Changelog

Version 1.0.94 November 2026

Independent canvases and workflows, with more reliable chat recovery

Connect content and creative workflows with nodes, reuse references and generation history, and get improvements to recovery, drafts, model selection, image downloads, reading, navigation, and subscription synchronization.

Independent canvases: Organize your work with text, file, and response nodes, with connections, groups, zoom controls, multimedia previews, and full-screen Markdown editing.
Canvas workflows: Create draft rules from natural language, review and enable them, then run them manually or after a selected generation completes. Follow step progress, inspect inputs and outputs, and retry failed steps.
Reuse references and history: Use @ to reference nodes and Library files, import favorites, and restore results together with their original inputs from generation history.
Generation recovery: Continue chat and canvas responses that reach the output limit, or retry with a lower limit when the available balance cannot cover the request.
Draft and setting protection: Better preserve drafts after failed or cancelled sends and when refreshing continued conversations, protect new edits made during recovery, and fix thinking settings during sign-in recovery.
More reliable model selection: Fixed incorrectly disabled models on shared pages, false conflicts after model list changes, and automatic routing errors, with better video parameters and input compatibility checks.
Image preview and download fixes: Generated image formats, download extensions, and file information now match the actual image content.
Reading and navigation improvements: Improved locating and expanding canvas lists and scrolling the sidebar, fixed overlapping chat outlines, and preserved your reading position when the window is resized.
Subscription synchronization fixes: Improved recovery after payment processing errors, reducing delays in subscription activation and renewal status updates.

Version 1.0.813 September 2026

Smoother Webpage, more reliable sharing and subscriptions

This update completes the multimodal model catalog, improves shared conversation continuation, Webpage, and generation stopping, and refines subscription information and plan browsing.

More complete multimodal model selection: Fixed missing image, audio, and video models in the available list, improved image generation capability detection, and reduced incorrect cleanup of model preferences.
More reliable sharing and continuation: Fixed blank shared pages when signed in and missing history when continuing a shared conversation, with better recovery and retries after refreshing.
Smoother Webpage: Fixed some generated content not being recognized as Webpage cards, reduced stalls when closing the sidebar, and improved long code expansion, syntax highlighting, and text selection preservation.
More reliable quick stops: Improved cancellation just after sending or while queued, without incorrectly marking older replies; refreshing after cancellation no longer reports that cancellation as a generation failure.
Completion indicators stay acknowledged: Opening a conversation acknowledges its current generation completion indicator and syncs this across tabs in the same browser, while preserving new indicators for later completions.
More accurate subscriptions and billing: Improved benefit and renewal status synchronization after plan changes, preserved valid benefits until scheduled changes take effect, corrected historical subscription amounts, and clarified scheduled plan changes versus renewal cancellation.
Plan browsing improvements: Improved long button labels on mobile and aligned the styling of plan comparison controls.

Version 1.0.75 September 2026

Updated plans, more flexible retries, and more reliable conversations

This update revises membership plans, improves API Key setup guidance, adds thinking effort selection when retrying, and makes everyday chats more reliable.

Updated membership plans: Free, Starter, and Plus are now the three plan options, with clearer benefit descriptions and easier plan browsing on mobile.
Clearer setup guidance: First-time users get help connecting and obtaining an OpenRouter API Key. AI requests use your own Key, and OpenRouter bills model usage separately.
Choose thinking effort when retrying: Keep one-click retries with the current model, or select a thinking effort supported by the model you want to retry with.
More reliable model selection: Available models stay consistent across chat, Echo, and settings, while invalid model preferences are cleaned up automatically.
Prevent empty conversations: Model availability is checked before the first message is sent, so a failed send does not leave an empty conversation behind.
Generation stopping and state synchronization fixes: Fixed responses still appearing to generate after stopping, incorrect states after restoring conversations, and stop states failing to sync across tabs.
Billing access fix: Users who previously paid and are now on the free plan can access billing management again.

Version 1.0.624 August 2026

More reliable sharing, generation, and media playback

This update improves shared reading and Webpage previews, makes long-running generation easier to stop, and strengthens everyday chat, account, and media reliability.

Anonymous shared conversations now stay focused on read-only viewing, while keeping jump-to-bottom and Webpage preview access; the table of contents steps aside while Webpage is open.
Webpage previews can now load declared, fixed-version browser libraries and styles more reliably in both regular and shared conversations.
Stopping Webpage generation or a long thinking response now finishes more reliably, preserves partial content, and reduces stalls while thinking is displayed.
When a media request cannot proceed, the conversation now receives a localized title instead of remaining untitled.
Highlights and annotations now update immediately after a change, keep successful feedback unobtrusive, and show more specific guidance when something fails.
Account, plan, and connection information now loads more reliably on the first visit, avoiding occasional blank or unavailable states.
Online audio and video continue playing after a temporary link expires, with seeking and partial playback handled more reliably.

Version 1.0.520 August 2026

Smarter thinking controls and more dependable Webpage previews

This update tailors thinking and response capacity to the selected model, while making Webpage creation, recovery, sharing, and previewing more reliable.

Thinking controls now match each model's capabilities, offering supported effort levels, Auto, Off, or a custom budget where available.
Chat and Webpage responses now use the selected model's available capacity more accurately, helping longer answers and complete Webpage files finish successfully.
Webpage cards now recover more reliably from interrupted streams and supported formatting variations, including complete HTML responses.
Shared conversations can now open HTML Webpage content in a safe, read-only preview without granting editing access.
Webpage previews can now load commonly used external libraries more reliably, with clearer failure feedback and a retry option when content cannot be displayed.
Copying Webpage content now confirms success or explains when it fails, and shared-conversation notices and cookie preference controls received small interface refinements.

Version 1.0.416 August 2026

More reliable chats, attachments, and formulas

This update improves the reliability of starting new chats, handling attachments, and displaying messages, while fixing issues that could cause duplicate generation or incorrectly rendered content.

New chats now enter the generation flow faster and appear in the sidebar immediately.
Fixes an issue where refreshing the page during generation could resubmit a message and produce duplicate replies.
Text and Markdown attachments no longer remain stuck uploading after a failure, and now provide clear failure and retry feedback.
When the current model cannot read an attached text file, Yorikai now tells you before sending instead of continuing without the attachment content.
Fixes the display of absolute values, norms, and other mathematical formulas containing vertical bars, including formulas inside tables.
Waiting dots now disappear promptly once the model starts thinking, writing, or generating media.

Version 1.0.314 August 2026

More complete, reliable references

When you reference text or media from a chat, the original content now stays available to the current conversation, retries, and follow-up messages.

Referenced text, images, videos, audio, and files now reach the model as their original content, rather than only a filename or reference card.
Retrying a reply keeps its original references without sending your message twice.
Follow-up messages can continue using previously referenced content, even after refreshing the page.
Video references can now be previewed, and models that cannot read video are clearly disabled when the conversation needs one.

Version 1.0.212 August 2026

Clearer membership and model access

This release fixes issues around membership states, model access, and settings navigation, while refining Echo presentation so key actions are easier to understand and complete.

Fixes an issue that kept Free members from selecting some higher-cost models when using their own API key.
Model restriction notices now link directly to plan upgrades or API key settings.
Pricing cards now show the appropriate get, upgrade, downgrade, or billing action for the current membership.
Refines Echo template bubbles and shadows so content is no longer obscured.
Official Echo author names now follow the interface language.

Version 1.0.0Kongren Edition9 August 2026

Yorikai’s first official release

Yorikai 1.0 marks the beginning of a unified, natural, and focused conversation experience, bringing multi-model chat, files, reusable knowledge, and assistant presets into one coherent workflow.

Use currently available AI models from one familiar conversation interface.
Switch models while keeping the conversation flow consistent, making exploration and comparison feel natural.
Upload files and reuse Library content so important information can take part in the current conversation.
Use Web Search to bring current information and source citations into supported conversations.
Use Echoes to save reusable assistant settings, instructions, and starter prompts.
Save messages and image, audio, and video files to Favorites for quick access later.
Generate images, audio, and video directly in Yorikai with supported models.
Purchase paid plans through Waffo’s hosted checkout, with payment details handled by Waffo.
Sync chats and Notebook content from Yorikai to Google Drive to keep a separate, up-to-date backup.