Skip to content

Latest commit

 

History

History
57 lines (40 loc) · 1.29 KB

CONTRIBUTING.md

File metadata and controls

57 lines (40 loc) · 1.29 KB

Contributing

Testing

We weren't supposed to connect to OpenAI's paid APIs while testing, so we chose some local services to use for test:

generateText, streamText, generateObject

It uses Ollama.

The model used may change as needed, currently llama3.2 and mistral-nemo.

ollama pull llama3.2
ollama pull mistral-nemo
ollama serve

generateSpeech

It uses openai-edge-tts.

You can quickly start one via docker or docker-compose:

docker run --rm -p 5050:5050 -e PORT=5050 travisvn/openai-edge-tts:latest
services:
  openai-edge-tts:
    image: travisvn/openai-edge-tts:latest
    ports:
      - 5050:5050

generateTranscription

It uses whisper.cpp and the ggml-large-v3-turbo-q5_0 model.

If you're using Nix, you can do this:

{ pkgs, ... }: {
  home.packages = with pkgs; [ openai-whisper-cpp ];
}

Then store the model files in the directory you want and start the server:

mkdir models
cd models
whisper-cpp-download-ggml-model large-v3-turbo-q5_0
cd ..
whisper-cpp-server --host 127.0.0.1 --port 9010 -nt -m models/ggml-large-v3-turbo-q5_0.bin --request-path /audio/transcriptions --inference-path ""