Nat-D / sebas

SEBAS is our virtual assistant

Setup

0. Install docker and setup Nvidia container toolkit

https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html

1. run ollama server (with GPU)

    docker compose -f docker-compose-gpu.yml up -d

or without gpu

    docker compose -f docker-compose.yml up -d

2. Download llama2 model. (The model is around 4 gb.)

    curl http://localhost:11434/api/pull -d '{
        "name": "llama2"
    }'

3. Try talk to the model

    curl http://localhost:11434/api/chat -d '{
        "model": "llama2",
        "messages": [
            {
            "role": "user",
            "content": "why is the sky blue?"
            }
        ]
    }'

Things to try out.

• System prompt
• Memory and Chatbot capability
• Using RAG
• Output JSON format and validation
• Output parsing
• Calling external functions
• Try Thai Language