DeepAsk Changelog
More details with pictures, please check LifeOS for Obsidian
2.0.22
- Fixed coding-agent context so DeepAsk includes the active note, current web page, and open tab summary when sending Anthropic, OpenAI-compatible, Claude Code, and Codex requests
2.0.21
- Fixed embedded MCP periodic daily-note lookup to honor Obsidian Daily Notes core folder and format settings, including month subfolders such as
Daily/06/YYYY-MM-DD.md
2.0.20
- Fixed cross-plugin settings updates polluting DeepAsk's authentication context when multiple LifeOS plugins are enabled, preventing occasional activation-state mixups
2.0.19
- Bundled the DeepAsk desktop Koa server into the main plugin file as a desktop-only dynamic import, removing the last external runtime file
- Aligned AI Chat's secondary integrity check with the installed
manifest.json, preventing false failures while the plugin instance is still reloading after an update
2.0.18
- Bundled the DeepAsk chat panel back into the main plugin file so AI Chat no longer depends on a separate
ai-chat.jsruntime file - Made the desktop Koa server loader resolve
koa-server.jsfrom the installed plugin directory and self-download the asset when an older updater missed it
2.0.17
- Fixed DeepAsk update packages missing split runtime files, which could make AI Chat and the local Koa server fail to load
2.0.15
- Fixed Codex CLI replies stopping after the first partial segment in DeepAsk
2.0.14
- Updated Ordinary Model Tools settings so built-in skills and embedded MCP tools are always enabled, while external skills and external MCP tools are disabled by default
- Split write/update tool permission from delete tool permission for ordinary model chat
- Fixed the Ordinary Model Tools settings layout so switches align correctly and the max tool iterations input no longer overlaps its label
2.0.13
- Fixed DeepAsk failing to load when the bundle mixed React 19.2.4 with react-dom 19.2.7
2.0.12
- Fixed the latest DeepAsk release missing the Ordinary Model Tools settings section in AI Connect
- Added a note that Ordinary Model Tools only apply to DeepAsk ordinary model chat and do not affect Claude Code / Codex CLI mode
2.0.11
- Fixed the Obsidian third-party plugin list sometimes hiding the settings button by registering the settings tab before vault indexing starts
2.0.10
- Added skill and MCP tool calling for ordinary AI model chat: models can read vault/user skills, call embedded LifeOS MCP tools, and connect to external MCP servers on desktop
- Added ordinary model tool-call settings, with write/delete tools disabled by default to prevent accidental note edits
- Added the AI Connect settings tab for DeepAsk HTTP Server (MCP) configuration and one-click LifeOS Skill install / update
- Fixed the missing MCP / Skill settings tab in DeepAsk
- Fixed settings white-screen and Invalid hook call crashes caused by antd / React resolving to different runtime copies
2.0.9
- Fixed AI model dropdowns in Obsidian's detached settings window drifting to the main window or becoming hard to select
- Fixed other settings dropdowns and active tab highlight behavior
- Fixed duplicate input borders in settings fields
- Fixed possible settings white-screen after upgrade and plugin activation integrity failures
2.0.0
- Added activation device management: view device usage, refresh the device list, and unbind unused devices from plugin settings
- Added a beta manual update entry in Activation & Upgrade, making it easier to get the latest plugin when automatic updates fail
- Added Anthropic configuration support while keeping the existing OpenAI-compatible services available
- Added more flexible model service configuration, so the same provider can keep separate settings for different connection modes
- Added Claude Code CLI and Codex CLI local coding-agent modes, allowing Obsidian context to be sent directly to local coding agents
- Added OpenAI and Anthropic identity marks in the chat page, making the active model service clearer
- Added an animated pre-response placeholder, so sending a message immediately shows that a reply is being prepared
- Improved context loading for large vaults, especially when DeepAsk is used alongside other LifeOS plugins

