rosgpt4all.gpt

Classes

GPT4All ROS Node.

Functions

Module Contents

class rosgpt4all.gpt.GPTNode

Bases: rclpy.node.Node

GPT4All ROS Node.

running = True

socket = None

socket_clients = []

queue = []

gpt_abort = False

model_name

model_path

allow_download

socket_path

device

system_prompt

prompt

eof_indicator

max_tokens

temp

top_k

top_p

repeat_penalty

repeat_last_n

n_batch

n_threads

ngl

verbose

sub

pub_generator

pub

pub_sentence

sentence = ''

model

session_thread

sock_thread

gpt_callback(token_id: int, response: str): Callback to handle self.gpt_abort signal.

shutdown(): Cleanup running threads.

parameter_callback(params): ROS dynamic reconfigure handler for RQT Gui

topic_callback(msg): Pass incoming message to GPT4ALL.

prompt_cleanup(prompt): Removes context until the end from prompt.

session(): Process incoming queue messages by GPT4ALL.

token_handler(token): Publishes to the generator and sentence publishers.

publish(s, pub): Publish GPT4ALL response.

print(s): Print to stdout and socket.

socket_server(): Serves GPT output to a AF_UNIX socket.

socket_send(bytes): Send bytes to all connected clients. Also maintains the self.socket_clients list.

rosgpt4all.gpt.main()