Overview

Relevant source files

• Cargo.lock
• Cargo.toml
• apps/desktop2/src/components/chat/interactive.tsx
• apps/desktop2/src/components/chat/trigger.tsx
• apps/desktop2/src/components/main/body/calendars.tsx
• apps/desktop2/src/components/main/body/contacts/index.tsx
• apps/desktop2/src/components/main/body/events.tsx
• apps/desktop2/src/components/main/body/folders.tsx
• apps/desktop2/src/components/main/body/humans.tsx
• apps/desktop2/src/components/main/body/index.tsx
• apps/desktop2/src/components/main/body/shared.tsx
• apps/desktop2/src/components/main/sidebar/index.tsx
• apps/desktop2/src/components/main/sidebar/profile/index.tsx
• apps/desktop2/src/routes/app/main/_layout.index.tsx
• package.json
• packages/obsidian/package.json
• packages/tiptap/package.json
• packages/ui/package.json
• packages/utils/package.json
• pnpm-lock.yaml

This page introduces Hyprnote, an AI-powered meeting notes application focused on privacy, local-first operation, and real-time transcription and enhancement. It summarizes the system's purpose, high-level architecture, and core technical approach. For details on specific subsystems, see the Architecture and Core Features sections.

Purpose

Hyprnote is a desktop application for capturing, transcribing, and enhancing meeting notes using on-device AI. Its primary goals are:

• Local-first operation: All audio, transcription, and AI processing runs on the user's device by default.
• Privacy: No meeting data leaves the device unless the user explicitly enables cloud integrations.
• Real-time workflow: Users can record meetings, view live transcripts, and generate AI-enhanced summaries or chat with an AI assistant.
• Extensibility: The system is built as a modular Tauri application with a plugin-based backend and a monorepo workspace.

This page covers:

• What Hyprnote is and its core value proposition
• High-level system architecture and major components
• Technology stack and workspace structure

For details on:

• Plugin system architecture, see [2.2]
• Audio and speech processing, see [4]
• AI and language model integration, see [5]
• User interface, see [6]

Sources: Cargo.toml10-254 pnpm-lock.yaml1-700

What is Hyprnote?

Hyprnote is a desktop application for private, AI-powered meeting notes. It enables users to:

• Record meetings and capture audio from microphone and speakers
• Generate real-time transcripts using on-device speech-to-text (STT) models
• Enhance and summarize notes using local large language models (LLMs)
• Search, organize, and review meeting content in a unified interface

Key Properties:

Aspect	Description
Architecture	Tauri 2.8 (Rust backend, plugin system) + React 19.2 (TypeScript frontend)
Privacy Model	Local-first: all ML models (STT, LLM) run on-device by default
Distribution	Native binaries, multi-channel (staging, nightly, stable)
Platforms	macOS 14.2+, Windows (WebView2), Linux (WebKitGTK 4.1)
Licensing	Keygen license validation, offline fallback supported

Distribution channels are configured via tauri.conf.json, tauri.conf.nightly.json, and tauri.conf.stable.json.

Sources: Cargo.toml10-254 pnpm-lock.yaml1-700

Core Capabilities

Feature-to-Plugin Mapping

Below is a mapping of user-facing features to backend plugin code entities. This diagram bridges the natural language feature space to the code entity space.

Diagram: "Feature-to-Plugin and Code Entity Mapping"

Core Features and Corresponding Code Entities:

Feature	Code Entity / Plugin	Description
Real-time Transcription	tauri-plugin-listener, hypr-audio, tauri-plugin-local-stt, hypr-whisper	Captures audio, performs VAD, streams to STT, stores transcript
AI Enhancement	tauri-plugin-template, hypr-template, tauri-plugin-local-llm, hypr-llm	Renders Jinja templates, streams LLM output for note enhancement
AI Chat	tauri-plugin-local-llm, hypr-llm	Context-aware chat, OpenAI-compatible API
Data Persistence	tauri-plugin-db, hypr-db-user	Stores sessions, events, words, tags, etc.
Offline Operation	All above plugins/crates	All models and data local by default
Integrations	plugins/calendar-apple, plugins/calendar-google, crates/nango	Calendar and OAuth integrations

Sources: Cargo.toml10-254 pnpm-lock.yaml1-700 plugins/db plugins/listener plugins/local-stt plugins/local-llm crates/audio crates/whisper crates/llm crates/template crates/db-user

System Architecture

High-Level System Overview

The Hyprnote system is organized as a layered desktop application, with a React frontend, a Rust backend, and a modular plugin system. The following diagram shows the major systems and their relationships, using code entity names for traceability.

Diagram: "Overall System Architecture (Code Entity Mapping)"

Architectural Layers and Key Code Entities:

Layer	Code Entity	Description
Presentation	apps/desktop2/src	React UI, state management, routing
IPC Bridge	Tauri IPC, @tauri-apps/api	Type-safe command bridge
Plugin Layer	plugins/*	Rust plugins for audio, STT, LLM, DB, etc.
Core Crates	crates/*	Audio I/O, LLM inference, DB core, etc.
Data Layer	SQLite via crates/db-user	Local data persistence

Sources: Cargo.toml10-254 pnpm-lock.yaml1-700 apps/desktop2/src-tauri plugins/listener plugins/local-stt plugins/local-llm plugins/db crates/audio crates/whisper crates/llama crates/db-core crates/db-user

Technology Stack

Backend

Component	Technology / Crate	Code Reference
Application Framework	Tauri 2.8	Cargo.toml90-110
Language	Rust 1.90.0+	Cargo.toml20-21
Database	SQLite via libsql	crates/db-core crates/db-user
Audio Processing	cpal, cidre (macOS), wasapi (Windows)	crates/audio
STT Engine	Whisper.cpp via hypr-whisper	crates/whisper plugins/local-stt
LLM Engine	llama.cpp via hypr-llama	crates/llama plugins/local-llm
Templating	minijinja	crates/template
IPC Type Safety	specta, tauri-specta	Cargo.toml209-211
Observability	tracing, Sentry	Cargo.toml201-203

Frontend

Component	Technology / Package	Code Reference
UI Framework	React 19.2	pnpm-lock.yaml145-146
Routing	TanStack Router	pnpm-lock.yaml81-83
State Management	Zustand, Tinybase	pnpm-lock.yaml156-173
Editor	Tiptap 3.x	packages/tiptap/package.json23-41
UI Components	Radix UI	packages/ui/package.json19-37
Styling	Tailwind CSS	packages/ui/package.json61
Build Tool	Vite	pnpm-lock.yaml211-213
Type Safety	TypeScript 5.8+	pnpm-lock.yaml208-210

Platform-Specific Features

Feature flags in Cargo.toml enable GPU acceleration and platform-specific audio backends:

• macOS: Metal GPU, CoreML
• Windows: Vulkan, DirectML, OpenBLAS
• Linux: Vulkan, ALSA

Sources: Cargo.toml10-254 pnpm-lock.yaml1-700 packages/ui/package.json1-65 packages/tiptap/package.json1-62

Project Structure

The Hyprnote repository is a monorepo with both Rust (Cargo) and JavaScript/TypeScript (pnpm) workspaces.

Workspace Organization

Diagram: "Monorepo Structure (Code Entity Mapping)"

Key Code Entities by Domain:

Domain	Code Entity	Path
Audio	hypr-audio	crates/audio
LLM	hypr-llm	crates/llm
STT	hypr-whisper	crates/whisper
Database	hypr-db-user	crates/db-user
Templates	hypr-template	crates/template
UI Components	@hypr/ui	packages/ui
Editor	@hypr/tiptap	packages/tiptap
Utilities	@hypr/utils	packages/utils

Sources: Cargo.toml10-254 pnpm-lock.yaml1-700 packages/ui/package.json1-65 packages/tiptap/package.json1-62 packages/utils/package.json1-30

Build and Deployment

Release Channels

Channel	Config File	Identifier	Update Endpoint
Staging	tauri.conf.json	com.hyprnote.dev	Not updatable
Nightly	tauri.conf.nightly.json	com.hyprnote.nightly	Nightly CDN
Stable	tauri.conf.stable.json	com.hyprnote.stable	Stable CDN

Each channel uses a different bundle identifier and update endpoint. LLM and STT models are bundled or downloaded as resources.

Build Pipeline

Diagram: "Build, Test, and Release Pipeline"

Build steps include:

• Frontend build with Vite (pnpm build)
• Rust backend build with Tauri (tauri build)
• Platform-specific packaging and code signing
• Artifact upload and release publish via CrabNebula Cloud

Sources: Cargo.toml10-254 pnpm-lock.yaml1-700

Development Environment

Prerequisites

Tool	Version	Purpose
Rust	1.90.0+	Backend compilation
Node.js	18+	Frontend tooling
pnpm	Latest	Package management
Python	3.x	Build scripts
Poetry	Latest	Python dependencies
protoc	25.7	Protocol buffer compilation

Platform-Specific:

• macOS: Xcode Command Line Tools
• Windows: Visual Studio Build Tools, MSVC
• Linux: libwebkit2gtk-4.1-dev, libasound2-dev, build essentials

Development Workflow

Development Commands:

Testing:

Type Safety Workflow:

1. Define Rust command in apps/desktop/src-tauri/src/commands.rs with #[specta::specta]
2. Add to collect_commands![] in lib.rs252-259
3. Run cargo test -p desktop export_types to generate bindings
4. Import from @/types in TypeScript: import { commands } from "@/types"

Sources: apps/desktop/package.json6-16 apps/desktop/src-tauri/src/lib.rs263-279 apps/desktop/src-tauri/tauri.conf.json7-11

---

Summary: Hyprnote is a sophisticated local-first desktop application that combines real-time audio processing, speech-to-text transcription, and AI-powered content enhancement. Built with Tauri and React, it leverages a plugin-based architecture for modularity and uses local ML models (Whisper, Llama.cpp) to ensure privacy. The system supports multiple platforms with GPU acceleration and provides optional cloud integrations while maintaining offline-first operation.