feat(media): add UI prototype wireframes to media gallery - classify .wireframe.json diagrams as wireframe category with canvas preview - link wireframe companion files to note references and deduplicate preview PNGs - add UI Prototypes category filter and single-line sidebar layout - clean up unused imports and lint warnings

CHANGELOG.md

@@ -6,6 +6,24 @@ All notable documentation and user-facing behavior changes are tracked in this f
 
 ### Added
 
+- Added **Dual-Stream Audio Capture & Meeting Recording**:
+  - Captures microphone and Windows desktop/system audio loopback simultaneously using Web Audio API AudioContext mixer (`src/utils/audioMixer.js`).
+  - Supports 3 capture modes: Meeting Mode (Mic + System Audio), Microphone Only, and System Audio Only.
+  - Live audio waveform visualizer canvas and recording timer with pause/resume support (`AudioRecorderBar.jsx`).
+  - Keyboard shortcut `Alt + V` to toggle audio/meeting recorder in the editor toolbar.
+- Added **Speech-to-Text (STT) Transcription Service & Model Settings**:
+  - Offline on-device transcription via Local ONNX Whisper (`whisper-tiny.en`, `whisper-base.en`, `whisper-small`).
+  - High-speed cloud transcription via Groq (`whisper-large-v3`) and OpenAI (`whisper-1`).
+  - Dedicated **Speech-to-Text** subtab in AI Settings (`AISettings.jsx`) for selecting engines, local models, capture defaults, and languages.
+  - Transcripts saved as structured companion `.json` files alongside audio in `media/audio/*.json` with timestamps, speaker attributions, key points, and action items.
+- Added **Landing Page Quick Capture Bar**:
+  - Integrated header toolbar on the landing dashboard (`LandingQuickCaptureBar.jsx`) for one-click Screen Snip, Desktop Recording, and Audio/Meeting Recording.
+  - Saves captured media directly to workspace media storage with zero note modifications.
+  - Notification toast provides direct 1-click `[Open Gallery]` action.
+- Added **Workspace Media Gallery Unused Media Filter & Audio Player**:
+  - Physical disk scanner (`media:list-disk-assets` IPC) discovers files in `media/`, `assets/`, and `images/`.
+  - New `⚠️ Unused / Orphans` filter tab and header stats indicator isolating unreferenced media files (`referenceCount === 0`).
+  - Interactive audio player preview (`AudioPlayerPreviewItem`) and orphan action buttons (`Copy Markdown Embed`).
 - Added **Multiple Developer-Friendly Fonts Support** (`View → Font`).
   - Choose between 5 curated IDE/developer-friendly typefaces: Inter (Default), JetBrains Mono, Fira Code, Cascadia Code, and Source Code Pro.
   - Changes cascade in real time across all app labels, navigation items, buttons, dialogs, and controls.

README.md

@@ -34,6 +34,9 @@ Notely is built with Electron + React and is designed for project notes, meeting
   - **Regex search** with validation and pattern matching for advanced queries.
   - **Code-aware search** to find patterns inside code blocks only.
 - Insert common Markdown snippets from the toolbar.
+- Record audio notes and full meetings with dual-stream audio capture (Microphone + Windows System Audio loopback) with automatic Speech-to-Text (STT) transcription powered by offline local ONNX Whisper or ultra-fast cloud providers (Groq whisper-large-v3, OpenAI whisper-1).
+- Quick capture directly from the landing dashboard with one-click screen snip, desktop recording, and meeting audio capture that saves directly to workspace media without altering notes until you embed them.
+- Explore and audit all diagrams, media, audio, and documents in the **Workspace Diagrams & Media Gallery**, featuring physical disk scanning, unreferenced orphan detection, and an **Unused Media** filter with audio playback and transcript preview.
 - Edit Markdown tables inline with a focused grid editor (row/column add/remove, alignment controls, and compact action chips).
 - Browse, annotate, optimize, and manage linked media.
 - Open note files in VS Code or the system default app.

THIRD_PARTY_NOTICES.txt

@@ -1,7 +1,7 @@
 THIRD-PARTY SOFTWARE NOTICES AND INFORMATION
 
 Notely incorporates components from the projects listed below.
-This file was generated automatically on 2026-09-24 by
+This file was generated automatically on 2026-09-26 by
 scripts/generate-notices.cjs. Do not edit by hand.
 
 Total third-party packages: 189

docs/ai/index.md

@@ -29,7 +29,12 @@ Notely features a modular, local-first AI platform designed around private data
 - Runs entirely offline using a local ONNX runtime for `BGE-small-en-v1.5` 384-dimensional dense vectors, with optional fallback to cloud embedding APIs.
 - Background worker process handles queue processing and debounced note indexing to prevent UI thread latency.
 
-### 4. In-App Tools Catalog & MCP Diagnostics
+### 4. Speech-to-Text (STT) Transcription Pipeline
+- Transcribe voice memos and meeting discussions into Markdown and companion JSON transcript files.
+- Runs offline via on-device WebAssembly/ONNX Whisper models (`whisper-tiny.en`, `whisper-base.en`, `whisper-small`), with optional sub-second cloud transcription via Groq or OpenAI Whisper.
+- Synchronized transcript view with key points and action items embedded into notes and the Media Gallery.
+
+### 5. In-App Tools Catalog & MCP Diagnostics
 - **MCP Tools Page** (`Ctrl/Cmd + Shift + M`): Interactive catalog to view tool definitions, test inputs, and verify outputs.
 - **AI Health / Diagnostics**: Real-time telemetry dashboard monitoring MCP connection status, client sessions, request volume, and error diagnostics.
-- **AI Settings** (`Ctrl/Cmd + Shift + ,`): Configure API providers, embedding options, and Knowledge Graph extraction confidence thresholds.
+- **AI Settings** (`Ctrl/Cmd + Shift + ,`): Configure API providers, embedding options, STT engines, and Knowledge Graph extraction confidence thresholds.

docs/ai/setup.md

@@ -38,7 +38,16 @@ Relationship extraction and entity graph generation:
 
 ---
 
-## 4. SQLite Database Locality
+## 4. Speech-to-Text (STT) Engine Setup
+
+Configure meeting and audio transcription inside **AI → AI Settings → Speech-to-Text**:
+- **Local ONNX Whisper (Offline)**: Download `whisper-tiny.en` (~40MB), `whisper-base.en` (~140MB), or `whisper-small` (~460MB) for private, zero-latency on-device transcription via WebAssembly/ONNX Runtime.
+- **Cloud Whisper (Groq / OpenAI)**: Sub-second cloud transcription using your saved Groq or OpenAI API key.
+- **Audio Capture Preference**: Set default recording mode to `Meeting: Mic + System Audio`, `Microphone Only`, or `System Audio Only`.
+
+---
+
+## 5. SQLite Database Locality
 
 All AI databases are workspace-scoped and stored inside the hidden `{workspace}/.notes-app/` folder to keep your data local and portable:
 1. `ai-embeddings.db`: Stores chunk text, line mappings, content hashes, and indexing queues.

docs/architecture.md

@@ -135,8 +135,8 @@ The following diagram shows the full request path from the React UI through each
 flowchart TD
     subgraph Renderer["Renderer Process (React / Vite)"]
         direction LR
-        ACP["AIChatPanel"] & AIS["AISettings"] & EBP["EmbeddingsPage"] & KGV["KnowledgeGraph"]
-        UAI["useAIAssistant hook"]
+        AIS["AISettings"] & MCP["MCPToolsPage"] & MED["WorkspaceDiagramsMediaPage"] & KGV["KnowledgeGraph"]
+        ARB["AudioRecorderBar"] & LQC["LandingQuickCaptureBar"]
     end
 
     subgraph Preload["Preload Bridge (preload.cjs)"]

docs/data-sync-security.md

@@ -45,13 +45,12 @@ Tips:
 
 If you do not share notes between devices, you can ignore this section.
 
-## 4. AI Features for Daily Use
+## 4. AI & MCP Integration for Daily Use
 
-1. Open **AI -> AI Settings**.
-2. Add the sign-in details for the AI service you want to use.
-3. Use AI chat, smarter search, and related-note features as needed.
-
-If the service you chose cannot do something, Notely shows a warning in AI Settings.
+1. Open **AI -> AI Settings** (`Ctrl/Cmd + Shift + ,`).
+2. Configure your preferred Speech-to-Text engine (Local ONNX Whisper or Cloud Groq/OpenAI) and API keys.
+3. Connect external AI assistants (Google Antigravity, Claude Desktop, Cursor) directly to Notely via the built-in MCP server on port `3700`.
+4. Use hybrid semantic search and knowledge graph discovery locally and securely.
 
 ## 5. When to Use Workspace Graph
 

docs/feature-availability.md

@@ -19,14 +19,17 @@ The matrix below details which features run entirely offline, which require loca
   { feature: 'Help Center', available: true, setup: 'No', internet: false },
   { feature: 'Tasks Dashboard', available: true, setup: 'No', internet: false },
   { feature: 'Version History (Git)', available: true, setup: 'No', internet: false },
-  { feature: 'Media Library', available: true, setup: 'No', internet: false },
+  { feature: 'Media Library & Disk Scanner', available: true, setup: 'No', internet: false },
   { feature: 'Embedded Terminal', available: true, setup: 'No', internet: false },
-  { feature: 'Screen Capture (Windows)', available: true, setup: 'No', internet: false },
+  { feature: 'Screen Capture & Snipping', available: true, setup: 'No', internet: false },
+  { feature: 'Screen Video Recording & Overlay', available: true, setup: 'No', internet: false },
+  { feature: 'Dual-Stream Audio & Meeting Recording', available: true, setup: 'No', internet: false },
+  { feature: 'Local Speech-to-Text (ONNX Whisper)', available: true, setup: 'Download Model (~40-460MB)', internet: false },
+  { feature: 'Cloud Speech-to-Text (Groq / OpenAI)', available: false, setup: 'Configure API Key in AI Settings', internet: true },
+  { feature: 'Model Context Protocol (MCP) Server', available: true, setup: 'No (Runs on port 3700)', internet: false },
   { feature: 'Mermaid Diagrams', available: true, setup: 'No', internet: false },
   { feature: 'Excalidraw Diagrams', available: true, setup: 'No', internet: false },
-  { feature: 'Workspace Graph', available: true, setup: 'No', internet: false },
-  { feature: 'Sync with other devices', available: false, setup: 'Pair Trusted Devices', internet: 'Local network' },
-  { feature: 'AI Chat & Rewriting', available: false, setup: 'Setup AI Provider', internet: true },
-  { feature: 'Meaning-based Search', available: false, setup: 'Setup AI Provider', internet: true },
-  { feature: 'Graph Clustering', available: false, setup: 'Setup AI Provider', internet: true }
+  { feature: 'Workspace Graph & Neural Extraction', available: true, setup: 'No (Local GLiNER2 ONNX)', internet: false },
+  { feature: 'Semantic Hybrid Search (BGE)', available: true, setup: 'No (Local BGE ONNX)', internet: false },
+  { feature: 'Sync with other devices (P2P)', available: false, setup: 'Pair Trusted Devices', internet: 'Local network' }
 ]" />

docs/feature-reference.md

@@ -79,9 +79,12 @@ During transfer:
 
 Open via **Workspace -> Diagrams & Media Gallery** (`Ctrl/Cmd + Alt + M`) or Command Palette:
 - Catalogs all used diagrams (inline Mermaid, Draw.io, Excalidraw), images, videos, audio, and PDF documents in the workspace.
+- **Physical Disk Scanner & Orphan Detection**: Scans workspace folders (`media/`, `assets/`, `images/`) to identify files present on disk that have zero note references.
+- **Unused Media Filter**: Quickly isolate unreferenced assets with the `⚠️ Unused / Orphans` filter tab.
+- **Audio & Media Preview**: Embedded audio players, video thumbnails, and transcript drawer with key points and action items.
 - Reuses Knowledge Graph layout aesthetics with collapsible category filters, search, and stat indicators.
 - Inspect details for any item with high-res/diagram live previews, file paths, and exact note references with line numbers and snippets.
-- Direct "Open Note" navigation into the editor.
+- Direct "Open Note" navigation and "Copy Markdown Link" actions.
 
 ## 2. Editor and Writing Experience
 
@@ -328,6 +331,26 @@ Notely supports area-based screen capture and full screen/window video recording
   - **Auto-Save & Linking**: Saves `.webm` files under `media/recordings/` and inserts `![Screen Recording](media/recordings/*.webm)` into the note.
   - **Note Preview & Modal Player**: Renders video links as interactive thumbnail cards with `▶ Play Video` badges and frame decoding. Clicking opens a full-screen playback modal with Download and Copy Path options.
 
+### Audio & Meeting Recording with Speech-to-Text
+
+Record voice notes or entire conference calls (Zoom, Google Meet, Teams, YouTube) directly into Notely with synchronized audio and transcription.
+
+- **Dual-Stream Audio Capture**:
+  - **Meeting Mode**: Simultaneously captures your local microphone and Windows desktop/system audio loopback, combining them into a single stereo track with Web Audio API mixer.
+  - **Microphone Only**: Dedicated voice memo recording.
+  - **System Audio Only**: Captures speakers/audio loopback without mic.
+- **Speech-to-Text (STT) Transcription**:
+  - **Local ONNX Whisper (Offline)**: Runs locally in-app via WebAssembly/ONNX runtime (`whisper-tiny.en`, `whisper-base.en`, `whisper-small`). No internet connection or cloud API keys required.
+  - **Cloud Whisper**: High-speed cloud transcription using Groq (`whisper-large-v3`) or OpenAI (`whisper-1`) with API keys configured in AI Settings.
+- **Transcript Storage**:
+  - Audio files are stored under `media/audio/*.webm`.
+  - Transcripts are saved alongside audio in `media/audio/*.json` containing segmented timestamps, speaker attributions, key points, and action items.
+- **Workflow Differences**:
+- **Media Gallery Integration & Transcripts Filter**:
+  - The **Diagrams, Media & PDFs** gallery (`Workspace -> Diagrams, Media & PDFs`) includes a dedicated **Transcripts** filter category alongside Diagrams, Images, Videos, Audio, PDFs, and Documents.
+  - Selecting a transcript in the gallery opens a rich inspector showing the AI executive summary, timestamped speaker segments, and a one-click button to copy the full transcript text.
+  - **On-Demand Transcript Generation**: Any existing audio or video recording in the Media Gallery has a **Generate AI Transcript** button (`✨`). Clicking it decodes the media track, resamples it to 16kHz mono Float32, runs Whisper (Local ONNX or Cloud Groq/OpenAI based on your AI Settings), and saves a companion transcript JSON file in `media/audio/`.
+
 ### Media preview tools
 
 Use zoom and media-aware preview controls to inspect assets.
@@ -373,40 +396,27 @@ This helps recover from aggressive crops or accidental edits.
 - Excalidraw previews have their own right-click actions.
 - If the diagram started from an image, you can switch back to the original image later.
 
-## 9. AI Assistance
-
-### AI settings
-
-Set up AI services and sign-in details in **AI -> AI Settings**.
-
-You can set up:
-
-- a writing assistant service for chat and rewriting
-- a separate service for smarter search and graph features
-
-### AI chat and commands
-
-Use AI for writing support, note understanding, and quick content actions.
-
-### Semantic features
-
-When AI search data is turned on, Notely can find related notes by meaning, not just exact words.
-
-### What each AI service can do
-
-Not every AI service supports every feature.
+## 9. AI & Machine Intelligence
 
-Notely shows capability warnings in settings when a selected provider cannot run a feature.
+### Privacy-First Architecture
 
-You can also choose which AI model to use in AI settings.
+Notely features a local-first, privacy-conscious AI architecture. Rather than relying on cloud-dependent chat sidebars, AI functionality is centered around:
+1. **Model Context Protocol (MCP) Server**: Local standard interface on port `3700` (`http://127.0.0.1:3700/mcp` and `/sse`) for external AI assistants (Google Antigravity, Claude Desktop, Cursor).
+2. **Local Speech-to-Text (STT)**: On-device Whisper transcription via WebAssembly/ONNX, plus optional cloud transcription.
+3. **Local Vector Embeddings & Hybrid Search**: SQLite FTS5 combined with local `BGE-small-en-v1.5` dense embeddings.
+4. **Offline Knowledge Graph**: Neural zero-shot entity and relation extraction powered by local `gliner2-multi-v1-onnx`.
 
-### AI operation feedback
+### AI Settings & Configuration
 
-Longer AI actions show progress and completion messages so you know something is happening.
+Configure AI services and endpoints in **AI -> AI Settings** (`Ctrl/Cmd + Shift + ,`):
+- **LLM Provider Setup**: Configure Google Gemini, Groq, or OpenAI / OpenAI-compatible endpoints with custom Base URLs for cloud STT and external workflows.
+- **Speech-to-Text (STT) Settings**: Choose between Local ONNX Whisper (`whisper-tiny.en`, `whisper-base.en`, `whisper-small`) and Cloud Whisper (Groq `whisper-large-v3`, OpenAI `whisper-1`).
+- **Feature Toggles**: Toggle pattern learning, embedding generation, relationship discovery, and adjust the neural entity extraction confidence threshold slider.
 
-### How recent the AI search data is
+### MCP Diagnostics & Interactive Testing
 
-Notely shows whether its AI search data is fresh or getting old.
+- **MCP Tools Catalog** (`Ctrl/Cmd + Shift + M`): Inspect all 7 enterprise tools (`search`, `read_note`, `edit_note`, `manage_tasks`, `manage_diagrams`, `workspace_overview`, `git_control`), run live tool test calls, and inspect JSON payloads.
+- **MCP Telemetry & Diagnostics**: View live connection status, active client sessions, request rates, execution latency, and error logs.
 
 ## 10. Peer-to-Peer Sync
 

docs/keyboard-shortcuts.md

@@ -55,6 +55,8 @@ Use shortcuts to navigate and edit notes quickly across Notely.
 | Toggle Outline Panel | `Ctrl/Cmd + Alt + L` | Editor |
 | Open Reference Note | `Ctrl/Cmd + Shift + K` | Editor |
 | Insert Reference Link | `Ctrl/Cmd + Shift + L` | Editor |
+| Capture Screen Snip | `Ctrl/Cmd + Shift + S` | Editor |
+| Toggle Audio / Meeting Recorder | `Alt + V` | Editor |
 
 ## Search & Find Panel
 

docs/settings-reference.md

@@ -154,6 +154,19 @@ Open **AI -> AI Settings**.
 
 Use lower temperature for predictable output. Use higher temperature for brainstorming or variation.
 
+### Speech-to-Text (STT) setup
+
+Open the **Speech-to-Text** tab in AI Settings:
+
+- **Transcription Engine**:
+  - `Local ONNX Whisper (Offline, In-Browser / WebAssembly)`: default offline engine; runs private on-device transcription.
+  - `Groq Cloud Whisper`: utilizes Groq's high-throughput `whisper-large-v3` for sub-second transcriptions using your configured Groq API key.
+  - `OpenAI Cloud Whisper`: utilizes OpenAI's `whisper-1` model using your configured OpenAI API key.
+- **Local Whisper Model**: Select from `whisper-tiny.en` (~40MB, fastest), `whisper-base.en` (~140MB, balanced), or `whisper-small` (~460MB, multilingual).
+- **Default Audio Capture Mode**: Choose default mode when launching recorder (`Meeting: Mic + System Audio`, `Microphone Only`, `System Audio Only`).
+- **Primary Spoken Language**: Specify expected audio language (or `Auto-Detect`).
+- **Auto-transcribe toggle**: Automatically triggers Speech-to-Text generation as soon as an audio or meeting recording is stopped.
+
 ### Data and privacy controls
 
 - Shows where AI-related app data is stored on your device