Prompt used: generate a simple website about the vector AI A...

Prompt

Prompt used: generate a simple website about the vector AI Agent apple shortcut on routinehub- self contained html Vector AI Agent Your personal AI assistant, built entirely inside a Siri Shortcut. Vector is a powerful, privacy-first AI agent that runs directly on your iPhone — no app store download required. Built by Mincofficial & Elc29, Vector turns Siri Shortcuts into a fully conversational AI with tool-calling capabilities, screen awareness, memory, and multi-model support. Whether you need help coding, managing your schedule, researching online, or just having a conversation, Vector is designed to handle it all with intelligence, speed, and a personality that evolves with you. What Makes Vector Different Unlike traditional AI apps that lock you into a single interface, Vector lives inside the Shortcuts app you already use every day. This means: No new app to learn — Vector feels native to iOS because it IS native to iOS Deep system integration — Access contacts, calendars, reminders, and more directly Privacy by design — Your data stays on your device unless you explicitly choose to share it Always evolving — Vector rewrites its own personality file after every conversation, becoming more attuned to how you communicate True agentic behavior — Vector doesn't just chat; it takes action. It can send messages, set alarms, browse the web, execute code, generate images, and manage files all in one continuous loop What Vector Can Do Vector isn't just a chatbot. It's a full-featured AI agent with access to dozens of tools and capabilities: Communication Send iMessages — Text anyone in your contacts without opening the Messages app Send WhatsApp messages — Direct WhatsApp integration for international messaging Send emails — Compose and send emails with custom subjects and recipients Make phone calls — Dial numbers directly from your conversation Search contacts — Find phone numbers, emails, addresses, birthdays, and notes for anyone in your address book Productivity & Organization Set timers — Start countdown timers for cooking, workouts, or focus sessions Set alarms — Create alarms for specific times with custom labels Set reminders — Add reminders to your default list with optional notes Create calendar events — Schedule meetings with start/end times, all-day support, and notes Create Apple Notes — Save ideas, lists, and information directly to your Notes app Information & Research Web search — Search the internet with configurable snippet depth. Vector always follows up by browsing pages in full, not just relying on search summaries Browse web pages — Read full articles and sources to provide accurate, complete answers Execute Python code — Run calculations and data processing in a sandboxed environment with 64MB memory and up to 10,000 lines of code Get current weather — Vector always knows the weather at your location for contextual responses Read your screen — With your permission, Vector can take a screenshot and read what's on your display to provide contextual help Read your clipboard — Vector can see what's copied to help with pasting, referencing, or continuing tasks Creative & Development Generate images — Create AI art from text prompts using Pollinations image generation Save code artifacts — Store HTML, Python, JavaScript, and other code files in a dedicated Vector-Artifacts folder View artifacts — Display saved files, with HTML files rendering fully in a web view Delete artifacts — Clean up old projects and files Text-to-speech — Convert any text into spoken audio files Tell jokes — Pull random jokes from the Official Joke API when you need a laugh System & Navigation Open URLs and deep links — Launch websites, app-specific pages, or Siri Shortcuts directly Access deep link information — Vector knows hundreds of iOS deep links for Settings, apps, and system features Run other shortcuts — Vector can trigger other Siri Shortcuts as part of its workflow Memory & Personality Short-term memory — Vector remembers facts you tell it and references them in future conversations Soul file — A continuously evolving personality document that Vector rewrites after every interaction, capturing your preferences, communication style, ongoing projects, and relationship dynamics Customization mode — When enabled, Vector actively modifies its behavior based on your feedback ("be more concise," "use more humor," etc.) Model Selection Vector supports a wide range of AI models through Pollinations AI, letting you choose the perfect brain for every task. Switch models anytime through Vector Config. Featured Models — The Big Three Model Badge Description Minimax M3 🧠 Smart The leading open-source model. Surpasses GPT-5.5 and Gemini 3.1 Pro on coding benchmarks and approaches Opus 4.7. Features MiniMax Sparse Attention (MSA) with a 1 million token context window, native multimodality (image and video input), and frontier agentic capabilities. Best for complex coding, long-horizon reasoning, and advanced tasks where accuracy matters most. Minimax M2.7 ⚖️ Balanced 230B parameter Mixture-of-Experts model with 10B active per token and a 200K context window. Deeply agentic — capable of building complex harnesses and completing elaborate productivity workflows. This is the recommended default for most users because it strikes the perfect balance between intelligence and speed. Step 3.7 Flash ⚡ Fast StepFun's 198B sparse MoE tuned for multi-step reasoning, tool use, and long-context workflows. Pushes up to 400 tokens per second with a 256K context window. Perfect for fast, reliable agentic work when you need answers quickly. Complete Model Lineup Smart (🧠) — Maximum capability for complex tasks These models prioritize reasoning depth, coding ability, and long-context understanding over raw speed. Choose these when working on difficult problems, writing complex code, or doing deep research. Minimax M3 — The leading open-source model. Surpasses GPT-5.5 and Gemini 3.1 Pro on coding benchmarks (SWE-Bench Pro: 59.0%) and approaches Opus 4.7. Features MiniMax Sparse Attention (MSA) with a 1 million token context window, native multimodality (image and video input), and frontier agentic capabilities. The first and only open-weight model to combine frontier coding, 1M context, and native multimodality. Specialty: Complex coding, agentic work, and long-horizon reasoning. Kimi K2.6 — Moonshot AI's flagship open-source model. 1T parameters, 262K context, multi-turn tool calling, and agent swarm capabilities. Specialty: Agentic work, coding, and long-horizon planning. Deepseek V4 Pro — DeepSeek's flagship built for long-horizon reasoning and real-world agent work. 1.6T-parameter MoE with 49B active per token and a 1M-token native context. Capable of repo-level coding, document-heavy analysis, and multi-step automation. Specialty: Coding, math, and reasoning. GLM-5.1 — Z.AI's 754 billion parameter MoE model with standout performance in front-end UI generation, full-stack scaffolding, and React/Vue component creation. Scores 45.3 in Claude Code evaluation. Specialty: Coding and web development. Balanced (⚖️) — The sweet spot of speed and intelligence These models deliver strong performance without the latency of the largest models. Ideal for daily use, general chat, and most productivity tasks. Minimax M2.7 — The recommended default. MiniMax's first model to deeply participate in its own evolution. 230B MoE, 10B active, 200K context. Built for agentic workflows, software engineering, and office tasks. Specialty: Agentic coding and productivity. Qwen3.6 Plus — Alibaba's hosted flagship with drastically enhanced agentic coding capabilities. Supports 201 languages and dialects, making it the best choice for multilingual users. Specialty: Multilingual agentic work. Fast (⚡) — Maximum speed for everyday tasks These models prioritize response time and efficiency. Perfect for quick questions, summarization, brainstorming, and situations where you need immediate feedback. Step 3.7 Flash — StepFun's high-frequency MoE built for production agents. 198B parameters, ~11B active per token, 256K context, up to 400 tokens/s. Specialty: Fast agentic work and coding. Gemma 4 26B A4B — Google's efficient open MoE bringing multimodal reasoning to consumer hardware. 26B parameters with ~4B active per token, 256K context, handles text, images, and documents across 140+ languages. Specialty: Very fast and reliable general chat. GPT-5.4 Nano — OpenAI's smallest workhorse optimized for high-volume coding, tool use, and sub-agent fleets. 400K-token context with 128K max output. Specialty: General use at high speed. GPT-5-Nano — OpenAI's fastest and cheapest GPT-5 variant. Ultra-low latency with a 400K token context window. Specialty: Rapid interactions and developer tools. Deepseek V4 Flash — DeepSeek's hosted speed flagship with enhanced agentic coding and strong multilingual capability. Specialty: Fast math and coding. Mistral Small 4 — Mistral's newest small model with amazing text generation speed and significant improvements over previous versions. Specialty: Fast general chat. Qwen3 Coder 30B — Alibaba's most agentic code model using the efficient 30B-A3B variant. Specialized for code generation, completion, and multi-turn tool calling. Specialty: Fast coding. Budget & Experimental Amazon Nova Micro — Amazon's text-only model optimized for cost and speed. 128K token context, excellent for text summarization, translation, interactive chat, and basic coding. Specialty: Very fast speed on a budget. Mistral Small 3.2 — For experimental use only. Not recommended for production tasks. How Vector Works The Conversation Loop Vector operates on a unique perpetual loop architecture: You speak or type your request Vector builds context — Current date, weather, your name, memory, screen content, clipboard, and its evolving soul file The AI decides — Using function calling, Vector determines which tools to use (if any) Tools execute — Vector sends messages, searches the web, sets timers, generates images, etc. Results return — The AI sees what happened and decides next steps Response delivery — You get a natural language answer, and the loop continues Vector can call multiple tools simultaneously for efficiency and always sends a status notification when working on longer tasks so you know it hasn't frozen. Input Handling Vector intelligently processes whatever you share: - Images & Photos — Converts to base64 and sends to a vision-capable model for rich description - Safari Web Pages — Extracts URL and page content for analysis - URLs — Fetches and reads the full page content - PDFs — Extracts text for summarization and Q&A - Rich Text — Converts to HTML for processing - Files — Attempts text extraction for readable formats Memory Management Vector keeps your last 12 messages in active context. When conversations grow longer, it intelligently trims older tool messages while preserving the system prompt and recent exchanges. This ensures Vector stays responsive while maintaining conversation coherence. Privacy & Security What Stays On Your Device Vector is designed with privacy as a core principle: All config files stored in /Vector-AI/ folder in the Files app All artifacts stored in /Vector-Artifacts/ folder No analytics, telemetry, or tracking of any kind No crash reporting or data collection by the creators Version check only contacts RoutineHub's public API to compare version strings — no personal data included What Gets Sent Externally AI messages — Your message plus system context (date, clipboard, screen text, weather, memory) is sent to Pollinations AI for response generation Web searches — Queries sent to the MIKLIUM API Code execution — Python code sent to the MIKLIUM API sandbox Image generation — Prompts sent to Pollinations image API You control your data: - Clear memory anytime via Vector Config - Reset all data by deleting the /Vector-AI/ folder - Toggle Privacy Mode to disable screen and clipboard sharing - Use your own Pollinations API key instead of the default Vector Config Companion Vector Config is the settings hub for your AI agent. Access it by telling Vector "run the Vector Config shortcut" or launch it directly. Available Settings 🗣️ Voice/Text Mode - Switch between typing conversations and voice-based interaction - Voice mode uses iOS dictation for input and text-to-speech for responses 🤖 Models - Browse all 15+ available models with detailed descriptions - See capability badges (Smart, Balanced, Fast) at a glance - Change your default model instantly - Model descriptions include parameter counts, context windows, and specialty use cases 🧠 Memory - View what Vector currently remembers about you - Delete memory to fix bugs or start fresh - Note: Short-term memory stores one primary memory at a time for efficiency ✍️ Customization - Toggle Customization Mode on/off - When on, Vector rewrites its soul file every round based on your interactions - When off, Vector uses a static personality - Just ask Vector to "be more friendly," "be more concise," etc. and it adapts 💻 Artifacts - View all saved code files and documents - Display HTML files fully rendered in a web view - Delete individual artifacts - Publish HTML artifacts to the web (submitted via email for hosting) 📇 Name - View and update what Vector calls you - Personalizes greetings and conversational depth 🔒 Privacy Mode - On: Vector does NOT read your screen or clipboard - Off: Vector CAN see your screen and clipboard for contextual help 🌸 Connect Pollinations - Link your own Pollinations account for additional features - Import your API key from compatible sources ⏬ Import Data - Export a prompt from your current AI provider - Paste the response to generate a personalized soul.txt file - Transfer your AI relationship from ChatGPT, Claude, or other services 🎬 Start Vector - Launch the main Vector shortcut directly from Config 🗑️ Delete All Data - Reset Vector to factory settings - Deletes memory, name, and config files - Preserves your artifacts folder

A system prompt was added to support web rendering