Knodge 2.0.7 - What has changed for you since 2.0.5

By Dein Knodge-Team

Knodge 2.0.7 - What has changed for you since 2.0.5

Short version: faster chat, cleaner mobile operation, more control over blog publishing, more robust background operation. Everything you notice in everyday life in one note.

Chat & AI

  • Voice mode in chat Hold down the mic and dictate continuously, the AI can also read out answers. (#422, #428)
  • Live dictation in 4-second chunks ** You see written text almost in real time instead of a big block at the end. (#428)
  • Faster standard model ** New chats now run on Gemini 3.1 Flash Lite - noticeably faster with the same quality. (#469, #457)
  • Tool calls (web search, database, images) appear as a small step list above the answer instead of cluttering up the chat. (#466)
  • Input remains fluid in long threads ** No more typing stuttering even after 200+ messages. (#493)
  • Directory of sources can be expanded. Used documents are grouped and expanded on request. (#439, #443)
  • Tables easier to read ** Clearer background, higher contrast - even on mobile. (#438)
  • **"General knowledge active" badge ** As soon as the AI goes beyond your documents, you see it immediately. One click deactivates it again. (#430)
  • Answer watchdog If an answer gets stuck, you get a "Reload" button instead of endless spinners. (#487)

Workspace & Mobile

  • Unified pill input. Desktop and mobile use the same chat pill: plus menu on the left, mic and send on the right. (#434, #426)
  • Plus menu always there ** Even if your hub has no special function activated, you can reliably find "New chat" in the Plus menu. (#436)
  • Slim mobile header On the smartphone, the workspace header is permanently 32 px high, the chat no longer has a header - more space for content. (#427, #409)
  • **Left icon bar ** On mobile, the left sidebar remains visible as a narrow icon column, you can collapse it if you need the whole screen. (#410, #464)
  • Edge swipe ** Swiping inwards from the edge of the screen opens the respective sidebar as an overlay. (#399)
  • Fullscreen for chat & documents Clicking on the maximize icon hides the rest, closing the document automatically brings you back. (#418)
  • Check off notes directly Status checkmark (TODO/OK) is now next to the file name, one click is enough. (#401)
  • "All publications" in the sidebar. Direct link to the blog overview from every workspace. (#483)

Blog & SEO

  • Translate and publish per language.** You decide separately which languages of a blog story go live - no more unwanted auto-translations. (#472, #475)
  • **Drafts with login paywall ** Non-published articles end up in the sitemap, but are only readable after login - good for SEO, safe for embargo. (#450, #452)
  • Draft teasers on the public overview ** Visitors can see that more is coming without spoiling content. (#456)
  • Managers see draft badges ** You can see at a glance which posts are not yet public. (#451)
  • Faster list page The blog overview is rendered on the server side - page is there immediately when called up, even for crawlers. (#455)
  • Soft-404 gone. Portal-Home points correctly to /blog, checkout pages deliver real body instead of empty shell. (#476, #477)
  • Private blog articles visible again. Members can also see internal posts reliably on mobile. (#470, #471)

Documents & Editors

  • Editor sharing per document. Managers explicitly share documents with editors via Lock/Users toggle - default remains private. (#416)
  • Publish own notes Editors may "provide" and rename their _user_* PDFs and CSVs themselves. (#444)
  • File names become clickable ** Knodge recognizes naked file names in chat sources and automatically links them to the document. (#443)
  • New industry presets. Coaching and marketing incl. own icons; "Science" now translated correctly. (#445)

Stability & Performance

  • Login survives network interruptions. Auth token is hardened against short offline phases and deployments - no more "Please log in again" after a WLAN change. (#437)
  • Cleanup job forgives DB stalls Nightly cleanups no longer abort on a single connection error. (post-#487)
  • React query persists Data you just saw is immediately available again after reload - the rest is reloaded in the background. (post-#487)
  • Hover-Prefetch. As soon as you hover over a link, the target page is prepared. (post-#487)
  • **Cache hit banner shorter ** The "As of HH:MM" display now disappears after 4 instead of 10 seconds. (#496)
  • Routing-Bugfix Workspace. Old document links are automatically redirected to the correct workspace and the document is opened. (post-#483)

In planning

  • Integrate Claude models In addition to Gemini, you should also be able to select Anthropic Claude models as the default or fallback chat model in your hub.
  • Optimize chat history ** Long threads are summarized more intelligently instead of just being cut off - relevant passages remain in context, token costs are reduced.
  • Dynamic top_k in RAG ** Instead of a fixed 25 hits, the AI decides for itself how many document passages it pulls depending on the question - short factual questions pull less, complex analyses pull more.

The order and exact form are still open. If you would like something specific for one of the points, please let us know.

Under the hood

We have changed the default chat model to Gemini 3.1 Flash Lite Preview, raised the RAG default to top_k = 25 (more precise hit lists) and rebuilt the Stripe webhook so that subscription renewals are activated idempotently via checkout.session.completed, customer.subscription.* and invoice.paid. Server-side rendering for blog lists and the hardened auth token together ensure that the platform feels the same as on the desktop, even with a shaky connection and on mobile devices.

Version 2.0.7 - May 2026

Feedback, wishes or discovered a bug? Write to us directly in the chat or by email to support@knodge.eu - we read every message.