> This experimental project is under rapid development and lacks basic safeguards. Until a stable `1.0` release, only run this repository on devices without sensitive information or access to paid services.
By centering this project on [Open Interpreter](https://github.com/OpenInterpreter/open-interpreter), the **01** is more natural, flexible, and capable than its predecessors. Assistants built from this repository can:
**Note:** The [RealtimeSTT](https://github.com/KoljaB/RealtimeSTT) and [RealtimeTTS](https://github.com/KoljaB/RealtimeTTS) libraries at the heart of the 01 are the work of [Kolja Beigel](https://github.com/KoljaB). Please star those repositories and consider contributing to those projects!
- Mac, Windows, and Linux are supported by running `poetry run 01`. This starts the [01 server](https://01.openinterpreter.com/software/run) and a client that uses your `ctrl` key to simulate the 01 light.
- The 01 light is an ESP32-based, push-to-talk voice interface. Build documentation is [here.](https://01.openinterpreter.com/hardware/01-light/materials)
- It works by connecting to the [01 server](https://01.openinterpreter.com/software/run).
<br>
**We need your help supporting & building more hardware.** The 01 should be able to run on any device with input (microphone, keyboard, etc.), output (speakers, screens, motors, etc.), and an internet connection (or sufficient compute to run everything locally). [Contribution Guide ↗️](https://github.com/OpenInterpreter/01/blob/main/CONTRIBUTING.md)
If you stream raw audio bytes to `/` in [Streaming LMC format](https://docs.openinterpreter.com/guides/streaming-response), you will receive its response in the same format.
Inspired in part by [Andrej Karpathy's LLM OS](https://twitter.com/karpathy/status/1723140519554105733), we run a [code-interpreting language model](https://github.com/OpenInterpreter/open-interpreter), and call it when certain events occur at your computer's [kernel](https://github.com/OpenInterpreter/01/blob/main/software/source/server/utils/kernel.py).
To communicate with different components of this system, we introduce [LMC Messages](https://docs.openinterpreter.com/protocols/lmc-messages) format, which extends OpenAI’s messages format to include a "computer" role:
To customize the behavior of the system, edit the [system message, model, skills library path,](https://docs.openinterpreter.com/settings/all-settings) etc. in the `profiles` directory under the `server` directory. This file sets up an interpreter, and is powered by Open Interpreter.
To specify the text-to-speech service for the 01 `base_device.py`, set `interpreter.tts` to either "openai" for OpenAI, "elevenlabs" for ElevenLabs, or "coqui" for Coqui (local) in a profile. For the 01 Light, set `SPEAKER_SAMPLE_RATE` in `client.ino` under the `esp32` client directory to 24000 for Coqui (local) or 22050 for OpenAI TTS. We currently don't support ElevenLabs TTS on the 01 Light.
