Public Release Timeline
Version Milestones
Release Timeline
Coming soon
Version 1.3.0
- Unread state and reminder dots are now part of the core navigation loop. Chat threads track unread replies more accurately, the Chats tab now reflects active unread conversations, and the Settings tab can surface pending app updates without scanning the whole app.
- App updates are now surfaced inside a clearer in-app flow. Settings can show a new-version badge, suppress reminder dots per released version after the user handles them, and open the App Store product sheet directly without kicking the user out into a separate app switch.
- Group collaboration is easier to read and less noisy. Valid plain-text mentions in agent replies now render more cleanly, deleted members no longer create false unread state, and collaboration context stays closer to the current thread instead of drifting toward stale carry-over.
- Agent editing and prompt boundaries are more resilient. Persona-specific draft prompts now survive step switching, user-facing names are normalized more defensively, and prompt metadata is sanitized closer to request time so saved content stays intact while model input stays cleaner.
- Reply presentation keeps tightening around real model behavior. Streaming output now gates half-finished collaboration or mention tokens, retry states recover more cleanly, failure bubbles read better, and per-model context budgets stay closer to each model's actual window.
Latest release
Version 1.2.0
- First-run onboarding and empty-state guidance became much more intentional. New users now move from profile setup straight into creating their first Agent, while Chats and Agents empty states react to whether the account already has Agents or conversations instead of repeating the same blank slate.
- Agent setup started behaving more like a real creation pipeline. Saving an Agent now runs a bootstrap step first, so identity summaries and first greetings land more consistently, and deleted-Agent states reconcile more cleanly with the thread list.
- Provider execution became far more robust than in the 1.1 line. Direct model calls gained a dedicated execution layer, clearer in-chat failure feedback for auth, quota, rate-limit, and upstream service problems, plus better attachment reuse instead of repeating unnecessary work.
- Daily chat behavior kept getting steadier. Group avatars returned to a shared layout system, streaming auto-scroll calmed down, preview text read more naturally, and thread state stayed closer to the real message lifecycle.
- Settings and presentation polish continued shipping alongside the core work. Avatar loading, sharing surfaces, storefront copy, and release materials all tightened up as the 1.2 line moved from preview into public release.
Earlier release
Version 1.1.0
- Model-aware multimodal handling landed in the product. Supported models can now take native image and file input, while unsupported ones fall back to on-device local grounding so attachment context still carries into the conversation instead of disappearing.
- Agent identity got more intentional. Built-in persona templates and deeper prompt and role configuration give users a clearer starting point for shaping distinct agents, while stronger identity continuity helps those agents feel more consistent over time.
- Shared group context and carry-over memory feel smarter. Group collaboration now leans harder into shared thread context, cleaner handoffs between agents, and stronger carry-over memory, so the same agent feels more natural across both one-on-one chats and group threads.
- Mentions, message handling, and on-device polish all improved. Mention replacement is less error-prone, group thread previews are cleaner, copied messages strip internal tokens, chat bubbles respond better to screen width, and long threads feel smoother to read, scroll, and revisit.
Launch milestone
Version 1.0.0
- Agent setup was already part of the launch product. The 1.0 line already included local agent creation, provider presets, identity fields, and prompt configuration, so Pokebot shipped with real configuration control from day one.
- Privacy-aware direct connections were there from day one. Pokebot launched without a developer-operated relay. Direct provider requests, local chat history, local session memory, and in-app privacy reminders were already part of the 1.0 experience.
- Multi-agent chat shipped as a first-class feature. One-on-one chats and group chats were both in the launch app, with thread-isolated history on the surface and shared agent session continuity behind the scenes so the same agent could already carry context across both.
- @mention routing, A2A collaboration, and Stop Thinking were all launch features. From 1.0 onward, users could route messages precisely inside group chat, let agents collaborate when needed, and still keep the whole reply flow visible and interruptible.