SystemSculpt v4.0.2 — compact streaming status pill, friendlier status labels, refreshed loader styles

TL;DR

The streaming status now appears as a compact pill with clearer labels and updated loading visuals.

Changed

• Streaming status (“Thinking…”) redesigned into a small pill that includes a spinner, making it less obtrusive.
• Status labels updated to more descriptive terms: Thinking, Writing, Using tools, Preparing.
• Loader visuals refreshed: CSS‑based loader styles replace the older braille spinner, keeping screen‑reader announcements intact.

SystemSculpt v4.0.1 — Improved tool error recovery, batch file actions, model ID fix

TL;DR

Adds automatic recovery from tool schema errors, safer batch file operations, and corrects model identifier handling.

The plugin is now more resilient when tool definitions contain errors, processes multiple file actions in controlled batches with clear notices, and sends model IDs to the server in a proper format.

Added

• Tool schema fallback – When a tool’s schema validation fails, the plugin automatically creates chunked calls and continues working instead of stopping.
• Batch filesystem operations – Filesystem actions are now processed in batches, limited by a maximum number of operations, and users receive a notice with typed results.

Changed

• Tool definitions – maxItems now matches the defined limit and descriptions are clearer for users.

Fixed

• Model ID normalization – Model identifiers are correctly formatted before being sent to the server, preventing malformed paths for OpenAI, Groq, and other providers.

Breaking Changes

None

SystemSculpt v4.0.0 — Quick‑edit widget, improved model list, smarter file suggestions

TL;DR

The update adds a fast‑access quick‑edit widget, normalises model identifiers for a clearer list, expands file suggestions to include sub‑folder notes, and refines streaming and credit handling for a smoother editing experience.

Summary

This release focuses on client‑side enhancements: a new quick‑edit UI component, clearer AI model naming, broader file suggestion coverage, more reliable streaming responses, and default monthly credit handling. Minor UI tweaks improve the overall usability of SystemSculpt within Obsidian.

Added

• Quick Edit Widget & Floating Widget styles – a new UI component for rapid editing, with floating‑style design and updated tool‑call handling.
• File suggestion enhancement – suggestions now include Markdown files located in sub‑folders, making custom prompts easier to discover.

Changed

• Model ID normalisation – Groq model identifiers are now shown in a consistent groq/<id> format, simplifying the model selection UI.
• Streaming improvements – more reliable server‑sent events parsing and footnote hints, reducing interruptions during AI‑generated content streams.
• API credits handling – default monthly credits are applied automatically, and a budget‑check tool helps you stay within your credit limits.

Fixed

• Minor UI adjustments to prevent timeouts on admin‑related pages (now faster to close).

Breaking Changes

• Legacy Groq model IDs (groq/openai/...) are automatically mapped to the new groq/... format

SystemSculpt v3.1.0 — Namespace management, status‑bar enhancements, and targeted embedding rebuild

TL;DR

Version 3.1.0 adds a full namespace manager for embeddings, shows active embedding details in the status bar, and gives you a new command to refresh embeddings for the current model.

Added

• Embeddings Namespaces panel in the settings page – view all stored embedding configurations, see provider/model details, switch to a different namespace, or delete unused ones.
• Status bar now displays the active embedding provider, model, and schema version (e.g., OpenAI • ada • v1).
• “Rebuild Embeddings (Current Model)” command – quickly refresh embeddings for the currently selected provider/model without touching other namespaces.
• Embeddings quiet period setting (default 1200 ms) to debounce rapid changes and avoid unnecessary re‑embedding.
• Query‑embedding cache – vectors are reused when the same content is encountered again, and the system now tracks folder moves/renames correctly.
• Quick‑Edit improvements – automatic registration of filesystem tools, a fallback to an internal server, and a notification when no changes are applied.

Changed

• Status bar behavior refined: always visible with a placeholder when idle, updates faster, and clearly indicates when embeddings are disabled.
• Embeddings storage is now namespace‑aware, preserving existing vectors when you switch providers or models.
• Introduced embedding schema versioning to keep different configurations from conflicting.
• Quick‑Edit now logs submit failures for easier troubleshooting.

Fixed

• Removed an unnecessary subscription status mapping that simplified user creation.
• Resolved a hotkey conflict that prevented the Quick‑Edit command from working.
• Stabilized OpenRouter streaming by switching to a non‑streaming SSE wrapper, preventing intermittent connection issues.

Breaking Changes

• None. All existing workflows continue to work; new features are additive.

SystemSculpt v3.0.0 — Real-time token counter, enhanced streaming, new file-edit flow

TL;DR

Everything streams smoother, you can now see live token counts while typing, and a new Quick Edit modal lets you modify any file with natural-language instructions.

Added

• Live token counter in the chat input: watch your message, context files, system prompt, and tool calls consume tokens in real time; tap for a detailed breakdown by category.
• Quick Edit modal (Cmd+Enter or Ctrl+Enter while editing a note): open a small overlay, type “add headings to this note” or “convert to bullet list”, and see the changes stream in with an inline diff before you accept.
• Interactive settings panel on empty chats: toggle model, agent mode, web search, and more via pill-style buttons without opening the settings tab.
• Automatic microphone recovery: recordings continue seamlessly when you swap audio devices (e.g., AirPods connect/disconnect).
• Inline file-operation previews for move, trash, and folder creation—see exactly what will change before approving tool calls.
• Enhanced token display for tool calls: each tool shows its token cost inline and in the result modal.

Changed

• Streaming experience now preserves the order of reasoning and content blocks, removes UI flicker, and gates updates on newlines for smoother reading.
• Tool result storage moved from standalone messages into the assistant message’s tool_calls array, reducing clutter in chat history.
• Model selector opens instantly from the chat view and fills progressively; provider names are shown beside every model.
• Token language simplified: “context usage” replaces “tokens” in most labels.
• Input placeholder now reads “Search your vault, link notes, write ideas” for clearer intent.

Fixed

• Chat scroll behavior no longer jumps when images or long tool results load.
• Malformed JSON in tool arguments is auto-repaired instead of failing.
• Tool calls with duplicate or missing IDs are handled gracefully across different provider formats.
• Server error messages display correctly even when the upstream returns plain text.
• Legacy consecutive-assistant messages merge cleanly when loading old chats.

SystemSculpt v2.4.0 — Centralized Agent Tools, Favorites Management, Smarter Tool Fallbacks

TL;DR

Manage Agent Mode tools in Settings, enjoy a new favorites manager, and see gentler notifications when AI models can’t use tools.

Added

• New Favorites Manager – a full UI to search, filter, bulk-select and organize your favorite models.
• Agent Tools section to Chat Settings – enable or disable Agent Mode tools directly in the settings panel; no more pop-up modals.
• Inline notification for tool fallbacks – when a model can’t use tools, a subtle footnote appears in the response instead of a disruptive popup.

Changed

• Agent Tools button removed – the gear-wheel button no longer appears in the chat view; everything is handled in Settings.
• Favorites handling overhauled – lists now stay in sync across the app, favorites are automatically deduplicated, and you can pin them to the top of every model picker.
• Model pickers and modals refined – keyboard navigation, selection feedback, and empty-state messages updated for smoother operation.

Fixed

• Template default selections now persist – choosing a default template in the Templates tab is immediately saved and remembered.

SystemSculpt v2.3.0 — Smarter Agent Mode, Streamlined Tools, Simplified Setup

TL;DR

Agent Mode now works out-of-the-box with automatic read-only vault access and a simple vault-icon toggle, while all filesystem tools got shorter, friendlier names.

What's New

Agent Mode Enhancements

• One-click vault exploration – flip the new vault icon in the chat header and every read-only tool (find, read, list, search, etc.) is enabled and auto-approved; no extra settings or dialogs
• Automatic prompt merging – when Agent Mode is on, your chosen system prompt gets a short agent prefix behind the scenes so you can keep your favourite prompt and still get full agent capabilities

Filesystem Tools

• Shorter, clearer names – read_files → read, grep_vault → search, create_directories → create_folders, and ten more; configurations migrate automatically
• Storage analysis improvements – new “ignore system files” option and a concise summary when nothing matches your filters

Global Consistency Settings

• Use Latest Everywhere toggle – switch it on and every new chat, title generator, and post-processor automatically uses your most recent model and prompt choice
• Consistent model labels – every menu now shows provider tags (e.g., [CLAUDE] Sonnet 3.5) so you always know which service is answering

Performance & Polish

• Faster folder listings – file previews are now capped to 12 small files (≤200 KB) to keep things snappy
• Simplified first run – the old guided tour is gone; the welcome modal now offers a single “Get Started” button and closes immediately
• Cleaner UI spacing and a handful of small layout fixes

SystemSculpt v2.2.1 — Local LLM auto-detection, fixed OpenRouter streaming, visual menu divider

TL;DR

One-click discovery and setup of local LLM servers plus more reliable connections to OpenRouter and other providers.

Added

• Local LLM Auto-Detection – Settings and setup modals now include a “Detect Local Providers” button that finds LM Studio (port 1234) and Ollama (port 11434) automatically and adds them ready to use.
• Stream Safety – A 30-second inactivity timeout prevents stuck responses when a provider keeps the connection open longer than expected.

Changed

• Provider Menu Layout – A subtle divider now separates your saved providers from the “add” and “settings” actions, making long menus easier to scan.

Fixed

• OpenRouter Connections – Removed extra headers that were triggering CORS errors, so OpenRouter (and similar providers) now connect reliably.
• Header Case Sensitivity – Response detection now works regardless of how providers capitalize their headers.
• Streaming Edge Cases – Better handling of comment-only lines and mixed-format deltas keeps responses smooth across a wider range of providers.

SystemSculpt v2.2.0 — Copy buttons on code blocks, faster streaming

TL;DR

Smoother chats, and one-click code copying.

Added

• Copy buttons for code blocks – Hover any code block in chat to reveal a copy button with instant confirmation
• Auto-filled bug reports – Feedback now opens a pre-filled GitHub issue with your environment details
• Embeddings settings for everyone – Provider choice and local scan options moved out of Advanced mode

Changed

• Feedback goes to GitHub – Replaced the website feedback form with direct GitHub issues for faster help

Fixed

• Bold and list formatting restored – Asterisks and bullets no longer get mangled in assistant replies
• Reasoning drawer width – Prevents the reasoning section from growing during streaming
• Hover colors unified – Message content parts keep consistent backgrounds on hover
• Context files land in the right place – Files now insert just before your latest message instead of at the start

feat: SystemSculpt v2.1.0 — Claude 4 models, instant drawers, flat chat layout

TL;DR

SystemSculpt 2.1.0 adds the new Claude 4 family, strips away sluggish animations, flattens the chat UI, and polishes every interaction.

✨ New Features

• Claude 4 models – Opus 4.1 / 4 and Sonnet 4 (200 k context, extended thinking) with automatic alias resolution (claude-opus-4-1 → claude-opus-4-1-20250805).
• Hover toolbar – one-click copy, edit, delete, reply on every message; swipe-out version on mobile.
• Debug toolkit – toggle verbose logs, export error dumps to SystemSculpt/Logs.

🎯 UX & Performance

• Instant drawers – reasoning and tool-call drawers open/close with 0 ms delay.
• Flat chat layout – removed message grouping and glassy styling; cleaner, faster, full-width interface.
• Slash-command polish – Enter no longer inserts newlines in confirmation dialogs; focus returns instantly.

🔧 Fixes & Reliability

• Reasoning preserved verbatim – bold markers, spacing, and original formatting stay intact.
• Title generation – better model selection, fallback logic, and validation.
• Streaming & API – robust error handling and logging in Anthropic/OpenAI adapters.

🏗️ Developer & Build

• Build logger – colorful, timestamped build output with icons.
• Bundle validation – in-memory esbuild checks catch issues before real builds.

feat: SystemSculpt v2.0.0 — unified vault search, embeddings overhaul, production build

TL;DR

Everything is faster, cleaner, and ready for production: lightning-fast hybrid vault search replaces legacy search, the entire embeddings stack was rebuilt for speed and reliability

Highlights

• Vault Search: Press Cmd+K for instant lexical + semantic search across all notes, complete with relevance scoring, drag-drop to chat, and a new ribbon icon.
• Embeddings Re-architected: From a monolithic 2 700-line manager to a modular Storage/Processor/Search design with chunk-level indexing, multi-provider support, configurable caching, and O(log n) vector search—startup is 5–10× faster and memory usage is adaptive.

Search & UX

• SystemSearchModal: Hybrid lexical/semantic search, background indexing, filter toolbar, live results.
• Ribbon icons: Fast vault-wide search and Similar Notes panel toggles.
• Cmd+K shortcut: Opens unified search anywhere in Obsidian.

Embeddings & Performance

• Modular core (core/Storage, core/Processor, core/Search) with dependency injection.
• Vector indexing: LSH-based index + optimized VectorSearch for 10–100× faster similarity queries.
• Multi-provider: SystemSculpt, OpenAI; Voyage AI has been removed but you can still use it with your own API key if you'd like; factory + circuit-breaker pattern.
• Memory management: Dynamic cache sizing (20-500+ items), auto-cleanup, no more leaks.
• Auto-processing: True set-and-forget; starts 3 s after layout-ready, 10 s status refresh, 15 s cache.
• Token safety: Pre-flight token estimation, dynamic batching, oversized-file truncation with user notice.

New User Onboarding

• Welcome modal + guided tour: Replaces old Tips modal; spotlight effect, keyboard navigation, fallback content.
• Setup hero: 3-step “Get set up in minutes” flow with live status badge and test connection.

Miscellaneous

• Settings refactor: 6 consolidated tabs, searchable settings UI, responsive design.
• Styling refresh: Glass-morphism UI, unified buttons/cards, improved mobile layout.
• Tooling UI: Batch tool-call approver with concurrency limits, timeout, auto-approve.
• AI agents: Marketing content generators + UI knowledge base for 50+ elements & 15 workflows.

feat: SystemSculpt v1.5.2 — fixed support for OpenAI reasoning models

TL;DR

OpenAI's reasoning models (o1/o3/o4) now work perfectly without any configuration errors.

✨ What's New

• Full Support for OpenAI Reasoning Models
  • The o1, o3, and o4 series reasoning models now work seamlessly without any configuration errors.
  • SystemSculpt automatically handles the special requirements of these models, so you can just select them and start using them.
  • No more temperature parameter errors - the system knows these models don't support temperature settings and handles it gracefully.

🛠 Improvements

• Smarter Model Handling
  • The system now automatically detects reasoning models and adjusts settings accordingly.
  • Better error messages if you try to use unsupported features with specific models.
  • Improved compatibility checking ensures you always know which features work with your chosen model.

feat: SystemSculpt v1.5.1 — Improved web search stability and mobile compatibility

TL;DR

Web search now works more reliably with better mobile support and smoother performance.

🛠 Improvements

• More Stable Web Search

  • Web search now uses a more reliable connection method that works consistently across all AI models.
  • Fixed issues where web search could sometimes fail or return incomplete results.
  • The system now handles web search requests more efficiently, reducing potential errors.

• Better Mobile Experience

  • Web search results now display properly on mobile devices and tablets.
  • Improved streaming of search results on mobile for smoother real-time updates.
  • Fixed display issues where web search indicators weren't showing correctly on mobile.

• Smoother Performance

  • Removed unnecessary debug messages that could clutter your view.
  • Optimized how web search data flows through the system for faster response times.
  • Better handling of network interruptions during web searches.

feat: SystemSculpt v1.5.0 — More reliable connections for global users

TL;DR

SystemSculpt now works more reliably worldwide with improved server infrastructure, plus the Similar Notes Panel keeps finding connections automatically.

✨ What's New

• Better Global Connectivity

  • Switched to a more robust global provider to handle our growing international user base.
  • Significantly improved connection reliability, especially for users outside North America.
  • Faster response times and fewer timeouts no matter where you're located.

• More Robust Infrastructure

  • Enhanced server-side error handling means fewer disconnections during long conversations.
  • Improved retry mechanisms automatically recover from temporary network issues.
  • Better handling of network switches (like going from WiFi to mobile data).

• Comprehensive Documentation

  • Brand new documentation covering everything from getting started to advanced features.
  • Step-by-step installation guides and troubleshooting tips.
  • Clear explanations of all AI features, settings, and integrations.

🛠 Improvements

• Similar Notes Panel Auto-Processing

  • The "Start Processing" button now properly enables continuous background processing.
  • Once you start processing, the panel keeps finding new connections as you work, without needing to restart.
  • No more confusion about when processing is active - it just works continuously in the background.

• Responsive Interface Improvements

  • Better handling of the toolbar on touch devices like tablets.
  • Improved interaction states and visibility across different screen sizes.
  • Smoother experience when switching between devices.

• Enhanced Credit Management

  • Credits now separate into monthly (refreshing) and additional (permanent) types for better flexibility.
  • Improved tracking shows you exactly how your credits are being used.
  • Option to purchase additional credits that never expire.

feat: SystemSculpt v1.4.3 — Better agent mode reliability and smooth large text handling

TL;DR

Agent mode now works more reliably, pasting large amounts of text no longer freezes the interface, and large content in chat history is easier to manage.

✨ What's New

• More Reliable Agent Mode

  • Agent mode now only activates when the proper SystemSculpt model is available, eliminating confusing "Model unknown" errors.
  • If the agent model isn't available, the feature is automatically disabled instead of causing problems.
  • Agent mode settings now work consistently across all parts of the interface.

• Smooth Large Text Pasting

  • Pasting extremely large amounts of text no longer freezes the interface.
  • You'll see helpful warnings when pasting large content, with size limits to keep things running smoothly.
  • Large text shows as a compact placeholder (like [PASTED TEXT - 472 LINES OF TEXT]) instead of cluttering your input field.

• Better Large Content Management

  • Large messages in your chat history now show a preview instead of overwhelming the interface.
  • When you encounter large content, you can easily view it in a full-screen modal, save it as a file, or copy it to your clipboard.
  • The viewing modal now uses Obsidian's native design for a better, more familiar experience.

🐛 Key Fixes

• More Stable Chat Processing

  • Fixed issues that could cause chat conversations to get stuck or display errors during streaming.
  • Improved how the plugin handles unexpected data to prevent crashes.
  • Chat operations now recover more gracefully when problems occur.

• Agent Mode Display Issues

  • Fixed problems where agent mode could show incorrect model information.
  • System prompt indicators now display properly when loading saved agent conversations.
  • Eliminated cases where agent mode could activate with the wrong model.

feat: SystemSculpt v1.4.2 — Enhanced embeddings experience, streamlined setup, and improved AI model management

TL;DR

Major embeddings workflow improvements with real-time status feedback and better search capabilities; streamlined provider setup with persistent preferences; new system prompt management and chat deletion features.

✨ What's New

• Real-Time Embeddings Status Bar

  • Added a dedicated status bar that provides live feedback during embeddings processing, so you always know when the system is working on your content.
  • Displays current processing state and helps you understand when embeddings operations are complete.

• Enhanced Embeddings Search Capabilities

  • Introduced semantic similarity search for finding related notes using embeddings technology.
  • Added intelligent UI that shows a disabled state when embeddings are turned off, with clear guidance on how to enable them.

• System Prompt Settings Tab

  • New dedicated settings section for managing system prompts and AI behavior customization.
  • Provides centralized control over post-processing settings and custom prompt configurations.

• Quick Setup Provider Buttons

  • Added one-click setup buttons for popular AI providers (OpenAI, Anthropic, OpenRouter, Ollama, LM Studio) to streamline initial configuration.
  • Features a new compact provider card layout that makes setup cleaner and more organized.

• Delete Chat Slash Command

  • You can now delete individual chat conversations directly from the chat interface using the new /delete command.
  • Includes a confirmation dialog to prevent accidental deletions and safely moves chats to your trash.

• Automatic Anthropic Endpoint Correction

  • The plugin now automatically detects and fixes malformed Anthropic API endpoints, ensuring your custom provider configurations work correctly.
  • Provides clear notifications when corrections are made and permanently updates your settings.

🛠 Improvements

• Persistent Provider Selection Preferences

  • Your model selection modal now remembers which providers you prefer to see, eliminating the need to reselect them each time.
  • Provider toggles save instantly without requiring modal closure, creating a smoother selection experience.

• Upgraded Embeddings Model

  • Switched to Voyage AI's voyage-3-large model for improved embeddings quality and performance.
  • Enhanced processing animations provide better visual feedback during embeddings operations.

• Streamlined Recorder Settings

  • Renamed post-processing setting to "Post-Processing Enabled" for better clarity.
  • Removed unnecessary button from the Recorder tab to create a cleaner interface.
  • Improved guidance text in Title Generation settings to help you configure prompts more effectively.

• Enhanced Agent Mode Management

  • Centralized agent mode logic eliminates inconsistent behavior and ensures the system prompt indicator displays correctly when loading chats.
  • Fixed race conditions that could cause UI elements to display incorrectly after chat loading.

• Improved Anthropic Integration

  • Cleaned up streaming implementation to work within Obsidian's platform limitations, providing more reliable responses.
  • Better error handling with more informative messages when issues occur.

⚠️ Heads-Up for Power Users

The embeddings system now uses Voyage AI's latest model, which may require re-processing existing embeddings for optimal performance. The system will handle this automatically when needed.

feat: SystemSculpt v1.4.1 — File attachment improvements, custom provider reliability improvements

TL;DR

Fixed critical custom provider model selection issues that were preventing users from accessing their custom AI models.

🐛 Key Fixes

• File Attachment in Chat

  • Fixed issue where users couldn't add files to chat conversations anymore.

• Custom Provider Model Selection

  • Fixed "Model unknown is unavailable with this provider" error that prevented users from selecting custom provider models.
  • Resolved issue where custom provider requests were incorrectly routed to OpenAI instead of the intended custom provider.
  • Fixed model names becoming corrupted when switching between providers.

🙏 Special Thanks

Big thanks to these community members who reported issues and helped identify problems:

• meeps - Reported file attachment issues in chat conversations
• haaa - First to report OpenAI provider "Stream processing failed" errors in v1.4.0
• esardi - Provided detailed error logs and confirmed the issue across multiple systems
• ronnykhalil - Identified OpenRouter routing problems and provided stability feedback
• grayerbeard - Tested across mobile platforms and helped isolate provider-specific issues

Your feedback helps make SystemSculpt better for everyone!

feat: SystemSculpt v1.4.0 — Embeddings fully overhauled, chat & agent mode streamlined, plus backup and UI refinements

TL;DR

Major embedding system overhaul using IndexedDB for speed and memory efficiency; chat and Agent Mode streamlined for faster access; plus new transcription auto-submit, backup service, vault-structure defaults, and cohesive UI updates.

✨ What’s New

• Embeddings System Overhaul

  • Migrated all embeddings storage to IndexedDB with efficient on-demand loading.
  • Added migration logic from legacy JSON and automatic pruning of stale entries.
  • Introduced cache limits (50 entries) and file-size checks (5 MB max) to prevent memory bloat.

• Chat & Agent Mode Streamlining

  • autoSubmitAfterTranscription setting enables chats to auto-submit post-transcription, reducing clicks.
  • Removed Agent Mode confirmation modal; Agent Mode toggles instantly for advanced workflows.
  • defaultIncludeVaultStructure setting auto-inserts vault outline into new chats for richer context.

• Automatic Backup Service

  • Implemented AutomaticBackupService for scheduled settings backups; manage frequency in Backup tab.

• Recording Settings Documentation

  • Added detailed markdown guide for recording behaviors (auto-transcription vs. post-processing).
  • Updated AudioTranscriptionModal to prioritize clean output over post-processing for accuracy.

• Embeddings Exclusions UI

  • New exclusion options in Embeddings settings: respect Obsidian folder/file exclusions and omit chat history.
  • Dynamic chip-based interface for managing exclusion lists.

🛠 Improvements

• Refined CSS across recorder, embeddings view, settings, and setup components for consistent spacing and visuals.
• Enhanced error handling in VersionCheckerService and ChangeLogTabContent to gracefully handle GitHub API rate limits.

feat: SystemSculpt v1.3.1 — Leaner, Smarter, Better Search

TL;DR

Removed outdated token counting overhead, supercharged vault search with intelligent scoring, and added powerful storage analysis tools.

✨ What's New

• Intelligent Search That Actually Gets You Find what you need faster with our new relevance scoring system. The AI now understands context better — searching for "daily notes" will prioritize files with both words together, not just random mentions of "daily" or "notes" scattered across your vault.

• Vault Storage Analyzer Curious where all your disk space went? The new storage analysis tool helps you understand your vault's footprint — find large files, see storage by file type, or analyze growth over time. Perfect for spring cleaning your knowledge base.

🛠 Improvements

• Lightning-Fast Search Results Completely rebuilt the search engine from the ground up. No more waiting for results — search happens instantly as you type, with the most relevant files always appearing first.

  • Smart keyword variations (searches for "myNote", "my-note", and "my_note" automatically)
  • Intelligent path scoring (boosts files in relevant folders, deprioritizes archives)
  • Clear relevance explanations showing why each result matched

• Cleaner, Faster Plugin Removed the legacy token counting system that was slowing things down. The plugin now uses a more efficient character-based approach that's both faster and more accurate.

  • Eliminated unnecessary dependencies for smaller plugin size
  • Improved startup performance by removing blocking operations
  • Better memory management for large vaults

• Smarter Provider Management Custom AI providers that fail repeatedly now auto-disable after 3 failures, preventing endless error loops. You'll get a clear notification when this happens so you can fix the issue.

🐛 Key Fixes

• Search Actually Works Now Fixed multiple issues where search would return empty results or freeze Obsidian. The new singleton pattern ensures only one search engine runs at a time.

• Embeddings Detection Fixed The plugin now correctly detects when embeddings are available, fixing the "No embeddings found" error that some users were experiencing.

• No More Startup Freezes Eliminated several blocking operations that could cause Obsidian to freeze on startup, especially for larger vaults.

🔒 Safety First

• Better API URL Handling Enhanced development mode support with automatic URL validation and correction. The plugin now intelligently switches between development and production servers based on your environment.

• Improved Error Recovery Added comprehensive error handling throughout the search system with proper cleanup on plugin unload.

— Mike

feat: SystemSculpt v1.2.2 — Agent Smarter, You Work Easier

TL;DR

Your SystemSculpt experience just got a big upgrade, making the agent more powerful, your notes easier to find, and giving you clearer insights into how you're using the system.

✨ What’s New

• Your Agent Got a Brain Boost

  • I've upgraded the agent (the part that helps you get things done) to a brand new, more powerful Google model. This means it's smarter, faster, and more consistent in how it helps you.

• Smarter Note Finding & Organization

  • I've made big improvements to how your notes are organized and how fast you can find them. The system now cleverly manages your note data, making everything run smoother.
  • There's a new feature that lets your agent find notes that are similar to what you're looking for, even if they don't have the exact words. This means more relevant results, faster.

• See Your Vault at a Glance

  • You can now easily see the entire structure of your vault (your folders and notes) right within your chat. It's like having a map of your knowledge base always at your fingertips.
  • When you're mentioning notes in chat, you'll also see your vault structure, making it super intuitive to pick the right file.

• Understand Your Usage, Instantly

  • A new on-screen counter now shows you exactly how much 'space' your prompts, notes, and conversations are taking up. This helps you understand and manage your usage better.

🛠 Improvements

• Clearer Messages & Fewer Headaches

  • I've made the system's internal messages and error reports much clearer, which helps me fix things faster if something goes wrong.
  • I've also improved how the system handles situations where it can't find certain information, so you'll see fewer confusing errors.

• Agent Instructions are Crystal Clear

  • I've fine-tuned the agent's internal instructions, so it understands its tasks better. This means it's more decisive and efficient when searching your vault and helping you out.

• Smoother Visuals & Feedback

  • When the system is working on something, you'll now see a cleaner, more modern loading indicator. It's a small change, but it makes the experience feel more polished.
  • I've also adjusted how messages appear in your chat, making them flow more naturally and feel more intuitive.

• Faster & Safer File Handling

  • I've optimized how the system reads and searches through your files, making it quicker and ensuring it doesn't try to send too much information at once.
  • If a file is too big to display fully, the system will now tell you clearly, so you're never left guessing.

• More Reliable Data Management

  • I've made the way the system identifies your messages more consistent, leading to better tracking and organization of your conversations.
  • I've also made the system smarter about checking your context files (the notes you're actively working with), ensuring that only valid files are included and your usage count is always accurate.

• Behind-the-Scenes Polish

  • I've cleaned up a lot of the code behind the scenes, making the system more flexible and easier to update in the future. This means more smooth improvements for you down the line.
  • I've also simplified some technical parts of the system, which helps me keep everything running smoothly and reliably.

feat: SystemSculpt v1.2.1 — Search, Validation, and Plugin Refinements

TL;DR

Improved search intelligence, enhanced code validation tools, and a cleaner, more focused plugin experience.

✨ What’s New

• Smarter Search Results

  • Your search function now intelligently categorizes results for clearer, more organized outcomes.
  • The search engine has been optimized to remove limitations on how many files it processes, ensuring more comprehensive results without sacrificing performance.

• Enhanced Error & CSS Validation

  • Our error checking tool now comes with detailed documentation, making it easier to understand its usage, exit codes, and output information.
  • A new CSS validation feature helps you identify duplicate classes, unused styles, and opportunities for optimizing your stylesheets.

• Streamlined Plugin Experience

  • We've removed development-specific features, like the standalone development mode indicator and the API server toggle, from production builds. This leads to a cleaner, less cluttered user interface.

🛠 Improvements

• Better Search Feedback

  • You'll now receive more detailed messages when searches time out or find no matches, guiding you on how to refine your search queries.
  • Search results are now scored and sorted based on how dense the matches are and how recently they were found, ensuring the most relevant information is prioritized.

• Clearer Code Checks

  • Improved console output for TypeScript and deprecated code checks now provides clearer guidance and actionable items, helping you maintain your code more easily.
  • New internal structures for CSS class and usage tracking enhance the code's overall organization and maintainability.

• Refined Core

  • The core service of the plugin has been simplified, improving its internal state management for better overall performance.

feat: SystemSculpt v1.2.0 — my biggest usability, performance & safety leap yet

TL;DR

Clearer tool-risk labels, a new autonomous “Vault Agent”, smoother streaming, refined UI, stronger mobile support, and a leaner, faster core.

✨ What’s New

• Safe Tool Categories Tools you enable are now split into Read and Write groups, with extra warnings before anything can modify your vault.

• Vault Agent (BETA) Opt-in mode that lets the AI roam your vault and act semi-autonomously. A dedicated info screen explains benefits and cautions before you switch it on.

• Full-Screen Mermaid Previews Click any diagram to pop it into a spacious, theme-aware viewer.

• Slash Command Palette Type “/” in chat to surface common actions without leaving the keyboard. Works great on mobile.

• Scroll-to-Bottom Shortcut One tap jumps you to the latest message; it hides itself when you’re already there.

• Adjustable Chat Font Size Choose Small, Medium, or Large text per chat for comfortable reading.

🛠 Improvements

• Tool Calls Feel Native

  • Collapsible drawers show live progress, results, and streamlined approve/deny buttons.
  • If multiple tools queue up, approve or reject them all in one click.
  • Auto-executed tools now match the neutral reasoning style—no more bright success blocks.

• Smarter Diff Viewer

  • Cleaner inline view with context trimming notices and horizontal scrolling when needed.
  • Under-the-hood editing is lighter, which means faster load times.

• Real-Time Streaming

  • Words appear token-by-token just like ChatGPT, then auto-format into full Markdown.
  • Reasoning sections collapse once they finish to keep chats tidy.

• Vault Filesystem Tools

  • Bigger read windows, richer metadata in results, and new bulk actions for moving or trashing items.
  • Search results now show just enough context without flooding the chat.

• Workspace Intelligence

  • Opening a file re-uses existing tabs when possible and splits sensibly on mobile.
  • New “Open Chat History” button lets you inspect the raw Markdown behind any conversation.

• Performance & Build

  • Development builds happily talk to localhost servers while production builds stay locked down.
  • Audio transcription size limit quadrupled from 500 MB to 2 GB.

🐛 Key Fixes

• Tool drawers no longer stick in a “pending” look after you approve them.
• Streaming text never cuts off or overlaps, and auto-scroll behaves on long chats.
• Chat saves can’t overwrite themselves mid-load, protecting your history.
• The O4 Mini model now executes tools reliably thanks to smarter message prep.

🔒 Safety First

• Enabling auto-approval on any write-capable tool now requires an explicit confirmation step.
• Every file-system path is validated and sandboxed to your vault.
• When a tool fails, the AI now sees a structured error instead of getting silently blocked, allowing graceful recovery without leaking sensitive details.

⚠️ Heads-Up for Power Users

1. Auto-accept setting changed — it’s now a list of specific tools instead of one global switch.
2. Custom themes may need tweaks if they relied on the old, bespoke CSS hooks; SystemSculpt now follows Obsidian tokens everywhere.
3. Legacy chat files still load fine, but new chats use a cleaner sequential format under the hood.

📊 Release Snapshot

• Roughly 9 300 lines added and 7 800 removed, yet the install package is 38 KB smaller.

Thank you for all the feedback that shaped this release. Dive in, explore, and let me know how the new SystemSculpt helps you craft better ideas—faster and safer than ever.

— Mike

feat: enhanced chat UI, resume chat command, improved model selection, system prompts, and error handling

Chat Interface Enhancements

• Resume Chat from Any History File: Resume previous chats directly from the command palette when viewing any chat history file. Two ways to continue conversations - the automatic resume button or through keyboard shortcuts.

• Smarter Model Selection: Prevents incompatible AI models from being selected during chat. When a model becomes unavailable, see a simple "Choose Different Model" button to quickly switch.

• No More Repeated Prompts: First-time users only asked once to select their preferred AI model when starting the plugin.

Visual and Interface Improvements

• Cleaner Layout: Model and system prompt indicators sit side-by-side at the top of the chat, each taking up half the width. On mobile devices, they stack vertically.

• Enhanced Visual Feedback:

  • Model indicator displays a bot icon
  • System prompt indicator shows a sparkles icon
  • Both indicators update instantly when you make changes
  • Context drawer highlights when files are added and removes highlight when cleared
  • "Add Context Files" button located inside the drawer

• System Prompt Selection:

  • Choose from pre-built prompts for general use or concise responses
  • Create and save custom prompts
  • Search and filter through available prompts
  • Preview prompts with syntax highlighting before selecting
  • Integrated with chat settings

Reliability Improvements

• Better Error Handling: Handles interruptions and cancellations gracefully, preventing confusing error messages.

• Instant Updates: All changes save automatically and reflect immediately in the interface.

Visual Polish

• Refined Appearance:
  • Smoother hover effects and transitions
  • Better dark mode support
  • Improved spacing and typography
  • More prominent visual indicators
  • Consistent styling across all elements

Enhanced plugin user experience with improved drag-and-drop, file handling, and message generation interactions.

• Improved drag-and-drop functionality in search results and folder views, adding intuitive "Drag All to Context" and "Copy Results" buttons and visual enhancements for easier item management.
• Enhanced chat file handling with better error checks, legacy format support, and more robust YAML frontmatter validation to ensure smoother file operations.
• Upgraded embedding provider management by allowing seamless switching without disabling features and by adding debouncing to UI actions for more reliable interactions.
• Refined message input behavior during response generation to keep the input and controls accessible, with clear feedback to avoid confusion.

Improve SystemSculpt provider integration for clearer settings and improved UI feedback.

• Added new user settings to enable/disable the SystemSculpt provider and specify a fallback when other providers fail.
• Updated the model selection interface to display the SystemSculpt option only when a valid license is present.
• Enhanced error handling in situations where provider configuration issues arise, including a refreshed primary action button style for better visual consistency.
• Streamlined the logic behind including the SystemSculpt provider, ensuring users experience more reliable and straightforward functionality.

Enhanced plugin experience with updated naming, improved license handling, advanced search features, and OpenAI API key validation.

• Updated terminology from "Cloud Models" to "AI Models" for clearer, consistent naming in models and settings.
• Enhanced license validation with more accurate feedback and error handling, ensuring users are immediately informed about their license status.
• Introduced a Global Search tab with a modern, OmniSearch-inspired design, enabling users to search across all notes with intuitive semantic hints.
• Upgraded the Similar Notes view by adding a tabbed interface and an empty state progress bar, improving navigation and visual feedback during searches.
• Added a validation button for OpenAI API keys in the embeddings section, allowing users to verify their key's validity directly.

Improve the plugin's UI and synchronization features with new modals, consistent styling, and enhanced backup and changelog functionality.

• Added a SystemSculpt models modal in settings that dynamically displays model count and handles errors during model loading.
• Introduced a custom provider modal alongside consistent component styling guidelines to streamline provider management.
• Developed a changelog tab that integrates GitHub releases and displays a timeline UI for easier feature tracking.
• Updated the manual backup process by replacing the browser prompt with a styled input field for a smoother user experience.
• Enhanced note synchronization with improved file-open and chat-loaded handling.
• Refined model selection in settings and modals by transitioning to a consistently used identifier.

Improve model detection accuracy and enhance chat favorites functionality

• Refined native model detection logic to ensure more accurate identification of models from API responses, providing more reliable model selection.
• Introduced chat favorites management in the UI, including toggling, sorting, and filtering options within the chat modal.

feat: Enhance embeddings service and API client with improved error handling and directory management

• Increased critical initialization timeout from 2 seconds to 7 seconds in main.ts to allow for longer startup processes.
• Updated EmbeddingsApiClient to handle API key requirements more flexibly for OpenAI and custom providers, ensuring better error messaging.
• Introduced a new method in EmbeddingsService to determine the effective embeddings directory based on provider settings, improving directory management.
• Added checks for custom provider configuration in EmbeddingsService to prevent embedding generation if the endpoint is not set.
• Enhanced utility functions for normalizing endpoint URLs and averaging embeddings, improving robustness and error handling.
• Updated UI components to reflect changes in settings and provide clearer user feedback regarding embedding configurations.

feat: Remove migration utilities and modal from storage management

• Deleted MigrationUtility and MigrationModal classes to streamline storage management. - Updated StorageManager import in main.ts to reflect the removal of migration-related components. - Adjusted storage initialization logic to eliminate migration checks, simplifying the startup process.

Enhanced search, embeddings, and settings for a more streamlined user experience

• Introduced a new OmniSearch modal that displays recent files when the search query is empty, with improved keyboard navigation and enhanced styling for better visibility. - Upgraded embeddings management with a refreshed UI that includes clear status feedback, metadata display (showing file dates and content previews), an integrated retry option for skipped files, and an auto-resume setting. - Improved embedding UI consistency by consolidating styles and refining progress indicators and layout for a cohesive design. - Updated embedding provider settings to support only OpenAI with a new default model, ensuring optimized performance and cost efficiency. - Enhanced the Similar Notes view by adding pagination and a Load More button for easier navigation through search results. - Introduced a user-friendly Backup & Restore modal for streamlined settings backup and restoration. - Streamlined embedding processes by automating repair functions, removing the need for manual interventions, and simplifying the overall interface.

Improve embedding configuration experience for users by adding new provider options and a dedicated settings tab

• Introduced support for multiple embedding providers (OpenAI, LM Studio, and custom endpoints) to offer flexible configuration. - Added dynamic settings in the embeddings tab, including a dropdown for provider selection with corresponding API key and endpoint options, as well as a confirmation prompt for recalculating embeddings. - Consolidated OpenAI API key and embedding settings into a dedicated embeddings tab for easier access and management.

refactor: Update references from Patreon to SystemSculpt license in various components and documentation

Added new prompt configuration modals and improved transcription handling for a streamlined user experience

• Introduced a modal to configure title generation prompts with options for presets and custom prompt files, accessible directly from the settings tab. - Added a tabbed modal for post-processing prompt editing, replacing the previous version to provide clearer options for preset selection or custom file usage. - Enhanced prompt operations with improved error handling and loading indicators during configuration. - Improved transcription handling by checking for existing markdown files before saving and updating notices for auto-paste settings.

New user-facing improvements in transcription settings and UI enhancements

• Added a Clean Output Only option in transcription settings for simplified output. - Implemented a countdown timer to auto-close the transcription notice after completion. - Introduced a microphone selection interface with refresh functionality in the settings. - Integrated custom transcription endpoint settings, allowing users to input endpoint URL, API key, and model selection.

Enhance transcription workflow with dynamic processing status updates for clear, real-time feedback

• Display clear processing status indicators (e.g., chunking, processing, uploading, success, error) during transcription - Provide real-time progress updates within the transcription modal for enhanced user awareness - Streamline transcription notices to offer direct status updates and improved error messaging

Improve transcription experience with expanded progress details and audio analysis tools

• Introduced an expanded progress modal for audio transcription featuring a new, user-friendly styling and toggle options for expanded/minimized views. - Enhanced the progress display to show detailed updates, including chunk processing status and error information. - Increased timeout settings to better support the handling of larger audio files and optimize SRT formatting. - Added audio diagnostic tools that analyze chunking performance and provide user-friendly reports and recommendations for optimal transcription settings.

Enhanced document handling, progress feedback, and citation display for improved user experience

• Improved document handling by streamlining file addition, processing, and Markdown conversion with integrated image extraction to enhance UI feedback. - Introduced a styled progress modal with a progress bar, steps, and interactive buttons to visually guide users during document processing. - Refined citation presentation by adjusting styling, removing content snippets, and enhancing web search integration for clearer and more effective citation displays.

Refactor TasksService, DirectoriesTabContent, TemplateManager, FileContextManager, and InputHandler for improved directory management and code cleanliness

• Cleaned up whitespace and formatting across multiple files for consistency. - Enhanced directory creation logic to utilize DirectoryManager where available, with fallbacks for legacy methods. - Updated event handling in DirectoriesTabContent to create directories only when explicitly requested. - Improved template loading and suggestion mechanisms in TemplateManager, ensuring better user experience. - Streamlined context file management in FileContextManager, including better error handling and logging. - Refined input handling in InputHandler, ensuring smoother user interactions and clearer code structure.

feat: Similar Notes enhancements and AI Agent rules

• Added context awareness to Similar Notes with ability to hide notes already in context - Implemented multi-vector search capabilities with support for chat content - Added hover preview functionality for similar notes without opening files - Enhanced search input with clear button and improved visual feedback - Implemented client-side token limit checks for search queries and content - Added comprehensive AI Agent rules for model handling and chat system architecture - Implemented automatic model switching when selected models are unavailable - Enhanced error handling with improved user feedback - Added documentation for code standards, component architecture, and error handling - Updated server URL for development mode - Improved event handling for chat messages to trigger updates in Similar Notes View

feat: system search and embeddings overhaul New Features * System Search modal with keyword and semantic options * Similar Notes ribbon view surfaces context from the active note * Text Input modal lets you run similarity searches on any phrase Improvements * Background embedding updates keep results fresh without manual action * State manager and watchdog recovery make embedding progress more reliable * Faster polling delivers near instant status and modal refresh UI * Refined Similar Notes view styling and loading behavior * Streamlined status indicator for clearer feedback Fixes * Removed obsolete files and scripts, cleaner repository and fewer errors * Fixed duplicate loading indicators and metadata inconsistencies

feat: Enhance web search functionality and update input handler styles

feat: Add favorites support for models

• You can now favorite your go-to models for quicker access - Toggle favorites directly from the model list - Filter to show only your favorites - See a friendly message when no favorites are set

feat: Add feedback submission link to settings tab UI Changes: - Added a clearly visible feedback link at the top of the SystemSculpt settings panel in Obsidian. - Styled the link using native Obsidian components to match user themes without custom CSS. New Features: - Users can now submit feedback, suggest features, or report bugs directly from within the plugin.

fix: Enhance recorder functionality

• Added new recorder component styles in recorder.css and integrated them into the main CSS. - Refactored RecorderService to improve audio recording handling and error management. - Enhanced UI for recording with a modal and improved status updates during recording. - Added support for auto-transcription of recordings and improved user notifications.

Enhance chat configuration, modal interactions, and template management for an improved user experience

• Introduced a Default Chat Presets modal that allows users to configure default AI models and system prompts for new chats. - Enhanced chat settings integration with a new default settings section and improved change handling. - Upgraded modal interfaces with standardized button styling, optional icon support, and refined interactions. - Added modals for submitting response improvements and saving AI responses as notes with folder suggestions. - Streamlined template management with improved model selection, final prompt editing, and better handling of empty note warnings. - Improved model validation and fallback mechanisms to enhance the overall response experience.

Improved chat interactions, chat settings with enhanced UI consistency and export/save features

• Streamlined chat editing and modal interactions with consistent button components and improved layout for a more intuitive experience. - Enabled local system prompts for easier customization within the chat interface. - Introduced a redesigned settings tab with clearly organized sections and new options for title generation, automatic transcription, and directory management. - Secured chat content display by escaping HTML characters and enhancing formatting with markdown callouts. - Enhanced file handling with confirmation prompts to prevent accidental overwrites, added markdown export capabilities, and incorporated a new Save as Note feature for seamless content saving.

fix: resolve startup errors and improve logging • Fixed critical startup error by removing conflicting .srt extension registration • Increased log storage limit from 10 to 500 logs to capture early initialization logs • Standardized log copying methods across error notices and settings for consistency • Moved error collector initialization to the beginning of the file for complete log capture • Improved error reporting with click-to-copy functionality for easier troubleshooting • Added comprehensive pattern matching to ensure all SystemSculpt logs are captured • Removed Obsidian version from log output • Added diagnostic logging to verify error collector initialization sequence • Added timeout handling for directory operations to prevent hanging • Fixed race condition between DirectoryManager and TemplateManager initialization

Implement startup special-case bug fixes and enhancements

• Fixed race condition between DirectoryManager and TemplateManager initialization, resolving UI components error messages. - Added comprehensive error reporting with click-to-copy functionality for easier troubleshooting. - Implemented system-wide log collection with a Copy Logs for Support button in the Advanced settings tab. - Enhanced error handling and logging to improve user feedback and diagnostics.

Enhance user-facing version display and model selection refresh functionality

• Display the current plugin version in the settings panel with a dedicated refresh button for easy update checks - Introduce a refresh button in the model selection modal to dynamically reload available models