Liberty

KO | EN

gitlite — search

#javascript #python #react #android #rust #deep-learning #linux #docker #macos #nodejs #vue #awesome-list #ios #openai

(★ 10)

Liberty 是一款面向桌面端的会议音视频处理工作台，围绕“本地转写、AI 总结、结果整理”这一完整链路设计。项目基于 Tauri 2、React、TypeScript、Rust 与 Python 构建，重点强调本地可运行、桌面端工作流完整以及配置集中管理。

Liberty 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

<h1 align="center">Liberty</h1>

<p align="center">
  A desktop workspace for meeting media processing.
</p>

<p align="center">
  Local transcription · Speaker diarization · AI summarization · Result organization
</p>

[English](./README.md) | [简体中文](./README.zh-CN.md)

Liberty is a local-first desktop application for meeting media processing. The current frontend uses React, the desktop shell uses Tauri 2, native services are implemented in Rust, and local transcription runs through a managed Python 3.10.17 runtime with a FunASR runner. It can process meetings fully through local SQLite and the managed runtime, while still keeping an optional remote backend mode.

![Liberty Create Job](https://raw.githubusercontent.com/whalesky-labs/Liberty/main/docs/images/ScreenShot_2026-05-21_192125_725.png)

![Liberty Desktop Companion](https://raw.githubusercontent.com/whalesky-labs/Liberty/main/docs/images/ScreenShot_2026-05-21_192138_208.png)

![Liberty Pet Store](https://raw.githubusercontent.com/whalesky-labs/Liberty/main/docs/images/ScreenShot_2026-05-21_192159_758.png)

## Current Capabilities

- Create meeting jobs through the desktop file picker with local audio and video files.
- In local mode, process one local file with the managed Python 3.10.17 runtime, FunASR runner, and ffmpeg.
- Track transcription, speaker diarization, task logs, processing duration, failure reasons, and retries.
- Manage OpenAI-compatible models, summary templates, AI summary windows, repeated summary runs, and the active summary result.
- Review transcripts, filter by speaker, rename speakers, open meeting notes, and work from a dedicated result workspace.
- Export transcript TXT, notes Markdown, bundled Markdown, and formal DOCX meeting minutes.
- Manage meeting members with Excel import/export, department, sort order, and recorder metadata.
- Switch Chinese/English UI, automatic/light/dark theme, transparent/tinted glass style, and accent color.
- View diagnostics for platform matrix, database schema version, runtime status, security baseline, and exported desktop-pet diagnostic logs.
- Use the desktop companion, 255-level growth, LP wallet, Pet Store, inventory, daily free blind box, daily check-in, make-up check-in, gift boxes, redeem center, item detail window, and native desktop pet rendering.

## Runtime Modes

### Local Mode

When `backendUrl` is empty, the app uses local SQLite and local Tauri commands. If the user has not manually configured a Python path, the app automatically installs or repairs the managed runtime based on runtime status.

Local mode includes:

- Python, ffmpeg, download sources, and model endpoints described by `runtime-manifest.json`.
- The local transcription runner under `python/funasr-runner/`.
- Job creation, execution, retry, and log synchronization in `apps/desktop/src-tauri/src/local_jobs.rs`.
- Runtime installation, download-source selection, validation, warmup, and logs in `apps/desktop/src-tauri/src/local_runtime/`.
- SQLite storage for jobs, transcripts, AI summaries, members, settings, and pet data.

Current local job execution processes one file with a local path. In local mode, selecting multiple files keeps the last selected file.

### Remote Mode

When `backendUrl` is set, the frontend uses `shared/services/remote/meetingApi.ts` to call the remote meeting API. Remote mode keeps job creation, listing, detail, and retry entry points, but the managed local runtime is not required before use.

### AI Summary Pipeline

AI summarization is not automatic after transcription. From the workbench, the user opens the AI summary window, chooses the model, template, speaker/timestamp options, generates a summary, and saves a summary run.

AI requests are sent by the Rust `local_ai` module to an OpenAI-compatible endpoint. Model API keys are stored through the system credential store: macOS Keychain or Windows Credential Manager.

### Pet Pipeline

The pet pipeline is a local companion system outside the core meeting flow. Current pet rules are centralized in the pet system notes: real work is the main growth source, LP is a local reward point, food grants fixed growth, and daily blind boxes, daily check-ins, make-up tickets, gift boxes, and redeem keys are local benefit or operations entry points. App startup tries to sync desktop pet state, but pet loading failure must not block the main window.

See [docs/pet-system.md](./docs/pet-system.md) for the current implementation notes.

## Technical Architecture

| Layer | Current implementation |
| --- | --- |
| Desktop shell | Tauri 2 |
| Frontend | React 19 + TypeScript + Vite |
| Routing | In-repo lightweight RouterContext |
| Native services | Rust, Tauri commands, SQLite, system credentials, DOCX/XLSX processing |
| Local transcription | Python 3.10.17 + FunASR runner + ffmpeg |
| Local storage | SQLite through bundled `rusqlite` |
| AI interface | OpenAI-compatible Chat Completions |
| Desktop pet rendering | macOS AppKit private API + Windows GDI/Win32 |

## Project Structure

```text
.
├─ apps/
│  └─ desktop/
│     ├─ src/
│     │  ├─ app/                 React shell, navigation, lightweight routing
│     │  ├─ assets/              Frontend images and store assets
│     │  ├─ features/            Jobs, AI, settings, members, pet, store pages
│     │  └─ shared/              i18n, components, services, types, global styles
│     └─ src-tauri/
│        ├─ capabilities/        Tauri window permission boundaries
│        ├─ resources/           Runtime manifest, DOCX template, pet resources
│        ├─ src/                 Rust commands, database, runtime, exports, pet
│        └─ tauri.conf.json      Tauri config and bundled resources
├─ python/
│  └─ funasr-runner/             Local transcription runner and Python deps
├─ scripts/                      Startup, runtime preparation, release checks
├─ docs/
│  ├─ architecture/              Architecture and release readiness docs
│  ├─ images/                    README screenshots
│  ├─ ai/                        AI process docs and migrated historical records
│  └─ pet-system.md              Current pet system notes
├─ Cargo.toml                    Rust workspace
├─ package.json                  pnpm workspace scripts
└─ pnpm-workspace.yaml
```

## Main Screens

- `New Job`: choose local media, title, language, speaker diarization, and hotwords.
- `Jobs`: review job status, processing time, file metadata, detail, retry, and delete actions.
- `Job Detail`: inspect input, status, progress, logs, failure reason, and workbench entry.
- `Workbench`: review transcripts, filter by speaker, rename speakers, open AI summary and notes windows, export results.
- `Models` / `Model Editor`: maintain OpenAI-compatible model configuration.
- `Templates` / `Template Editor`: maintain AI summary templates.
- `Members` / `Member Editor`: maintain members, departments, sort order, and recorder metadata with Excel import/export.
- `Settings`: appearance, locale, managed runtime, manual Python, ASR parameters, remote backend, diagnostics.
- `Pet Center`: view pet level, cumulative growth, stage, events, desktop behavior, and interactions.
- `Pet Store`: view LP, catalog, inventory, equipment, food/tool usage, and item details.
- `Daily Blind Box`: open up to 10 free local boxes per day; rewards come from the Pet Store, excluding pets.
- `Daily Check-In`: view streaks, reward calendar, history, claim rewards, or use a make-up ticket.
- `Redeem Center`: enter locally verified redeem keys for LP, growth value, or pet items, and view device-local redemption history.

## Local Data

SQLite currently stores:

- App settings and managed runtime state.
- Jobs, input files, transcript segments, job events, and process-log snapshots.
- AI models, summary templates, summary runs, and active summary selection.
- Meeting members, departments, sort order, and recorder flag.
- Pet profile, desktop behavior settings, growth events, stage cosmetics, and level snapshots.
- LP wallet, inventory, economy ledger, milestone counters, daily blind box history, check-in records, daily free-claim state, and redemption history.

The schema is created in `apps/desktop/src-tauri/src/local_db/schema.rs`; migration versioning is maintained in `infrastructure/migrations.rs`.

## Development Commands

Install dependencies:

```bash
pnpm install
```

Start the frontend:

```bash
pnpm desktop:dev:web
```

Start the Tauri desktop app:

```bash
pnpm desktop:tauri dev
```

Build the frontend:

```bash
pnpm desktop:build:web
```

Build the desktop app:

```bash
pnpm desktop:tauri build
```

Run the full check:

```bash
pnpm check
```

`pnpm check` runs frontend typecheck/build, version/platform/security checks, Rust fmt/test/clippy.

## Supported Platforms

| Platform | Rust target | Validation level | Runtime backend |
| --- | --- | --- | --- |
| macOS Apple Silicon | `aarch64-apple-darwin` | Primary | FunASR |
| macOS Intel | `x86_64-apple-darwin` | Primary | FunASR |
| Windows x64 | `x86_64-pc-windows-msvc` | Primary | FunASR |
| Windows x86 | `i686-pc-windows-msvc` | Extended | sherpa-onnx |

## Documentation

- [Pet system notes](./docs/pet-system.md)
- [Enterprise desktop architecture](./docs/architecture/enterprise-desktop-architecture.md)
- [Release readiness](./docs/architecture/release-readiness.md)

## Notes

- The frontend uses React/TSX.
- Local mode uses the client-downloaded managed runtime by default, with manual Python override available in Settings.
- Local execution depends on ffmpeg, the Python runner, and model resources; incomplete runtime state is marked `repair_required`.
- Runtime download sources are controlled by `runtime-manifest.json` and the Environment & Models selection in Settings; sources must be real reachable URLs, not placeholder assets.
- Formal DOCX meeting minutes use `apps/desktop/src-tauri/resources/templates/meeting-minutes.docx`.
- The pet system is a local reward and companion system. It does not include real-money payments, top-ups, trading, or leaderboards; the daily blind box is a free benefit that does not consume LP, sell attempts, or use paid probability drops.

## License

This project is licensed under the MIT License. See [LICENSE](./LICENSE) for details.

Liberty

Liberty 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Liberty

Liberty 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README