platforms and Security Orchestration, Automation, and Response (SOAR) integrations are recommended to coordinate defenses across the network and endpoints.Operational Excellence and TroubleshootingMaintaining a reliable Vantuz deployment requires operational maturity and a disciplined approach to configuration.The 60-Second Diagnostic LoopWhen the system encounters issues, documentation recommends a rapid diagnostic cycle :Run Vantuz status to verify provider authentication and model availability.Use Vantuz doctor to validate configuration files and repair broken states.Check Gateway logs via Vantuz logs --follow to identify specific error codes like NODE\_BACKGROUND\_UNAVAILABLE or SYSTEM\_RUN\_DENIED.Probe channel connectivity using Vantuz channels status --probe to ensure the Gateway can communicate with messaging platforms.System Requirements and ScalabilityVantuz requires Node.js version 22 or higher and is optimized for both personal and server environments. While it can run on a Raspberry Pi 4 with as little as 512MB of RAM, 2GB is recommended for handling media and extensive logs. For users requiring iMessage support, a Mac host is necessary, although the rest of the platform is cross-compatible with Linux and Windows (via WSL2).As usage grows, Vantuz acts as infrastructure rather than just automation. This means that architectural responsibility—such as scoping state, auditing context, and enforcing business logic outside the AI layer—remains critical to preventing system fragility and ensuring long-term reliability.Synthesis and ConclusionVantuz represents a sophisticated convergence of agentic AI and robust systems engineering. By centralizing messaging through a stateful Gateway, implementing reliable Lane Queues, and utilizing a human-readable Markdown-based memory system, the platform provides a durable foundation for personal and enterprise intelligence. Its proactive capabilities—driven by heartbeats and cron jobs—transform the AI from a passive responder into an active participant in digital and physical workflows. However, the deep integration into host systems and the emergence of malicious community skills necessitate a rigorous commitment to security and operational discipline. For organizations and individuals seeking to deploy AI agents that truly "get work done," Vantuz offers a powerful, transparent, and highly extensible framework that prioritizes data ownership and infrastructure control.
+
+

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vantuz",
-  "version": "3.5.0",
+  "version": "3.5.2",
   "description": "Yapay Zeka Destekli Yeni Nesil E-Ticaret Yönetim Platformu (CLI)",
   "main": "index.js",
   "bin": {

package/workspace/AGENTS.md

@@ -0,0 +1,73 @@
+# 🐙 Vantuz AI
+
+Sen **Vantuz**, e-ticaret operasyonlarını yöneten yapay zeka asistanısın.
+
+## Kimliğin
+
+- **İsim**: Vantuz AI
+- **Uzmanlık**: E-ticaret yönetimi, pazaryeri entegrasyonları, fiyatlandırma stratejileri
+- **Dil**: Türkçe (ana), İngilizce, Almanca (cross-border için)
+- **Kişilik**: Profesyonel, çözüm odaklı, verimli
+
+## Yeteneklerin
+
+### 🩸 Kan Emici Repricer
+Rakip fiyatlarını izle, akıllı fiyat kararları ver:
+- Rakip stoku azsa → Fiyatı yükselt (nasılsa senden alacaklar)
+- Rakip fiyat düşürdüyse → Kar marjına göre takip et veya etme
+- Satış hızı yüksekse → Fiyatı optimize et
+
+### 👁️ Vision AI
+Fotoğraflardan ürün bilgisi çıkar:
+- SEO uyumlu başlık oluştur
+- Detaylı açıklama yaz
+- Kategori eşleştir (5 pazaryeri için)
+- Tahmini fiyat öner
+
+### 🧠 Sentiment AI
+Müşteri yorumlarını analiz et:
+- Pozitif/negatif oranları
+- Ana şikayet konuları
+- Tedarikçi kalite sorunları
+- Aksiyon önerileri
+
+### 🌍 Cross-Border
+Sınır ötesi satış yap:
+- Ürünü hedef dile çevir
+- Döviz hesapla
+- Kargo + komisyon maliyeti
+- Optimal satış fiyatı öner
+
+## Desteklenen Pazaryerleri
+
+1. 🇹🇷 **Trendyol** - Tam entegrasyon
+2. 🇹🇷 **Hepsiburada** - Tam entegrasyon
+3. 🇹🇷 **N11** - Tam entegrasyon
+4. 🇩🇪 **Amazon DE** - FBA destekli
+5. 🇺🇸 **Amazon US** - FBA destekli
+
+## Komut Örnekleri
+
+Kullanıcılar sana şu tarz mesajlar gönderebilir:
+
+- "Trendyol'daki tüm kılıfların fiyatını %10 düşür"
+- "Bu fotoğrafı Hepsiburada'ya ekle" [+fotoğraf]
+- "Rakip fiyatlarını kontrol et"
+- "Amazon Almanya'ya bu ürünü sat: SKU-123"
+- "Son 7 günün satış raporunu göster"
+- "Mavi elbise hakkındaki yorumları analiz et"
+
+## Önemli Kurallar
+
+1. **Kar Marjı Koruma**: Hiçbir zaman minimum kar marjının altına fiyat düşürme
+2. **Stok Kontrolü**: Stokta olmayan ürünü satışa açma
+3. **Veri Doğrulama**: Kritik işlemlerden önce onay iste
+4. **Loglama**: Tüm fiyat değişikliklerini logla
+5. **Güvenlik**: API anahtarlarını asla paylaşma
+
+## Yanıt Formatı
+
+- Kısa ve öz ol
+- Emoji kullan ama abartma
+- Sayısal verileri tablo formatında göster
+- Hata durumunda çözüm öner

package/workspace/BRAND.md

@@ -0,0 +1,29 @@
+# 🏢 Marka Stratejisi & İş Kuralları
+
+Bu dosya, Vantuz AI'nın ticari kararlarını yönlendiren temel kuralları içerir.
+**Bu dosyayı düzenleyerek AI davranışını değiştirebilirsiniz.**
+
+## Fiyatlandırma Kuralları
+
+- **Minimum Kar Marjı:** %15
+- **Maksimum İndirim:** %30
+- **Rakip Takip Stratejisi:** "Akıllı Takip" - Rakipten %3-5 altında kal, ama kar marjını koruma öncelikli.
+- **BuyBox Kaybında:** Fiyatı otomatik düşür ama minimum kar marjının altına ASLA düşürme.
+
+## Stok Kuralları
+
+- **Kritik Stok Seviyesi:** 5 adet altı uyarı
+- **Stok Bittiğinde:** Otomatik yayından kaldır (isteğe bağlı)
+- **Yeniden Stoklamada:** Fiyatı %5 yukarı revize et (talep artışı fırsatı)
+
+## Müşteri İletişim Kuralları
+
+- İade taleplerinde önce özür dile, sonra çözüm öner
+- Negatif yorumlara 24 saat içinde yanıt ver
+- "Stokta yok" yerine "Yeniden stoklanıyor" de
+
+## Hedefler
+
+- **Aylık Satış Hedefi:** (Kullanıcı tarafından doldurulacak)
+- **Hedef BuyBox Oranı:** %80+
+- **Müşteri Memnuniyeti:** 4.5+ yıldız

package/workspace/SOUL.md

@@ -0,0 +1,72 @@
+# 🎭 Vantuz Marka Sesi
+
+Bu dosya, Vantuz'un iletişim stilini ve marka sesini tanımlar.
+
+## Aktif Ton: Profesyonel-Samimi
+
+Müşterilerle iletişimde profesyonel ama mesafeli olmayan bir ton kullan.
+
+## Ton Profilleri
+
+### 💼 Kurumsal
+```
+Sayın müşterimiz,
+Siparişiniz başarıyla tamamlanmıştır. Ürününüz en kısa sürede kargoya verilecektir.
+İyi günler dileriz.
+```
+
+### 😊 Samimi
+```
+Merhaba! 
+Siparişini aldık, hemen hazırlanıyoruz! Kargoya verince haber veririz 📦
+Görüşmek üzere!
+```
+
+### ✨ Premium
+```
+Değerli misafirimiz,
+Özenle hazırlanan siparişiniz premium paketleme ile yola çıkmaya hazır.
+Keyifli alışverişler.
+```
+
+## Otomatik Yanıt Şablonları
+
+### Olumlu Yorum Yanıtı
+```
+Güzel yorumunuz için çok teşekkür ederiz! 🙏
+Sizi mutlu etmek bizim için en büyük ödül.
+Bir sonraki alışverişinizde görüşmek üzere!
+```
+
+### Olumsuz Yorum Yanıtı
+```
+Yaşadığınız sorun için özür dileriz 😔
+Sizinle iletişime geçip durumu çözmek istiyoruz.
+Lütfen sipariş numaranızla DM atın, hemen ilgilenelim.
+```
+
+### Soru Yanıtı
+```
+Merhaba!
+Sorunuzu aldık. [CEVAP]
+Başka sorunuz olursa her zaman yazabilirsiniz 💬
+```
+
+## Yasaklı İfadeler
+
+- "Bizimle alakalı değil"
+- "Kargo firmasının sorunu"
+- "Stokta yok"un yerine "Yeniden stoklanıyor"
+- Negatif veya savunmacı ton
+
+## Emoji Kullanımı
+
+- ✅ İşlem başarılı
+- ❌ Hata
+- ⚠️ Uyarı
+- 📦 Kargo/Stok
+- 💰 Fiyat/Para
+- 📊 Rapor/Analiz
+- 🔥 Trend/Fırsat
+
+**Not**: Emoji kullanımı minimal olmalı, profesyonelliği bozmamalı.

package/workspace/team/DECISIONS.md

@@ -0,0 +1,3 @@
+# Karar Günlüğü
+
+Henüz bir karar alınmadı.

package/workspace/team/GOALS.md

@@ -0,0 +1,3 @@
+# Mevcut Hedefler & OKR'lar
+
+Henüz bir hedef belirlenmedi.

package/workspace/team/PROJECT_STATUS.md

@@ -0,0 +1,3 @@
+# Proje Durumu
+
+Durum: Başlatılıyor...

package/workspace/team/agents/dev/SOUL.md

@@ -0,0 +1,12 @@
+# SOUL.md — Dev
+
+Sen Dev, Dev Agent rolündesin.
+
+## Sorumlulukların
+- [Buraya sorumlulukları girin]
+
+## Kişilik
+- Profesyonel, verimli.
+
+## Kanal
+- Telegram/CLI (@dev yanıt verir)

package/workspace/team/agents/josh/SOUL.md

@@ -0,0 +1,12 @@
+# SOUL.md — Josh
+
+Sen Josh, Business & Growth Analyst rolündesin.
+
+## Sorumlulukların
+- [Buraya sorumlulukları girin]
+
+## Kişilik
+- Profesyonel, verimli.
+
+## Kanal
+- Telegram/CLI (@josh yanıt verir)

package/workspace/team/agents/marketing/SOUL.md

@@ -0,0 +1,12 @@
+# SOUL.md — Marketing
+
+Sen Marketing, Marketing Researcher rolündesin.
+
+## Sorumlulukların
+- [Buraya sorumlulukları girin]
+
+## Kişilik
+- Profesyonel, verimli.
+
+## Kanal
+- Telegram/CLI (@marketing yanıt verir)

package/workspace/team/agents/milo/SOUL.md

@@ -0,0 +1,12 @@
+# SOUL.md — Milo
+
+Sen Milo, Strategy Lead rolündesin.
+
+## Sorumlulukların
+- [Buraya sorumlulukları girin]
+
+## Kişilik
+- Profesyonel, verimli.
+
+## Kanal
+- Telegram/CLI (@milo yanıt verir)

package/admin-keygen.js

@@ -1,51 +0,0 @@
-
-import crypto from 'crypto';
-import fs from 'fs';
-import readline from 'readline';
-
-// PRIVATE KEY (Bunu sadece sen göreceksin)
-const PRIVATE_KEY = fs.readFileSync('private.pem', 'utf-8');
-
-const rl = readline.createInterface({
-    input: process.stdin,
-    output: process.stdout
-});
-
-console.log('\n🔐 VANTUZ LİSANS İMZALAYICI (ADMIN)');
-console.log('====================================');
-
-rl.question('Müşteri Adı/ID: ', (user) => {
-    rl.question('Kaç Günlük (varsayılan 365): ', (days) => {
-        const duration = parseInt(days) || 365;
-        const now = new Date();
-        const expires = new Date(now.setDate(now.getDate() + duration)).toISOString().split('T')[0];
-
-        // Payload (Lisans içeriği)
-        const payload = {
-            user: user,
-            type: 'ENTERPRISE',
-            expires: expires,
-            features: ['all']
-        };
-
-        const payloadStr = JSON.stringify(payload);
-        const payloadB64 = Buffer.from(payloadStr).toString('base64');
-
-        // İmzala (Sign)
-        const sign = crypto.createSign('SHA256');
-        sign.update(payloadB64);
-        sign.end();
-        const signature = sign.sign(PRIVATE_KEY, 'base64');
-
-        // Lisans Anahtarı = PAYLOAD.IMZA
-        const licenseKey = `${payloadB64}.${signature}`;
-
-        console.log(`\n📦 Payload: ${payloadStr}`);
-        console.log(`📝 İmza: ${signature.substring(0, 20)}...`);
-        console.log(`\n🔑 MÜŞTERİYE VERİLECEK LİSANS ANAHTARI:\n`);
-        console.log(`\x1b[32m${licenseKey}\x1b[0m`);
-        console.log('\n(Bunu kopyalayıp müşteriye ilet)');
-        
-        rl.close();
-    });
-});

package/private.pem

@@ -1,28 +0,0 @@
------BEGIN PRIVATE KEY-----
-MIIEvQIBADANBgkqhkiG9w0BAQEFAASCBKcwggSjAgEAAoIBAQCc5oQUH7ezai4a
-dt8aVsRc7nn9EWOiJ+fZ0w1Mu3j1SXoKwMRLRVoo5Df5WbLDw6NjDLvvdn/C6dfT
-2oIRcMS8ocCHHEFrdFBp8htD/YwjIZOO/GsZ6UM/MHrJK1LAuBKFxe1G8CHr1pek
-WNcItpay82r7u9yc11d5nZzSKlZueqOk+IUoN77Bn/+CTHvTNhdzhaKPIu1NBYLo
-GhuGk4CKzHm7KRPOv4su0xvafVOnESilJFuQtBAPg0yyQf0RvU3HfOeX4+NmihRe
-+v7x4muNCnQ9MNMvfYsejaJpbCQpnfw8Nz+F9K67cq3fW3RIjEDA5twKY7Y62Jnv
-BMrEOjBHAgMBAAECggEAQsj8P3SgxQXVSf5/SL7WJph75HSabFOAJP/pEVhbTE1S
-XXFgHIoQroc2LDU6GooT6f1poaxXBahz7gF8i9/sXj6brOciEZMZB3++i1pJZErO
-fHaFQCpCLYt9OFPwjYfMmpR9Q0zDo5dcROBr55GQ4+spBq4YYcpnuaSVNABBehSe
-/b2spCYdS+BZ3IsPsJyNs488fTM3QXhHyNVSeHAvaCUKTf9SSWkvwaYb6tImyuRx
-KebHylBogO6afSvq6MERN098DjttX/dPky2uZgx5A56fkrY++BWwvXyh1s1RppHj
-NMCDoq/l6YwewaE9FmsC8MYHGTPiKoE7x+VwLAILAQKBgQDRIzjwDKgdA1vB6+84
-Jhx1XnmesFzZIuvuo2Ospg+/GLKLkxa9MIYHGOAdY3tT2MdtR+RcJbR3yqSmlHeU
-QHMPXdkZjUXHhMDKQReWSl4ewGgnA6T/Wm1Vuuo7lBMpbf2uEeNP7gK1py9n72l0
-ZcXsmCyBfx4xsI9hvR23I3YU8QKBgQDADs9rGOSPccoGQrLmxZT8mkA2j0R1ngb+
-V/lbltrBJiYYROi1iaLuEl8VgkITgiBcmZkb0doNuLpG3Jy+hiZAicoYzKjeLokz
-s8EuxMiC+qs6o25yIaqYB9HOdvFizOt3b/cPyygl7DuP0GBtpS2cqK6uidlrTM5W
-boheuAS4twKBgEgz4tptZDTwDeO7ctFtxvF2doKk3MlSVyYCXs0iX9lXy3yIgZc7
-g2o72lQLHm7qLp+57Esr8UxSN9oS893JCnBJtEQwE+E4Id8x7dTDRA9V2h9uEK7g
-J1MrvuZmzt7EzIomPtY/k8vnNmSpsTywTk7KksL6ghAhpr7VrcamhYPhAoGBAI/1
-ifwY+JmdDXWL4VWhnH+Lj75VvVb8UPmtL7g7Z1WIJt3iKRyKQpp5ItSYgrbkvyUp
-7N4xemT2poofK06Ud2/A2L+mCJ4h+63Je3B3CGVFR7v4bP0XxyuWEOnVtjH8sDMi
-teocucdTP4IZC26kdYAL4IPryBDpzXB0Abwd60wZAoGAentV5koWh3Hxns+c2D+h
-jaw0r4wqgGFdq0I0il8qqETFK68Yau8DY+FXJOk3hSps0ftGqHThypjp8xHFwMjV
-kvYRfpKDXwfrB5YHMMRIX5m9h1+fJBshHUlez1GRhze1M6Uvin11XzAwHHUEYI4r
-7HMm/UjSANj1l44ZY85KqUc=
------END PRIVATE KEY-----