jarvis - GitLite

KO | EN

gitlite — search

#ruby #proxy #openai #ai #github #windows #python #macos #csharp #cloud-native

jarvis

(★ 28)

A minimalist local voice assistant built in C++ using Whisper.cpp, llama.cpp, and Piper TTS.

파일 탐색기

ZIP 다운로드

- brain.hpp
- ollama.hpp
- recorder.hpp
- stt.hpp
- tts.hpp
- brain.cpp
- main.cpp
- recorder.cpp
- stt.cpp
- tts.cpp
buildMe.sh
LICENSE
README.md
setup.sh
testMe.sh

# CDN으로 사용하기

jsDelivr

jsDelivr는 공개 GitHub 리포지토리를 별도 설정 없이 CDN으로 즉시 서빙합니다. 버전과 파일을 고르면 웹페이지에 바로 붙일 수 있는 링크와 예시 코드가 만들어집니다.

버전

파일

링크

예시

명령어 용어집

이 문서에서 사용된 명령어를 모아봤습니다. 낯선 명령어가 있다면 펼쳐서 확인해보세요.

🔍

`piper-tts`

설명 보기 ▼

이 명령은 `piper`의 별칭입니다.

tldr piper

자세한 내용은 원본 명령을 참고하세요:

🔍

`piper`

설명 보기 ▼

빠르고 로컬에서 동작하는 신경망 기반의 텍스트 음성 변환 시스템.
<https://rhasspy.github.io/piper-samples>에서 음성 모델을 시도해보고 다운로드하세요.
더 많은 정보: <https://github.com/OHF-Voice/piper1-gpl>.

echo {{Thing to say}} | piper -m {{path/to/model.onnx}} -f {{outputfile.wav}}

텍스트 음성 변환 [m]모델을 사용하여 WAV [f]파일 출력(모델 경로에 대한 config 파일이 있을 경우):

echo {{Thing to say}} | piper -m {{path/to/model.onnx}} -c {{path/to/model.onnx.json}} -f {{outputfile.wav}}

[m]모델과 JSON [c]설정 파일을 지정하여 WAV [f]파일 출력:

echo {{Warum?}} | piper -m {{de_DE-thorsten_emotional-medium.onnx}} --speaker {{1}} -f {{angry.wav}}

여러 명의 화자가 있는 음성에서 특정 화자를 ID 번호로 선택:

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

![Platform](https://img.shields.io/badge/platform-Linux-blue)
![Made with C++](https://img.shields.io/badge/made%20with-C%2B%2B-00599C)
![Arch BTW](https://img.shields.io/badge/tested%20on-Arch%20Linux-blueviolet)

> [!WARNING]
> **🚨 DEPRECATED: This project is no longer maintained. 🚨**
>
> Jarvis was a fantastic first iteration, but he has been officially retired and replaced by a vastly improved, completely unified offline architecture.
>
> 👉 **Please check out the new and improved project here: [Sarah](https://github.com/almimony75/sarah)** 👈

> "A powerful, conversational AI assistant inspired by Jarvis—brought to life by sleepless nights and caffeine."

Welcome to **Jarvis**: your locally-running, privacy-respecting voice assistant. Built in pure C++ and powered by the open-source legends of `llama.cpp`, `whisper.cpp`, and `piper-tts`, this project aims to put a miniature Iron Man-style AI on your Linux box.

# 📑 Table of Contents

- [🧠 Jarvis](#-jarvis)
- [📑 Table of Contents](#-table-of-contents)
- [🤖 Features](#-features)
- [🗂️ Project Structure](#️-project-structure)
- [📦 Under the Hood](#-under-the-hood)
- [🔮 Roadmap](#-roadmap)
- [🧪 How to Summon Jarvis](#-how-to-summon-jarvis)
- [🧙 Summoning Ritual Complete](#-summoning-ritual-complete)
- [⚠️ Known Limitations](#️-known-limitations)
- [🧩 External Libraries](#-external-libraries)
- [**Disclaimer:** ](#disclaimer)

# 🤖 Features

- 🗣️ Full voice assistant powered by local LLMs
- 🧏 Offline speech recognition using `whisper.cpp`
- 🗯️ Text-to-speech via `piper-tts`
- 🔌 100% local – no internet, no cloud, no snooping
- ⚙️ Modular: STT, TTS, brain – all replaceable or extendable
- 💡 Simple setup with scripts to automate everything
- 🚀 Runs fine on Arch (btw)

# 🗂️ Project Structure

| Path         | Description                                    |
| ------------ | ---------------------------------------------- |
| `src/`       | C++ source files                               |
| `include/`   | C++ header files                               |
| `setup.sh`   | Installs all dependencies and downloads models |
| `buildMe.sh` | Compiles Jarvis using CMake                    |
| `testMe.sh`  | Runs a quick health check                      |
| `bin/`       | Compiled Jarvis binary lives here              |

# 📦 Under the Hood

| Component | Tech Used                                 |
| --------- | ----------------------------------------- |
| 💬 LLM    | `llama.cpp`                               |
| 🧏 STT    | `whisper.cpp`                             |
| 🔊 TTS    | `piper-tts`                               |
| 🧠 Lang   | C++ (because why not suffer beautifully?) |
| 🧪 Tests  | `testMe.sh` script                        |

Everything runs locally. No GPU? No problem. No cloud? Even better.

# 🔮 Roadmap

- Streaming responses
- Multithreaded architecture
- JSON-based config (change personality without recompiling!)
- Tool calling support (maybe your Jarvis will book your therapy)
- Self-destruct command (don’t worry, it only wipes the repo... probably)

# 🧪 How to Summon Jarvis

```
sudo ./setup.sh      # Installs all dependencies and fetches models
sudo ./testMe.sh     # Runs checks to make sure everything is installed
sudo ./buildMe.sh    # Compiles the beast
```

🟢 If everything worked, you'll find the executable in `./bin/jarvis`.

Then:

```
cd bin
./llama-server -m ../models/Meta-Llama-3.1-8B-Instruct.Q4_0.gguf  # to run the server
```

And in another terminal:

```
./jarvis
```

# 🧙 Summoning Ritual Complete

Thanks for checking out Jarvis. Contributions, feedback, and evil plans welcome.

> “Sometimes you just need a voice in your head. Preferably one that compiles.”

# ⚠️ Known Limitations

- Doesn’t actually make coffee or destroy the world... yet
- Only officially tested on **Arch Linux** (btw).  
   That said, if you install the required dependencies using your distro’s package manager, it _should_ work on other Linux distributions as well.  
   _(See line 92 in `setup.sh` for the dependency list.)_

# 🧩 External Libraries

Jarvis relies on several amazing open-source projects:

- [`llama.cpp`](https://github.com/ggml-org/llama.cpp) — Local LLM inference engine
- [`whisper.cpp`](https://github.com/ggerganov/whisper.cpp) — Offline speech recognition
- [`piper`](https://github.com/rhasspy/piper) — Text-to-speech synthesis

# **Disclaimer:**

-Not the real Iron Man’s Jarvis—mine consumes more memory and is 100% open-source. No billionaire genius required!