Status: Deprecated.

Humoid Llama Bot

Llama.cpp bot for use in the department of Humoids in OneLoveIPFS chat rooms.

Features

• Invokes llama.cpp daemon in chat mode
• Discord and Matrix for chat clients
• Native cross-posting across chat clients

Dependencies required

• nodejs and npm (Latest LTS, v18 minimum supported)
• Matrix account for the bot
• Discord application for the bot

Installation

1. Clone and compile llama.cpp.
2. Download the required models (i.e. ggml-model-q4_0.bin).
3. Copy the example .env file and configure.

cp .env.example .env

4. Install node modules.

npm i

5. Start bot

npm start

Configuration

An example config file has been included in .env.example file. Copy this to .env file and modify. In addition, the bridge may be configured through environment variables and command line args.

Environment variables are prefixed with HUMOID_LLAMA_ and uppercased.

Argument	Description	Default
--log_level	Sets the log level of the bot	info
--exec_path	Path to compiled llama.cpp executable
--exec_args	Additional llama.cpp arguments
--model_path	Path to ggml model
--prompt_file	Path to text file containing prompt
--ignore_prefix	Message prefix ignored by bot	~
--matrix_homeserver	Matrix homeserver where bot lives
--matrix_bot_token	Matrix bot user access token
--matrix_room_id	Matrix room ID that bot responds to
--discord_channel_id	Discord channel ID where bot lives in
--discord_bot_token	Discord bot token
--discord_loading_emoji_id	Suffix to append for response progress on Discord