diff --git a/CHANGELOG.md b/CHANGELOG.md index 5c97b2d50d..8855a1710f 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -9,6 +9,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0 ### Added +- 💸 **Much cheaper tool chats.** Citation instructions, memory recall, the skills list and attached knowledge are no longer rewritten into the system prompt or your message on every tool round, so earlier parts of the conversation stay identical between requests and providers can serve them from their prompt cache, which makes chats with tools much cheaper. - đŸĻž **Coding agents as models.** On an Open Terminal that offers agents, coding agents such as Claude Code, Codex and Pi now appear in the model picker (one agent per conversation) and can be chatted with or called by voice, with their reasoning effort and service tier as options and their tool requests approved in the chat. - đŸšĻ **Auto tool permissions.** Administrators can now turn on an Auto tool permission mode (off by default) next to Full access and Ask for approval, where a review model decides which tool calls run automatically and which ask for approval first. - 🔗 **Personal terminals from native apps.** Native apps on your own computer can now sign in with renewable sessions (90 days without use by default, set with "SESSION_EXPIRES_IN"), appear as a personal terminal in chats and receive notifications even with no page open, with terminal output that resumes after a dropped connection. @@ -24,7 +25,6 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0 - âŒ¨ī¸ **Faster typing after large pastes.** Typing in the chat input after pasting 16,000 lines of text is now about 117 times faster, down from 8.6 seconds to 74 milliseconds per keystroke. [#33729](https://github.com/open-webui/open-webui/pull/33729), [#12087](https://github.com/open-webui/open-webui/issues/12087) - ⚡ **Faster terminal rendering.** The terminal tab now renders with WebGL where the browser supports it and takes in bursts of output in larger batches, so heavy output is handled faster. - âąī¸ **Faster large chats.** Large chats now open about twice as fast, and marking a chat as read or unread or removing a tag from it is 4 to 6 times faster. [#33665](https://github.com/open-webui/open-webui/pull/33665), [#33666](https://github.com/open-webui/open-webui/pull/33666) -- 📨 **Stable prompts for caching.** Citation instructions, memory recall, the skills list and attached knowledge are no longer rewritten into the system prompt or your message on every tool round, so earlier parts of the conversation stay identical between requests and provider prompt caching can reuse them. - đŸĒļ **Lighter reasoning streams.** Long reasoning replies from providers such as OpenRouter now send far less data to the browser and to Redis while they stream. [#33698](https://github.com/open-webui/open-webui/pull/33698) - đŸŽī¸ **Faster chat access checks.** Sending messages, typing indicators, shared links, comments and file access no longer load the whole conversation just to check who may access it, which helps most on large chats. [#33667](https://github.com/open-webui/open-webui/pull/33667), [#33669](https://github.com/open-webui/open-webui/pull/33669) - 🚀 **Faster channels.** The channel list, the channel messages and the replied-to messages in a channel now load much faster, especially with many channels or many replies. [#33690](https://github.com/open-webui/open-webui/pull/33690), [#33691](https://github.com/open-webui/open-webui/pull/33691), [#33688](https://github.com/open-webui/open-webui/pull/33688)