Correctomatic tester with LLM

The purpose of this project is to generate a container that will correct exercises using a LLM model, integrated in the Correctomatic system.

How to use it

Testing the corrections

To test the corrections, you must first build the container with the following command:

export $(grep -v '^#' .env | xargs) && \
docker build -t correction-test-chatgpt \
    --build-arg API_KEY=$API_KEY \
    --build-arg LLM_PROVIDER=$LLM_PROVIDER \
    --build-arg MODEL=$MODEL \
    .

Then you should run the container with the following command:

# Must be absolute paths
PROMPT_FILE=$(pwd)/exercise/prompt.txt
EXERCISE_FILE=$(pwd)/exercise/example_exercise.txt

# Run the container
docker run --rm \
    -v "$EXERCISE_FILE":/tmp/exercise \
    -v "$PROMPT_FILE":/prompt.txt \
    correction-test-chatgpt

Once you're happy with the results, you can build the container and it will be ready to be used in the Correctomatic system.

Development

Prepare the virtual env:

pip -m venv .venv

TO-DO

Build the container

TO-DO

Run the container to develop the code

TO-DO

# Must be absolute paths
PROMPT_FILE=$(pwd)/exercise/prompt.txt
EXERCISE_FILE=$(pwd)/exercise/example_exercise.txt
LLM_PROVIDER=groq # or g4f

docker run --rm \
    -v "$EXERCISE_FILE":/tmp/exercise \
    -v "$PROMPT_FILE":/prompt.txt \
    -v "$(pwd)/app":/app \
    -e "LLM_PROVIDER=$LLM_PROVIDER" \
    correction-test-chatgpt

If you change the requirements.txt file you will need to rebuild the container.

Notes

pip install virtualenv python -m virtualenv .venv

pip install -r requirements.txt

pip install groq

Parece que usa el API de OpenAI: https://platform.openai.com/docs/api-reference/introduction

API Key: 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

export GROQ_API_KEY= https://console.groq.com/playground

docker build -t correction-test-chatgpt .

docker run --rm correction-test-chatgpt

docker run --rm
-v ./exercise/example_exercise.txt:/tmp/exercise
-v ./exercise/prompt.txt:/app/prompt.txt
correction-test-chatgpt

GPT4Free es un proyecto que intenta ofrecer la posibilidad de utilizar el modelo de lenguaje GPT-4 de forma gratuita. Lanza el strem a una lista de proveedores. Aquí tienes la lista: https://github.com/xtekky/gpt4free#-providers-and-models Los que están en verde funcionan y puedes lanzar directamente el strem a ese proveedor o modelo. Página principal de G4F: https://g4f.mintlify.app/get-started/introduction https://github.com/xtekky/gpt4free

Hay que instalar el módulo de Python: https://g4f.mintlify.app/get-started/quickstart pip install -U g4f

Código ejemplo de la página web: https://g4f.mintlify.app/get-started/usage

Huggingchat

HuggingChat https://github.com/Soulter/hugging-chat-api