FuLLMetal Agent Example

This project provides a Fullmetal Agent written in Node.js that interacts with Fullmetal API.

The main workflow this application is as following:

1. Register itself to Fullmetal API
2. Receive prompts from Fullmetal API
3. Generate a response to the received prompts using LLM running locally
4. Send the response to Fullmetal API

Installation

1. Clone the following repository

git clone https://github.com/m0dE/fullmetal-agent-example

2. Install the project and set your configuration parameters

cd fullmetal-agent-example

npm install

3. Obtain your FULLMETAL_API_KEY by following instructions mentioned here
   
   

4. Create .env file in root folder and add the following keys. Remember all these keys are important. If left blank then your agent may not work properly.

FULLMETAL_API_KEY=YOUR_FULLMETAL_API_KEY

#AGENT NAME
AGENT_NAME=

#MODEL INFORMATION
MODEL_NAME=
MODEL_FILE=

# -ngl N, --n-gpu-layers N number of layers to store in VRAM
NGL=28

Download a GGUF LLM file and save it in the models folder. For example, you can use https://huggingface.co/TheBloke/Wizard-Vicuna-7B-Uncensored-GGUF/resolve/main/Wizard-Vicuna-7B-Uncensored.Q2_K.gguf

mkdir models

cd models

wget https://huggingface.co/TheBloke/Wizard-Vicuna-7B-Uncensored-GGUF/resolve/main/Wizard-Vicuna-7B-Uncensored.Q2_K.gguf

CUDA SUPPORT

Please make sure you have build-essentials installed on your machine.
To install build-essentials

yum groupinstall "Development Tools"

To build node-llama-cpp with CUDA support please run this command inside of your project:

npx --no node-llama-cpp download --cuda

Usage

To run, type:

npm start