The biggest update since Knowledge Bases and Standing Agents — a new agent that acts, not just answers. NEW FEATURES • Autonomous Task Agents — Give an agent a goal instead of a fixed prompt, and it plans and executes its own sequence of tool calls to get there, re-planning after every result. Anything that writes data or has a real-world effect pauses for your explicit approval first, so you always see exactly what it's about to do before it happens. Generated images and spend are shown right in the run and tracked in Usage & Costs. Find it right next to Standing Agents. • Search Everything — A new search across chats, messages, agents, memories, and Knowledge Base documents at once. Opening a message result jumps straight to it in the conversation. • Standing Agents now tell you why — Agents show a "next run ≈" estimate, and if a scheduled run fails, you'll see why instead of just a stale "last ran" time. IMPROVEMENTS & FIXES • Hardened the streaming-response path against a rare crash reported by TestFlight testers. • Fixed a data race in model-list fetching that could occur when two screens looked up models at the same time. • Fixed Ollama tool-calling so multi-tool-call turns correctly match each result to the call it answers. • Fixed Remove Object and Colorize to no longer crash on very large photos, and no longer return a smaller image than the one you started with. • Fixed a download-cancel race that could leave a model download running after Cancel was tapped. • Fixed a Knowledge Base document counter that could drift after a failed re-import. • A broad pass over the app's typography and spacing for more visual consistency across every screen. Thanks for using Team AI! As always, we'd love to hear your feedback.

Memory improvements: - Fixed memory not saving new facts when chatting with on-device Apple Intelligence. - Memory Browser and Memory Map now correctly reflect fresh activity when you return to an older chat and keep talking. - New: choose a custom OpenAI-compatible embedding provider (e.g. Ollama) in Settings > Memory, with automatic model discovery. Image generation: - New: generate images using your own local Ollama connection, right inside Image Studio and chat. Fixes: - Copy and paste (press-and-hold) now works reliably across every text field, including Settings. - Standing Agents moved above New Group Discussion on the Agents tab for quicker access. - Fixed a data-loading issue affecting some builds. - Fixed a crash that could occur when generating certain gradient or abstract images in Image Studio. - Fixed a crash that could occur when generating a group discussion summary report. - Fixed an issue where choosing a custom embedding provider while creating a new Knowledge Base did not reliably save. - Fixed a rare issue on iPad where starting a new chat could show a blank screen instead of the chat view. - Fixed a rare data migration issue that could affect saved memories when upgrading from an older version of the app.

What's New in Version 1.4.0 Our biggest update yet — a whole new set of on-device, privacy-first tools that make Team AI far more capable, with no subscription and nothing to sign up for. NEW FEATURES Knowledge Bases — Chat with your own documents. Add PDFs, text, and Markdown files to a Knowledge Base, then attach it to any chat from its options. Everything is embedded and searched entirely on-device, so your documents never leave your iPhone. Standing Agents — Put an agent on a schedule. Set one up to run on its own (like a daily briefing or a weekly check-in) and have its latest results waiting for you. Roundtable Listen — Hear a group discussion read aloud, with a distinct voice for each agent, so you can follow a multi-agent debate hands-free. Weekly Memory Digest — A private "What I've learned about you" recap, distilled on-device and refreshed about once a week. One-Tap Boost — Not happy with an answer? Boost it, and the model will critique and refine its own response in a single tap. AI Judge for Model Comparison — Compare two models side by side and let an AI judge score which did better. MCP Server Directory — Browse and add Model Context Protocol servers from a built-in, one-tap directory. IMPROVEMENTS & FIXES Decision Records — Key decisions from group discussions are saved as durable memories. Image Recipes — Apply chained image-editing steps to any image in one tap. Monthly Spend Budget — Set a budget and track usage in Usage & Costs. Sharper on-device memory search using Apple's contextual embeddings. A refreshed, more professional look throughout. Numerous bug fixes and stability improvements. Thanks for using Team AI! As always, we'd love to hear your feedback.

What's New in Version 1.3.4 NEW FEATURES • True Image-to-Image Generation — Attach an existing image when generating in Image Studio or right from a chat, and use it (with an adjustable strength slider) to guide the result instead of starting from a blank prompt. A new "Generate Variation" action on any image in your history opens the composer with that image pre-attached. • iCloud Sync — Your chats, agents, and memories now sync privately across your devices via iCloud. • More On-Device Models — Added Qwen 2.5 Coder 1.5B, a lightweight coding model that runs on older iPhones; xLAM 2 8B, a dedicated function-calling/tool-use model; and a Neural-Engine-optimized variant of Stable Diffusion 1.5 for faster on-device image generation. IMPROVEMENTS & FIXES • Haptic feedback added throughout Image Studio, agent editing, memory management, and Settings, so saves, deletes, and generations all feel more responsive. • Redesigned Usage & Costs with a clearer summary card, per-provider and per-model breakdowns, and more accurate cost estimates. • Fixed a bug where Siri Shortcuts could resolve the wrong provider or model. • Fixed a rare issue that could prevent the app from opening on first launch after a device reboot. • Fixed Image Studio's edit menu overflowing on some screens; added delete-all and cancel-generation actions. Thanks for using Team AI! As always, we'd love to hear your feedback.

What's New in Version 1.3.3 NEW FEATURES • Image Studio — A dedicated tab for generating images, right between Chats and Agents. Pick a provider — on-device Stable Diffusion, Gemini, OpenAI, or your own local Ollama/LiteLLM server — choose a model, describe what you want, and generate. Every image you create is saved privately on your device and shown as browsable history, with save-to-Photos, share, and delete built in. • Generate Images From Any Chat — A new sparkles button next to the message input opens the same provider/model picker without leaving your conversation, so a generated image attaches directly to what you're saying — including from your own local Ollama/LiteLLM server. • Local Ollama & LiteLLM Image Generation — Point Image Studio at your own local Ollama or LiteLLM server for image generation, with the same negative prompt, steps, and seed controls as our other advanced providers. • On-Device Image Generation — Download real Core ML Stable Diffusion models (Stable Diffusion 2.1, Stable Diffusion XL, RealVisXL V5.0, and the single-step SD-Turbo) in Settings → Local Image Models and generate images with zero internet connection and no API key. • Five New On-Device Photo Tools — Upscale, Colorize, Remove Object, Restore Faces, and Deblur are now available on any image in Image Studio's history, whether you generated it or imported it from your Photo Library. All five run entirely on your device using a new on-device machine learning engine — nothing is uploaded anywhere. • Import From Photos — Bring any photo from your library straight into Image Studio to apply the new photo tools, or just build up your history alongside your generated images. • Smarter Local Model Compatibility — Settings → Local Models and → Local Image Models now show whether each model actually fits your specific device's memory, instead of a generic device-name guess. Both catalogs are also grouped by category so it's easy to scan. • More Local Models — Added SmolLM3 3B, Qwen 3.5 4B, Qwen 3.5 9B, and Gemma 4 12B to the on-device text model catalog. IMPROVEMENTS & FIXES • Fixed a bug where chat and group chat sessions using a local Ollama/LiteLLM server could silently pick the wrong model instead of the one you configured. • Fixed a bug where branching a group discussion into a new conversation could lose your custom provider settings. • Fixed a bug where stopping a response mid-generation could discard what had already been written instead of keeping it. • Image generation and photo tools now run smoothly in the background instead of momentarily freezing the app. • Moved "New Chat" to the top-right of the Chats tab, with a more compact inline header. • Fixed a bug where saving an image to Photos could silently fail to prompt for permission. • Updated the in-app App Guide and Tips & Features screens to cover Image Studio, on-device image generation, and the new photo tools. Thanks for using Team AI! As always, we'd love to hear your feedback.

NEW FEATURES Voice Mode — Let your AI agents read their responses aloud, hands-free. Great for multitasking or when you'd rather listen than read. Share Your Agents — Export any custom agent you've built and share it with friends, or import agents others send you. Perfect for sharing your favorite prompt setups. Live Activities — Follow group discussions right from your Lock Screen and Dynamic Island, so you always know who's speaking and how the conversation is progressing without opening the app. Faster Workflows — Multi-step workflows now run independent steps in parallel instead of one at a time, so complex pipelines finish faster. IMPROVEMENTS & FIXES Fixed an issue that could cause the app to crash or show incorrect messages during active group discussions. Fixed a bug where regenerating a response could occasionally leave stray messages behind. Improved reliability when connecting to custom tool servers.

Welcome to a massive update! We've added powerful new intelligence features, expanded our local model catalog, and squashed bugs to make Team AI smoother than ever. AI MEMORY SYSTEM Your AI now remembers. Team AI uses on-device vector embeddings to automatically extract and store key facts from every conversation with smart deduplication. Memory Browser: Search, pin, edit, and delete individual memories. Memory Timeline: A visual graph showing how your AI's knowledge grows over time. Per-Agent Scoping: Each agent maintains its own memory alongside shared cross-thread knowledge. Long-press any message and tap "Remember This" to save it manually. Everything stays 100% on-device and private. CROSS-CHAT KNOWLEDGE When you start a new chat, related context from past conversations is automatically surfaced. Tap "Include Context" to inject prior knowledge into the current thread. MODEL COMPARISON Compare two AI models head-to-head in real time. Choose any two providers, type a prompt, and see both responses side-by-side with speed and performance metrics. Available models now auto-load from each provider. Pick a winner when you're done. DISCUSSION PLAYBOOKS Group chats just got strategic. Choose from 6 structured discussion formats: Devil's Advocate: One agent relentlessly challenges every idea. Six Thinking Hats: de Bono's parallel thinking method. Red Team / Blue Team: Agents split into attackers vs. defenders. SWOT Analysis: Four agents each own Strengths, Weaknesses, Opportunities, or Threats. Socratic Dialogue: One agent only asks probing questions. Free Discussion: Open-ended conversation. CONSENSUS TRACKING AND VOTING Enable real-time consensus tracking during discussions. Turn on voting to let agents cast Agree/Disagree/Neutral at the end and see which ideas won majority support. CONVERSATION BRANCHING Long-press any message and tap "Branch from Here" to fork into an alternate timeline. Explore different directions without losing your original thread. CODE EXECUTION SANDBOX AI can now write and execute code directly in your chat. Results render inline with syntax highlighting. PROMPT LIBRARY Save your best prompts for reuse. Long-press any message to save it, then quickly insert saved prompts into any future conversation. WORKFLOW PIPELINES Build sequential AI pipelines where each step's output feeds into the next. Define custom steps like "Data Extraction, Analysis, Summary" and run them automatically. LOCAL MODEL EXPANSION (NOW 30 MODELS) We've added heavier, desktop-class models (up to 8GB) optimized for high-end devices like iPad Pro M4 and iPhone 15/16 Pro Max: Gemma 4 (1B and 4B) from Google. Llama 3.1 8B and Hermes 3 for reasoning and roleplay. Mistral Nemo 12B and Qwen 2.5 14B pushing mobile RAM limits. DeepSeek Coder V2 Lite (16B MoE) and CodeQwen 1.5 for programming. Plus more from Microsoft (Phi-3) and 01.AI (Yi 1.5). INSTANT PRESET AGENTS Your Agent Library now auto-populates on first launch with 16 specialized AI personas including Startup Founders, a Code Review Panel, and a Debate Club. BUG FIXES AND IMPROVEMENTS Real-time Storage UI: Deleting a local model now instantly updates storage statistics. Cost Tracking: Fixed local models accidentally showing estimated costs. Local inference is truly free. Feature Tips: The tips guide now opens properly from the toolbar. Workflow UX: New workflows auto-present configuration instead of a blank screen. Model Comparison UX: Configuration panel shows upfront on first launch. Model Picker: All model selectors now auto-fetch available models from your provider. Empty States: All new feature screens show helpful onboarding instructions. Thanks for using Team AI! If you enjoy it, please leave us a review.

Welcome to our biggest update yet! We’re thrilled to introduce a suite of powerful, enterprise-grade features that turn Team AI into your ultimate personal intelligence hub: Autonomous Agent Workflows Chain multiple AI prompts together into automated, sequential pipelines. Set up complex tasks and watch the AI seamlessly hand off outputs from one agent to the next without requiring you to intervene. Enterprise-Grade Code Execution Sandbox Our new secure JavaScript Sandbox allows your AI to write, test, and execute scripts dynamically right on your device. From complex math to data processing, the AI now truly has a mind of its own. Generative UI Components Experience chat like never before. The AI can now dynamically generate interactive UI elements and mini-applications directly inside your conversation thread to visualize data and give you richer answers. Local On-Device AI Processing We now fully support local, on-device Large Language Models (LLMs). Experience ultimate privacy and lightning-fast speeds with zero internet required when using your favorite local models. Siri Shortcuts Integration "Hey Siri, ask Team AI..." — You can now trigger powerful AI prompts, chat threads, and agent workflows using nothing but your voice. Perfect for hands-free assistance on the go. Other Improvements Added a comprehensive Help & Guide section in Settings to help you master all the new features. Various performance optimizations and under-the-hood enterprise security enhancements.

What's New in Team AI 1.2: • 14 On-Device AI Models Run the latest models entirely on your iPhone — Llama 4, Qwen 3, Gemma 3, Phi-4 Mini, SmolLM2, Mistral Small 3, DeepSeek R1, and Qwen 2.5 Coder. All offline and private. • Multi-Agent Group Chat Launch team discussions with one-tap presets — Startup Founders, Code Review Panel, Creative Writing Room, Debate Club, and Product Team. Agents debate, collaborate, and build on each other's ideas in real time. • Consensus Voting & Auto Summary Enable voting rounds where agents cast structured votes. Toggle auto-summary for a formatted recap when the discussion ends. • 7 Built-in Tools Calculator, Weather, Unit Converter, Timer, Text Summarizer, Image Description, and Image Generation — all on-device, no API keys needed. • Web Search & URL Fetch Agents can search the web and fetch page content autonomously when tools are enabled. • Rich Markdown Rendering Code blocks with syntax labels and one-tap copy. Bold, italic, links, and lists render beautifully in conversations. • Voice Input Tap the mic to dictate messages hands-free. Uses on-device speech recognition for privacy. • Suggested Replies Get 3 smart follow-up questions after each response. Tap to continue the conversation instantly. • Streaming Speed Indicator See real-time tokens-per-second while AI responses are generating. • Bookmarks & Regeneration Long-press any message to bookmark, copy, share, or regenerate a fresh response. • Search & Organization Search across all conversations. Pin important chats to the top. Swipe to delete. • Personas Start chats with built-in personas — Code Expert, Creative Writer, Concise Responder, Translator — or create your own. • Export Conversations Export any chat as Markdown or plain text to share or archive. • Usage & Cost Tracking Monitor token usage and costs across all conversations from Settings. • Stability Smart context windowing keeps the app fast and stable even in very long conversations. Auto-titles name your chats after the first exchange.

Stability & Performance • Resolved a critical crash that occurred after several messages in chat • Completely redesigned the streaming engine for dramatically smoother responses • Reduced main thread load during AI generation by up to 80% Chat Experience • Messages now always appear in correct chronological order • Added a "Thinking..." indicator while the AI prepares its response • Improved auto-scroll behaviour during long conversations • Added quick navigation buttons to jump to the top or bottom of any chat Group Discussions • Same stability improvements applied to multi-agent group conversations • Smoother streaming for all agents and orchestrator summaries Settings • Fixed colour and interaction styling in Custom Provider and MCP settings