Changelog

All notable changes to Gemini Scribe are documented here.

🎨 Gemini Scribe 4.11.0 - Design System, Plan Mode & Smarter Agent Runs v4.11.0

🎨 Design system overhaul — a refreshed visual identity: a signature Gemini accent (bold user message bubbles, a gradient send button, and a brand mark), a unified elevation/shadow scale, a gradient progress bar across all states, motion polish, and consistent icon sizing — all built on a new theme-adaptive design-token layer so the plugin adapts cleanly to light, dark, and custom themes (#1090, #1104, #1107, #1109, #1110, #1112)
📋 Plan Mode (opt-in) — a new toggle in the agent view (and a command) that has the agent lay out its plan before it starts acting, shown as a "Plan" pill; leave it off for the usual direct execution (#1046)
⚙️ Background execution by default for long-running tools — deep research and image generation now run in the background automatically, so the agent view stays responsive while they work and you can track them in the Background Tasks panel (#1085)
⏹️ Stop halts a streaming response immediately — the Stop button now cancels a mid-stream follow-up right away instead of waiting for it to finish, and follow-up requests stream live as they arrive (#1053, #1097)
🖥️ Ollama improvements — a single Ollama model field (replacing the per-task pickers) that persists independently, so switching Gemini → Ollama → Gemini no longer changes your Gemini chat model; automatic vision-capability detection so image-capable local models are recognized; and accurate token counts read from the model's own reporting (#1125, #1058, #1076, #1152)
🖼️ SVG attachments — SVG and SVGZ files are now rasterized to PNG before being sent to the model, so vector images work as attachments (#1082)
📱 Cleaner startup, especially on mobile — the plugin no longer triggers "attempted to load NodeJS package" notices when it loads, fixing startup toast spam and a latent mobile-compatibility gap (#1154)
🧠 Smarter long agent runs — context is now compacted during a long tool chain (not just before a turn), so extended agent runs stay within budget without dropping the work already in flight (#1074)
🔎 RAG polish — "Reindex All" is renamed "Rescan Vault" to match what it actually does, and the "show all files" affordance is now keyboard-accessible (#1056, #1159)
🐛 Fixes — Deep Research works again on @google/genai 2.x (#1151); agent tool logs no longer fold into the preceding reasoning block in session history (#1084); and error messages no longer mislabel non-network failures as connectivity problems (#1153)
🔧 Under the hood — the plugin's entire softened-lint backlog is now cleared and enforced at zero warnings and all 150 circular imports were eliminated, tidying the codebase and moving the Obsidian community-directory review toward clean (#1032)

4.11.0 is a large feature release. The headline is a full design-system overhaul: a new theme-adaptive design-token layer underpins a signature Gemini visual identity — bold user message bubbles, a gradient send button and brand mark, a unified three-step elevation/shadow scale, a gradient progress bar across all states, motion polish, and a consistent icon-size scale — so the interface feels more cohesive and adapts cleanly across light, dark, and custom themes. The agent gains an opt-in Plan Mode that lays out a plan before acting, background execution by default for long-running tools (deep research and image generation) so the agent view stays responsive, immediate Stop cancellation of a mid-stream streaming response, and mid-loop context compaction that keeps long tool chains within budget without losing in-flight work. Ollama support matures with a single model field that persists independently of your Gemini model (fixing a provider switch that previously reset your chat model), automatic vision-capability detection, and accurate token counts from the model's own reporting. SVG and SVGZ files can now be attached (rasterized to PNG first). Startup is cleaner — the "attempted to load NodeJS package" notices are gone, which also closes a latent mobile-compatibility gap. Fixes include Deep Research on @google/genai 2.x, session-history reasoning/tool-log separation, clearer error messages, and RAG affordance and keyboard improvements. Under the hood, the entire softened-lint backlog is cleared and enforced at zero warnings and all 150 circular imports were eliminated via a leaf plugin interface, moving the Obsidian community-directory review toward clean. Dependencies are unchanged (@google/genai 2.10.0).

🛠️ Gemini Scribe 4.10.2 - Interactions API transport fix & recoverable deletes v4.10.2

🛠️ Hotfix — 4.10.2 fixes two issues on top of the full 4.10 feature set (listed below):
🧪 Interactions API transport fixed — with "Use Interactions API" enabled, every request failed with a CORS error ("Failed to fetch") after the Gemini SDK update to @google/genai 2.10.0, which broke chat and summarization for anyone who had the opt-in transport on. It now routes through Obsidian's requestUrl again, so the Interactions path (streaming text/reasoning, tool calls, and grounded sources) works (#1045)
🗑️ Safer file deletion — when the agent deletes a file or folder it now follows your Obsidian "Deleted files" setting (move to system trash or the vault's .trash folder) instead of permanently removing it, so deletions are recoverable; includes minor correctness fixes (#1030)
🌍 Localized UI in 20 languages — the entire plugin interface (settings, modals, agent view, commands, notices) is now translated, auto-selected from your Obsidian language with graceful fallback to English
🧠 Model reasoning display — the agent shows the model's thinking inline in the tool activity block and persists it in session history, so you can follow how it reached an answer (#965)
🧪 Interactions API transport (experimental, opt-in) — a "Use Interactions API" setting routes Gemini chat through Google's newer Interactions API, with full streaming of text, reasoning, tool calls, and grounded sources. Off by default and runs statelessly (your conversation history stays on your device); turn it on under Settings → Agent config → API configuration, or leave it off to keep using the proven generateContent path
🗺️ Google Maps grounding tool — the agent can ground answers in Google Maps data for place and location questions (#964)
🧠 Per-use-case thinking depth — reasoning effort is tuned per task (completions think the least, agent chat the most) instead of one global setting, balancing latency and quality (#986)
⏳ Soft agent turn budget — long agent runs get a gentle reminder as they approach the turn limit, plus a one-shot extension, instead of stopping abruptly (#996)
📋 Copy buttons for tool calls — quickly copy tool-call parameters and results from the agent view (#960)
🔄 In-app model-list refresh — refresh the available Gemini model list without restarting (#908)
📦 More bundled skills — obsidian-markdown and json-canvas skills ship built in, with richer obsidian-bases knowledge (#963)
⬆️ Dependency refresh — @google/genai 2.10.0 and @allenhutchison/gemini-utils 1.1.0

4.10.2 is a hotfix for the 4.10 line with two fixes; there are no other functional changes. First, it restores the experimental Interactions API transport: after the Gemini SDK was updated to @google/genai 2.10.0 in 4.10.1, the SDK's internal client was restructured and the plugin's CORS workaround stopped taking effect, so every request made with "Use Interactions API" enabled failed in Obsidian's renderer with a "Failed to fetch" error — breaking chat and summarization for anyone who had turned the opt-in transport on. Requests are now routed through Obsidian's requestUrl again, so the Interactions path works as intended (streaming text and reasoning, tool calls, and grounded sources). The transport remains off by default. Second, file deletion performed by the agent now respects your Obsidian "Deleted files" setting — files are moved to the system trash or the vault's .trash folder rather than permanently deleted — so deletions are recoverable; this change also includes a few minor correctness fixes. Everything from the 4.10 feature set is unchanged and still applies: the first fully localized UI in 20 languages (settings, modals, the agent view, command names, and notices, selected automatically from your Obsidian language with a graceful fallback to English); inline model reasoning that is shown in the tool activity block and persisted in session history; the opt-in Interactions API transport (Settings → Agent config → API configuration); a Google Maps grounding tool; per-use-case reasoning depth; a soft agent turn budget with a one-shot extension; copy buttons for tool-call parameters and results; in-app refresh of the Gemini model list; and additional bundled skills (obsidian-markdown, json-canvas). Dependencies remain at @google/genai 2.10.0 and @allenhutchison/gemini-utils 1.1.0.

🌍 Gemini Scribe 4.10.1 - Multilingual UI, Reasoning Display & Interactions API v4.10.1

🔧 Notes-only patch — 4.10.1 completes the 4.10.0 release notes (which omitted several of the features below). No functional changes; this is the full 4.10 feature set:
🌍 Localized UI in 20 languages — the entire plugin interface (settings, modals, agent view, commands, notices) is now translated, auto-selected from your Obsidian language with graceful fallback to English
🧠 Model reasoning display — the agent now shows the model's thinking inline in the tool activity block and persists it in session history, so you can follow how it reached an answer (#965)
🧪 Interactions API transport (experimental, opt-in) — a new "Use Interactions API" setting routes Gemini chat through Google's newer Interactions API, with full streaming of text, reasoning, tool calls, and grounded sources. Off by default and runs statelessly (your conversation history stays on your device); turn it on under Settings → Agent config → API configuration, or leave it off to keep using the proven generateContent path
🗺️ Google Maps grounding tool — the agent can ground answers in Google Maps data for place and location questions (#964)
🧠 Per-use-case thinking depth — reasoning effort is now tuned per task (completions think the least, agent chat the most) instead of one global setting, balancing latency and quality (#986)
⏳ Soft agent turn budget — long agent runs get a gentle reminder as they approach the turn limit, plus a one-shot extension, instead of stopping abruptly (#996)
📋 Copy buttons for tool calls — quickly copy tool-call parameters and results from the agent view (#960)
🔄 In-app model-list refresh — refresh the available Gemini model list without restarting (#908)
📦 More bundled skills — obsidian-markdown and json-canvas skills ship built in, with richer obsidian-bases knowledge (#963)
⬆️ Dependency refresh — @google/genai 2.10.0 and @allenhutchison/gemini-utils 1.1.0

4.10.1 is a notes-only patch that completes the 4.10.0 release notes — there are no functional changes from 4.10.0. The original 4.10.0 notes covered only the Interactions API work and omitted several user-facing features that also shipped in this line. The full 4.10 feature set: the first fully localized UI (the entire plugin interface — settings, modals, the agent view, command names, and notices — is now translated into 20 languages, selected automatically from your Obsidian language with a graceful fallback to English); the agent now surfaces the model's reasoning inline in the tool activity block and persists it in session history; an experimental, opt-in Interactions API transport for Gemini behind a default-off "Use Interactions API" setting (Settings → Agent config → API configuration) that streams text, reasoning, tool calls, and grounded Google Search sources statelessly while keeping the proven generateContent path as the default; a Google Maps grounding tool for location questions; per-use-case reasoning depth; a soft agent turn budget with a one-shot extension; copy buttons for tool-call parameters and results; in-app refresh of the Gemini model list; and additional bundled skills (obsidian-markdown, json-canvas). Dependencies were refreshed including the Gemini SDK (@google/genai 2.10.0) and the shared gemini-utils package (1.1.0).

🌍 Gemini Scribe 4.10.0 - Multilingual UI, Reasoning Display & Interactions API v4.10.0

🌍 Localized UI in 20 languages — the entire plugin interface (settings, modals, agent view, commands, notices) is now translated, auto-selected from your Obsidian language with graceful fallback to English
🧠 Model reasoning display — the agent now shows the model's thinking inline in the tool activity block and persists it in session history, so you can follow how it reached an answer (#965)
🧪 Interactions API transport (experimental, opt-in) — a new "Use Interactions API" setting routes Gemini chat through Google's newer Interactions API, with full streaming of text, reasoning, tool calls, and grounded sources. Off by default and runs statelessly (your conversation history stays on your device); turn it on under Settings → Agent config → API configuration, or leave it off to keep using the proven generateContent path
🗺️ Google Maps grounding tool — the agent can ground answers in Google Maps data for place and location questions (#964)
🧠 Per-use-case thinking depth — reasoning effort is now tuned per task (completions think the least, agent chat the most) instead of one global setting, balancing latency and quality (#986)
⏳ Soft agent turn budget — long agent runs get a gentle reminder as they approach the turn limit, plus a one-shot extension, instead of stopping abruptly (#996)
📋 Copy buttons for tool calls — quickly copy tool-call parameters and results from the agent view (#960)
🔄 In-app model-list refresh — refresh the available Gemini model list without restarting (#908)
📦 More bundled skills — obsidian-markdown and json-canvas skills ship built in, with richer obsidian-bases knowledge (#963)
⬆️ Dependency refresh — @google/genai 2.10.0 and @allenhutchison/gemini-utils 1.1.0

This is the first release with a fully localized UI: the entire plugin interface — settings, modals, the agent view, command names, and notices — is now translated into 20 languages, selected automatically from your Obsidian language with a graceful fallback to English. The agent also now surfaces the model's reasoning: its thinking is shown inline in the tool activity block and persisted in session history so you can follow how it reached an answer. This release adds an experimental, opt-in Interactions API transport for Gemini — Google has made its Interactions API generally available, and a new "Use Interactions API" setting (Settings → Agent config → API configuration) lets you route chat through it while keeping the proven generateContent path as the default. The transport is stateless (the plugin continues to own and replay your conversation history locally) and supports the full experience: incremental streaming of text and reasoning, tool calls, and grounded Google Search sources; it's behind a default-off toggle because the underlying SDK surface is still marked experimental, so nothing changes unless you turn it on. Also new: a Google Maps grounding tool for location questions, per-use-case reasoning depth (latency-sensitive completions think minimally while the agent loop thinks the most), a soft agent turn budget with a one-shot extension, copy buttons for tool-call parameters and results, in-app refresh of the Gemini model list, and additional bundled skills (obsidian-markdown, json-canvas). Dependencies were refreshed including the Gemini SDK (@google/genai 2.10.0) and the shared gemini-utils package (1.1.0).

🔧 Gemini Scribe 4.9.1 - Vault Context Fix v4.9.1

🗂️ Fixed the "Initialize Vault Context" button — it sent a malformed model request and failed to generate AGENTS.md

This patch fixes the "Initialize Vault Context" / "Update Vault Context" button in the agent view. The request sent to the model was shaped so that the analysis prompt was discarded, causing the model to return a conversational reply that failed to parse and never produced an AGENTS.md file. The request is now built correctly and the feature works again.

🪝 Gemini Scribe 4.9 - Lifecycle Hooks, Stable Prefix Caching, Custom Endpoint v4.9.0

🪝 Lifecycle hooks engine — define hooks that fire on Obsidian events and run summarize/rewrite/command/agent actions, with full management UI and a settings toggle
⚡ Prompt-prefix caching stabilized — the agent's system instruction is now byte-stable across turns and tool follow-ups, restoring Gemini's implicit prefix cache so long sessions stop reprocessing history each turn (closes #763)
🌐 Custom API endpoint — new setting to point Gemini Scribe at a proxy or alternate Gemini-compatible endpoint, covering all SDK call sites
📐 Collapsible settings sections — settings page reorganized into foldable sections (foundation for closing #430)
🎯 Agent sidebar actions as commands — new sidebar actions registered as Obsidian commands, keyboard-bindable (closes #785)
🔐 MCP hardening — stdio server env vars now live in Obsidian's encrypted SecretStorage; offline or unreachable MCP servers no longer block plugin load
🛡️ Unified tool-policy — Sessions, Projects, Scheduled Tasks, and Hooks now share one policy model so per-feature permissions behave consistently
📂 Two-phase context compaction — long conversations now truncate older turns first, summarize second, for cleaner handling near the token limit
🧱 Per-provider package layout — src/api/providers/{gemini,ollama}/; GeminiClientFactory → ModelClientFactory (breaking change for plugins building on the SDK; OllamaClient and OllamaClientConfig are now in the public barrel)
🧪 Eval-harness maturity — baseline-compare with auto-bless, richer rubric matchers plus LLM-as-judge fallback, per-task timeouts, clean interrupt handling, and an eval-harness skill

This release introduces a lifecycle-hooks engine — define hooks that fire on Obsidian events (note modify, command palette, etc.) and run summarize/rewrite/command/agent actions with a dedicated management UI. Agent prompt-prefix caching has been stabilized (closes #763): the system instruction is now byte-stable across turns and tool follow-ups, restoring Gemini's implicit prefix cache so long chat or coding sessions no longer reprocess history each turn. A custom API endpoint setting lets you point Gemini Scribe at a proxy or alternate Gemini-compatible host. MCP servers are now resilient on offline machines and store stdio env vars in Obsidian's encrypted SecretStorage. Tool-policy is unified across Sessions, Projects, Scheduled Tasks, and Hooks. The settings UI gains collapsible sections, and new agent sidebar actions are registered as Obsidian commands. Under the hood: per-provider packages under src/api/providers/ (GeminiClientFactory → ModelClientFactory; breaking for SDK consumers, OllamaClient now in the public barrel), two-phase context compaction, tool-result payload truncation in older history, retry/backoff applied to non-chat direct SDK calls, and substantial eval-harness maturity work (baseline-compare, LLM-judge fallback, per-task timeouts, clean interrupt) plus broader test coverage.

✨ Gemini Scribe 4.8 - Scheduled & Background Tasks, Ollama, Activity Modal v4.8.0

⏰ Scheduled tasks — run agent tasks on a cron, time-of-day, or day-of-week schedule with full management UI
🌙 Missed-run catch-up — tasks that should have run while Obsidian was closed surface on startup for review
🛰️ Background tasks — deep research and image generation now run in the background with output consolidated under [state-folder]/Background-Tasks/
📊 Unified activity modal — Background Tasks and RAG status share a single tabbed view
🦙 Ollama provider — point the plugin at a local Ollama server for offline, local-model chat
🛑 Runaway-loop abort — repeated tool-loop detections within a turn now abort the turn with a clear notice instead of churning
🧪 Eval harness — new harness for agent-loop changes with multi-hop retrieval, loop-trap tasks, pass^k reliability, and per-run model overrides
📱 Mobile fixes — agent view layout on iOS and CatchUpModal now opens directly on mobile
🛠️ Headless agent loop — AgentLoop extracted from the agent view so scheduled and background runners share the same execution engine

This release introduces scheduled tasks: define agent tasks with cron, time-of-day, or day-of-week schedules and manage them with a dedicated UI. If Obsidian was closed when a run was due, a catch-up modal surfaces missed runs on startup. Long-running operations like deep research and image generation now run as background tasks, with all their output consolidated under [state-folder]/Background-Tasks/, and a unified activity modal gives you a single place to monitor background work alongside RAG indexing status. Ollama joins as a provider option for fully local chat. The runaway-loop detector now escalates to a per-turn abort with a user-visible notice. Under the hood, AgentLoop has been extracted so scheduled and background runners share the same engine as the UI, and a new eval harness with multi-hop, loop-trap, and pass^k reliability checks helps verify agent-loop changes. Includes mobile fixes for iOS layout and the catch-up modal, plus a migration of the test runner from ts-jest to Vitest.

✨ Gemini Scribe 4.7 - Projects, Session Memory & File Shelf v4.7.0

📂 Projects — scope agent sessions to a folder with custom instructions and tool/skill overrides
🧠 Session recall — the agent can search past conversations for relevant context
📎 File context shelf — unified drag-and-drop bar replaces the old context panel
📝 Diff review view — write and edit operations show a side-by-side diff before applying
📦 Bundled skills — built-in help and Obsidian knowledge skills, auto-generated from docs
📄 Binary file awareness — agent tools can read images, audio, video, and PDFs directly
🏗️ Layered prompt architecture — system prompts refactored into composable Handlebars sections
⚡ Strict TypeScript checks — property initialization and unused variable enforcement

This release introduces Projects, letting you scope agent sessions to a folder with custom instructions, permission overrides, and skill filters. Session recall gives the agent memory across conversations — it can search past sessions for relevant context. The new file context shelf replaces the old context panel with a compact drag-and-drop bar that auto-expands folder contents. Write and edit tool calls now show a diff review view so you can inspect changes before they're applied. Bundled skills ship built-in help and Obsidian knowledge that are auto-generated from the docs site at build time. Agent tools can now read binary files (images, audio, PDFs) encountered during tool execution. The prompt system has been refactored into layered Handlebars templates for better maintainability. Under the hood, a new event bus architecture decouples lifecycle hooks, and strict TypeScript checks improve code reliability.

✨ Gemini Scribe 4.6 - Context Management & Tool Permissions v4.6.0

🧠 Automatic context compaction — long conversations are summarized to stay within model limits
📊 Live token usage display with cached/uncached breakdown (opt-in in settings)
🛡️ Per-tool permission policies — configure allow, deny, or ask-every-time per tool
🔧 Collapsible tool activity bar — tool calls grouped into expandable sections
💭 Expandable thinking section replaces tooltip for model reasoning
📁 Fixed plugin init failure when folders already exist (Obsidian Sync)

This release adds automatic context management that monitors token usage and summarizes older conversation turns when approaching model limits. A new token counter (enable in settings) shows live usage with a breakdown of cached vs new tokens. Tool permissions can now be configured per-tool with allow/deny/ask policies. Tool calls are grouped into collapsible activity bars for cleaner chat. The thinking model output now uses an expandable section instead of a tooltip. Also fixes a plugin initialization failure when using Obsidian Sync.

🔧 Gemini Scribe 4.5.1 - iOS Compatibility Fix v4.5.1

🍎 Fixed plugin failing to load on iOS due to Node.js module imports

This patch fixes the plugin failing to load on iOS devices. The issue was caused by Node.js module imports that are not available in the mobile environment.

✨ Gemini Scribe 4.5 - MCP Servers, Skills & Secure Storage v4.5.0

🔌 [Experimental] MCP server support with HTTP transport and OAuth authentication
🧠 Agent skills system following the agentskills.io specification
📎 Unified drag-and-drop pipeline with file type classification
🔐 API key migrated to Obsidian SecretStorage for secure handling
🔍 Search box in Add Context Files modal for quick filtering
⚡ Default chat model changed to Flash for free API key compatibility
🔄 RAG indexing now starts when toggled without requiring plugin reload
📚 New documentation site with expanded FAQ and contributing guides

This release adds experimental MCP (Model Context Protocol) server integration with HTTP transport and OAuth support, letting you connect external tools to your agent. The new agent skills system allows extensible AI capabilities following the agentskills.io spec. Your API key is now stored securely via Obsidian's SecretStorage (OS keychain). File attachments have been unified with smart classification for text, media, and binary files. The default chat model is now Flash for better compatibility with free API keys.

✨ Gemini Scribe 4.4 - Deep Research & Quality of Life v4.4.0

🔬 Deep research migrated to gemini-utils ResearchManager
🛡️ Trusted Mode with extended vault tools
🔤 Fixed HTML entities appearing in chat responses
📝 HTML markup now preserved in saved chat history
🔧 Better empty response messages list executed tools
✨ Subtle fade-in animation for confirmation results

This update brings an improved deep research engine, trusted mode for streamlined tool execution, and several quality-of-life fixes. HTML entities returned by Gemini models are now properly decoded, and markup is preserved in saved history files. Empty response messages now list which tools were executed for better transparency.

🔧 Gemini Scribe 4.3.1 - Setup Experience Fix v4.3.1

🔑 Fixed plugin setup for new users without API key
⚙️ Settings are now always accessible, even before configuration
🛡️ Graceful error handling when API key is missing
🔄 Plugin automatically activates when API key is added

This update fixes a critical issue where new users could not access settings to configure their API key. The plugin now loads partially when unconfigured, allowing access to settings. Once configured, the plugin automatically activates without needing to restart Obsidian.

✨ Gemini Scribe 4.3 - Images & Selection Actions v4.3.0

🖼️ Multimodal image support - attach images to your chats
✨ Explain Selection - get AI explanations of selected text
❓ Ask about Selection - ask questions about selected text
🔍 Folder and tags filtering for semantic search
🔗 Fixed @ mentions to use proper wikilink paths
⌨️ Fixed IME composition issues for international keyboards
🛑 Fixed Stop button to properly halt pending tool executions
📦 Removed obsidian-dataview dependency

This update adds multimodal capabilities - you can now attach images to your agent conversations. New selection-based actions let you quickly explain or ask questions about any selected text via the right-click menu. Semantic search now supports folder and tag filtering. Several bug fixes improve reliability for international users and tool execution.

🔧 Gemini Scribe 4.2.1 - RAG Stability & New Features v4.2.1

🐛 Fixed RAG re-indexing on every Obsidian restart
📄 PDF and attachment indexing support
⏸️ Pause/resume commands for RAG sync
📊 Detailed status modal with file lists and search
🔄 Resume interrupted indexing after crash/restart
⚡ Rate limit handling with automatic retry
💾 Incremental cache saves for durability

This update brings major stability improvements to RAG indexing. The vault no longer re-indexes on every restart, and you can now index PDFs and attachments. New pause/resume commands give you control over syncing, and interrupted indexing can be resumed. The status modal now shows detailed file lists with search functionality.

✨ Gemini Scribe 4.2 - Semantic Search & Improved Errors v4.2.0

🔬 [Experimental] Semantic vault search using Google File Search API
🗂️ Background indexing keeps your vault searchable
💬 Clearer API error messages (quota, auth, rate limits)
🖼️ Fixed image model dropdown in settings
✏️ Fixed writing tool to respect YAML frontmatter

This update introduces experimental semantic search powered by Google's File Search API. When enabled in Advanced Settings, your vault is indexed in the background, allowing the AI to search by meaning rather than just keywords. Also includes improved error messages that clearly explain API issues like quota limits or authentication problems.

🐛 Gemini Scribe 4.1.2 - Writing Tool Fix v4.1.2

✏️ Fixed writing tool to properly respect YAML frontmatter
📝 Content is now correctly placed after frontmatter blocks
🔍 Added edge case handling for malformed frontmatter
📚 Improved documentation for YAML handling

This update fixes an important issue where the writing tool would incorrectly place content at the very beginning of files, overwriting or disrupting YAML frontmatter. The tool now properly detects and preserves frontmatter blocks (defined by --- delimiters), placing new content after them as intended.

🐛 Gemini Scribe 4.1.1 - Stability & UX Improvements v4.1.1

💬 In-chat confirmations - no more hidden modal dialogs
🧠 Fixed Gemini 3 thinking mode display
🔧 Fixed Gemini 3 function calling with thought signatures
⏱️ Agent timeout protection prevents infinite hangs
🎨 Better visual feedback during tool execution

This update focuses on stability and user experience. Confirmation dialogs are now inline in the chat, Gemini 3 models work properly with thinking mode and function calling, and the agent includes timeout protection to prevent getting stuck.

✨ Gemini Scribe 4.1 - Enhanced AI & Better UX v4.1.0

🌍 Multilingual support - prompts in your language
🧠 Gemini 2.5 Pro & Gemini 3 with thinking progress
🛑 Stop button to cancel long-running operations
💡 Dynamic example prompts based on your vault
🎨 Improved UI with icon buttons and progress indicators
🔍 Enhanced search with new file content tool

This update brings powerful new AI models, multilingual support, and major UX improvements. Includes important security fixes and better vault operations.

🎉 Welcome to Gemini Scribe 4.0! v4.0.0

🤖 Unified agent-first interface - one powerful chat mode
🔧 Tool calling built-in to every conversation
💾 Persistent agent sessions with full history
📦 Old history safely archived as readable markdown

This is a major update focused entirely on the powerful Agent Mode. The old note-based chat has been removed in favor of a unified agent experience with tool calling, persistent sessions, and better context management.

🔧 Gemini Scribe 3.3.2 - SDK & Context Fixes v3.3.2

⏱️ Added timer to thinking indicator in agent mode
📦 Updated @google/genai SDK to resolve urlContext type errors
🔧 Fixed Gemini API content format and restored file context

This maintenance release updates the Google GenAI SDK to fix type errors with urlContext, fixes API content formatting, and adds a timer to the thinking indicator in agent mode.

🐛 Gemini Scribe 3.3.1 - Popout Window & Model Fixes v3.3.1

📋 Fixed paste functionality in popout/detached windows
🔗 Fixed @ mention file chip insertion in popout windows
🤖 Fixed models not being used correctly for summary and rewrite tasks
🏷️ Fixed agent session auto-title generation
🧩 New reusable DOM context utility module

This release fixes several issues with popout/detached windows including paste and @ mentions, corrects model selection for summary and rewrite tasks, and adds a new DOM context utility for better window handling.

✨ Gemini Scribe 3.3 - Agent Mode v3.3.0

🤖 AI Agent with tool calling - read, write, search, and manage vault files
🔍 Google Search integration for web information
🌐 Web page fetching and analysis
💬 Session management with persistent conversation history
📎 Context file management with @ mentions
🛡️ Safety features: permissions, loop detection, system folder protection
📋 Copy button for user messages
🎨 Professional UI with collapsible tool messages and animations

A major feature release introducing Agent Mode - an AI-powered assistant that can actively work with your vault through tool calling. Includes comprehensive file operations, Google Search, web fetching, session management, and safety features like loop detection and system folder protection.

⚙️ Gemini Scribe 3.1 - Dynamic Parameters & Advanced Settings v3.1.0

🌡️ Smart temperature ranges that auto-adjust based on model capabilities
🎛️ Model-aware Top P controls with proper validation
🔧 New Advanced Settings panel for developer-focused options
🔍 Automatic model discovery with parameter limit detection
💾 Smart 24-hour model cache with manual refresh
🧪 Comprehensive new test coverage

This release adds dynamic model parameter controls that automatically adjust ranges based on your Gemini model's actual capabilities. The new Advanced Settings panel organizes developer options cleanly, and the model discovery system automatically fetches available models with their parameter limits.

🎉 Gemini Scribe 3.0 - Major Release v3.0.0

📁 Plugin state folder restructured for better organization
📝 Enhanced custom prompt system with command palette commands
🔍 Dynamic model discovery from Google's API
⚡ Real-time streaming responses in chat
🔧 Replaced low-level file operations with native Obsidian APIs
🚀 Automatic migration from flat folder structure

A major release with significant architectural improvements. Plugin data is now organized into subdirectories (History, Prompts, Agent-Sessions), custom prompts get command palette integration, and model discovery pulls available models directly from Google's API. Includes automatic migration for existing users.

⚡ Gemini Scribe 2.2 - Streaming Responses v2.2.0

⚡ Real-time streaming AI responses in chat
🔄 Intelligent streaming retry with automatic fallback
🧪 Comprehensive test coverage for streaming
📚 Added implementation planning guidelines
🐛 Fixed UI stuttering and duplicate content during display

This release adds real-time streaming support for AI responses. See responses appear progressively as they're generated, with smart fallback to standard responses if streaming isn't available. Includes new CI/CD workflows and improved developer experience.

🔧 Gemini Scribe 2.1 - Retry Logic & Stability v2.1.0

🔄 Retry logic for model API requests with exponential backoff
🧪 Improved test coverage for models and debug modules
🔤 Fixed HTML entity decoding in model responses
🔗 Fixed README links to new repository

This release adds robust retry logic for API requests with exponential backoff, fixes HTML entity handling in model responses, and improves test coverage across the codebase.

🐛 Gemini Scribe 2.0.3 - Context Handling Fix v2.0.3

🔧 Fixed linked notes being processed when Send Context is disabled
📄 Context Depth 0 now correctly includes only the active file
⚡ Prevented unnecessary file context processing when disabled

This release corrects context handling behavior. The Send Context toggle now strictly controls whether file content is gathered, and Context Depth of 0 properly includes only the active file without linked notes.

🐛 Gemini Scribe 2.0.2 - Stability Patch v2.0.2

🐛 Bug fixes and stability improvements

A stability patch with minor bug fixes following the 2.0.0 major release.

🎉 Gemini Scribe 2.0 - API Abstraction & History V2 v2.0.0

🏗️ Major internal refactor: API abstraction layer
📝 History V2 with improved storage and management
🔧 New model management system
⚠️ Breaking: significant internal changes

A major internal refactor introducing an API abstraction layer, new model management system, and History V2. This release modernizes the plugin's architecture for better extensibility and reliability.

🐛 Gemini Scribe 1.8.1 - Patch v1.8.1

🐛 Bug fixes and minor improvements

A minor patch release with bug fixes following the 1.8.0 release.

📱 Gemini Scribe 1.8 - Mobile Support v1.8.0

📱 Plugin now works on mobile with history enabled
📁 Flattened history filenames for cross-platform compatibility
🔄 Automatic migration of legacy history files

This release enables mobile support by flattening history filenames for cross-platform compatibility. Existing history files are automatically migrated to the new naming format.

Changelog ​