@ -1,159 +1,6 @@
# ---------- Dependencies ----------
import os
import asyncio
import faiss
from typing import Any , Optional , List
from contextlib import contextmanager
from pydantic import BaseModel , Field
from langchain import LLMChain , OpenAI , PromptTemplate
from langchain . chains . base import Chain
from langchain . experimental import BabyAGI
from langchain . embeddings import OpenAIEmbeddings
from langchain . vectorstores . base import VectorStore
from langchain . vectorstores import FAISS
from langchain . docstore import InMemoryDocstore
from langchain . chains . qa_with_sources . loading import load_qa_with_sources_chain
from langchain . agents import ZeroShotAgent , Tool , AgentExecutor
from langchain . text_splitter import RecursiveCharacterTextSplitter
from langchain . tools import BaseTool , DuckDuckGoSearchRun
from langchain . tools . file_management . read import ReadFileTool
from langchain . tools . file_management . write import WriteFileTool
from langchain . tools . human . tool import HumanInputRun
from swarms . tools import Terminal , CodeWriter , CodeEditor , process_csv , WebpageQATool
from langchain . experimental . autonomous_agents . autogpt . agent import AutoGPT
from langchain . chat_models import ChatOpenAI
from langchain . tools import tool
# ---------- Constants ----------
ROOT_DIR = " ./data/ "
# ---------- Tools ----------
openai_api_key = os . environ [ " OPENAI_API_KEY " ]
llm = ChatOpenAI ( model_name = " gpt-4 " , temperature = 1.0 , openai_api_key = openai_api_key )
tools = [
DuckDuckGoSearchRun ( ) ,
WriteFileTool ( root_dir = ROOT_DIR ) ,
ReadFileTool ( root_dir = ROOT_DIR ) ,
process_csv ,
WebpageQATool ( qa_chain = load_qa_with_sources_chain ( llm ) ) ,
# Tool(name='terminal', func=Terminal.execute, description='Operates a terminal'),
# Tool(name='code_writer', func=CodeWriter(), description='Writes code'),
# Tool(name='code_editor', func=CodeEditor(), description='Edits code'),#
]
# ---------- Vector Store ----------
embeddings_model = OpenAIEmbeddings ( )
embedding_size = 1536
index = faiss . IndexFlatL2 ( embedding_size )
vectorstore = FAISS ( embeddings_model . embed_query , index , InMemoryDocstore ( { } ) , { } )
# ---------- Worker Node ----------
# Define the input schema for the WorkerNode
# Define the input schema for the WorkerNode
class WorkerNodeInput ( BaseModel ) :
llm : Any = Field ( description = " Language model " )
tools : List [ Tool ] = Field ( description = " List of tools " )
vectorstore : VectorStore = Field ( description = " Vector store " )
@tool ( " WorkerNode " , args_schema = WorkerNodeInput )
class WorkerNode :
""" Useful for when you need to spawn an autonomous agent instance as a worker to accomplish complex tasks, it can search the internet or spawn child multi-modality models to process and generate images and text or audio and so on """
def __init__ ( self , llm , tools , vectorstore ) :
self . llm = llm
self . tools = tools
self . vectorstore = vectorstore
def create_agent ( self , ai_name , ai_role , human_in_the_loop , search_kwargs ) :
self . agent = AutoGPT . from_llm_and_tools (
ai_name = ai_name ,
ai_role = ai_role ,
tools = tools ,
llm = self . llm ,
memory = self . vectorstore . as_retriever ( search_kwargs = search_kwargs ) ,
human_in_the_loop = human_in_the_loop ,
)
self . agent . chain . verbose = True
def run_agent ( self , prompt ) :
tree_of_thoughts_prompt = """
Imagine three different experts are answering this question . All experts will write down each chain of thought of each step of their thinking , then share it with the group . Then all experts will go on to the next step , etc . If any expert realises they ' re wrong at any point then they leave. The question is...
"""
self . agent . run ( [ f " { tree_of_thoughts_prompt } { prompt } " ] )
worker_node = WorkerNode ( llm = llm , tools = tools , vectorstore = vectorstore )
# ---------- Boss Node ----------
class BossNode :
def __init__ ( self , llm , vectorstore , task_execution_chain , verbose , max_iterations ) :
self . llm = llm
self . vectorstore = vectorstore
self . task_execution_chain = task_execution_chain
self . verbose = verbose
self . max_iterations = max_iterations
self . baby_agi = BabyAGI . from_llm (
llm = self . llm ,
vectorstore = self . vectorstore ,
task_execution_chain = self . task_execution_chain
)
def create_task ( self , objective ) :
return { " objective " : objective }
def execute_task ( self , task ) :
self . baby_agi ( task )
# ---------- Inputs to Boss Node ----------
todo_prompt = PromptTemplate . from_template (
" You are a planner who is an expert at coming up with a todo list for a given objective. Come up with a todo list for this objective: {objective} " " "
)
todo_chain = LLMChain ( llm = OpenAI ( temperature = 0 ) , prompt = todo_prompt )
tools + = [
Tool (
name = " TODO " ,
func = todo_chain . run ,
description = " useful for when you need to come up with todo lists. Input: an objective to create a todo list for. Output: a todo list for that objective. Please be very clear what the objective is! " ,
) ,
Tool (
name = " AUTONOMOUS Worker AGENT " ,
func = WorkerNode . run_agent ,
description = " Useful for when you need to spawn an autonomous agent instance as a worker to accomplish complex tasks, it can search the internet or spawn child multi-modality models to process and generate images and text or audio and so on "
)
]
suffix = """ Question: {task}
{ agent_scratchpad } """
prefix = """ You are an Boss in a swarm who performs one task based on the following objective: {objective} . Take into account these previously completed tasks: {context} .
"""
prompt = ZeroShotAgent . create_prompt (
tools ,
prefix = prefix ,
suffix = suffix ,
input_variables = [ " objective " , " task " , " context " , " agent_scratchpad " ] ,
)
llm = OpenAI ( temperature = 0 )
llm_chain = LLMChain ( llm = llm , prompt = prompt )
tool_names = [ tool . name for tool in tools ]
agent = ZeroShotAgent ( llm_chain = llm_chain , allowed_tools = tool_names )
agent_executor = AgentExecutor . from_agent_and_tools ( agent = agent , tools = tools , verbose = True )
boss_node = BossNode ( llm = llm , vectorstore = vectorstore , task_execution_chain = agent_executor , verbose = True , max_iterations = 5 )
from swarms . tools . main import *
from swarms . agents . workers . worker_agent import WorkerNode
from swarms . agents . boss . boss_agent import BossNode
class Swarms :
def __init__ ( self , openai_api_key ) :
@ -161,7 +8,6 @@ class Swarms:
def initialize_llm ( self ) :
return ChatOpenAI ( model_name = " gpt-4 " , temperature = 1.0 , openai_api_key = self . openai_api_key )
#
def initialize_tools ( self , llm ) :
web_search = DuckDuckGoSearchRun ( )
@ -213,14 +59,7 @@ class Swarms:
llm = self . initialize_llm ( )
tools = self . initialize_tools ( llm )
vectorstore = self . initialize_vectorstore ( )
worker_node = self . initialize_worker_node ( llm , tools , vectorstore )
boss_node = self . initialize_boss_node ( llm , vectorstore )
task = boss_node . create_task ( objective )
boss_node . execute_task ( task )
worker_node . run_agent ( objective )
worker_node = self . initialize_worker
# class Swarms:
Kye
2 years ago