ai-mirow-projects
/
swarms
mirror of https://github.com/kyegomez/swarms
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

clean up

Browse Source
main 0.1.1
Kye 2 years ago
parent bdf5d873ef
commit 39ffd6acbe
3 changed files with 94 additions and 89 deletions
Show Stats Download Patch File Download Diff File

2
setup.py
Unescape View File

@ -3,7 +3,7 @@ from setuptools import setup, find_packages
setup( setup(
name = 'swarms', name = 'swarms',
packages = find_packages(exclude=[]), packages = find_packages(exclude=[]),
version = '0.1.0', version = '0.1.1',
license='MIT', license='MIT',
description = 'Swarms - Pytorch', description = 'Swarms - Pytorch',
author = 'Kye Gomez', author = 'Kye Gomez',

176
swarms/agents/swarms.py
Unescape View File

@ -21,18 +21,10 @@ from langchain.agents import ZeroShotAgent, Tool, AgentExecutor
from langchain import OpenAI, SerpAPIWrapper, LLMChain from langchain import OpenAI, SerpAPIWrapper, LLMChain
from swarms.agents.workers.auto_agent import worker_agent
worker_agent = worker_agent
# Define your embedding model # Define your embedding model
embeddings_model = OpenAIEmbeddings() # embeddings_model = OpenAIEmbeddings()
# Initialize the vectorstore as empty # Initialize the vectorstore as empty
import faiss import faiss
embedding_size = 1536
index = faiss.IndexFlatL2(embedding_size)
vectorstore = FAISS(embeddings_model.embed_query, index, InMemoryDocstore({}), {})
# from swarms.agents.workers.auto_agent import AutoGPT # from swarms.agents.workers.auto_agent import AutoGPT
@ -123,11 +115,23 @@ embeddings_model = OpenAIEmbeddings(openai_api_key="")
embedding_size = 1536 embedding_size = 1536
index = faiss.IndexFlatL2(embedding_size) index = faiss.IndexFlatL2(embedding_size)
vectorstore = FAISS(embeddings_model.embed_query, index, InMemoryDocstore({}), {}) vectorstore = FAISS(embeddings_model.embed_query, index, InMemoryDocstore({}), {})
####################################################################### => Worker Node
worker_agent = AutoGPT.from_llm_and_tools(
ai_name="WorkerX",
ai_role="Assistant",
tools=tools,
llm=llm,
memory=vectorstore.as_retriever(search_kwargs={"k": 8}),
human_in_the_loop=True, # Set to True if you want to add feedback at each step.
)
worker_agent.chain.verbose = True
####################################################################### => Worker Node
class WorkerNode: class WorkerNode:
def __init__(self, llm, tools, vectorstore): def __init__(self, llm, tools, vectorstore):
@ -170,101 +174,101 @@ worker_node.run_agent("Find 20 potential customers for a Swarms based AI Agent a
#======================================> WorkerNode #======================================> WorkerNode
class MetaWorkerNode: # class MetaWorkerNode:
def __init__(self, llm, tools, vectorstore): # def __init__(self, llm, tools, vectorstore):
self.llm = llm # self.llm = llm
self.tools = tools # self.tools = tools
self.vectorstore = vectorstore # self.vectorstore = vectorstore
self.agent = None # self.agent = None
self.meta_chain = None # self.meta_chain = None
def init_chain(self, instructions): # def init_chain(self, instructions):
self.agent = WorkerNode(self.llm, self.tools, self.vectorstore) # self.agent = WorkerNode(self.llm, self.tools, self.vectorstore)
self.agent.create_agent("Assistant", "Assistant Role", False, {}) # self.agent.create_agent("Assistant", "Assistant Role", False, {})
def initialize_meta_chain(): # def initialize_meta_chain():
meta_template = """ # meta_template = """
Assistant has just had the below interactions with a User. Assistant followed their "Instructions" closely. Your job is to critique the Assistant's performance and then revise the Instructions so that Assistant would quickly and correctly respond in the future. # Assistant has just had the below interactions with a User. Assistant followed their "Instructions" closely. Your job is to critique the Assistant's performance and then revise the Instructions so that Assistant would quickly and correctly respond in the future.
#### # ####
{chat_history} # {chat_history}
#### # ####
Please reflect on these interactions. # Please reflect on these interactions.
You should first critique Assistant's performance. What could Assistant have done better? What should the Assistant remember about this user? Are there things this user always wants? Indicate this with "Critique: ...". # You should first critique Assistant's performance. What could Assistant have done better? What should the Assistant remember about this user? Are there things this user always wants? Indicate this with "Critique: ...".
You should next revise the Instructions so that Assistant would quickly and correctly respond in the future. Assistant's goal is to satisfy the user in as few interactions as possible. Assistant will only see the new Instructions, not the interaction history, so anything important must be summarized in the Instructions. Don't forget any important details in the current Instructions! Indicate the new Instructions by "Instructions: ...". # You should next revise the Instructions so that Assistant would quickly and correctly respond in the future. Assistant's goal is to satisfy the user in as few interactions as possible. Assistant will only see the new Instructions, not the interaction history, so anything important must be summarized in the Instructions. Don't forget any important details in the current Instructions! Indicate the new Instructions by "Instructions: ...".
""" # """
meta_prompt = PromptTemplate( # meta_prompt = PromptTemplate(
input_variables=["chat_history"], template=meta_template # input_variables=["chat_history"], template=meta_template
) # )
meta_chain = LLMChain( # meta_chain = LLMChain(
llm=OpenAI(temperature=0), # llm=OpenAI(temperature=0),
prompt=meta_prompt, # prompt=meta_prompt,
verbose=True, # verbose=True,
) # )
return meta_chain # return meta_chain
def meta_chain(self): # def meta_chain(self):
#define meta template and meta prompting as per your needs # #define meta template and meta prompting as per your needs
self.meta_chain = initialize_meta_chain() # self.meta_chain = initialize_meta_chain()
def get_chat_history(chain_memory): # def get_chat_history(chain_memory):
memory_key = chain_memory.memory_key # memory_key = chain_memory.memory_key
chat_history = chain_memory.load_memory_variables(memory_key)[memory_key] # chat_history = chain_memory.load_memory_variables(memory_key)[memory_key]
return chat_history # return chat_history
def get_new_instructions(meta_output): # def get_new_instructions(meta_output):
delimiter = "Instructions: " # delimiter = "Instructions: "
new_instructions = meta_output[meta_output.find(delimiter) + len(delimiter) :] # new_instructions = meta_output[meta_output.find(delimiter) + len(delimiter) :]
return new_instructions # return new_instructions
def main(self, task, max_iters=3, max_meta_iters=5): # def main(self, task, max_iters=3, max_meta_iters=5):
failed_phrase = "task failed" # failed_phrase = "task failed"
success_phrase = "task succeeded" # success_phrase = "task succeeded"
key_phrases = [success_phrase, failed_phrase] # key_phrases = [success_phrase, failed_phrase]
instructions = "None" # instructions = "None"
for i in range(max_meta_iters): # for i in range(max_meta_iters):
print(f"[Episode {i+1}/{max_meta_iters}]") # print(f"[Episode {i+1}/{max_meta_iters}]")
self.initialize_chain(instructions) # self.initialize_chain(instructions)
output = self.agent.perform('Assistant', {'request': task}) # output = self.agent.perform('Assistant', {'request': task})
for j in range(max_iters): # for j in range(max_iters):
print(f"(Step {j+1}/{max_iters})") # print(f"(Step {j+1}/{max_iters})")
print(f"Assistant: {output}") # print(f"Assistant: {output}")
print(f"Human: ") # print(f"Human: ")
human_input = input() # human_input = input()
if any(phrase in human_input.lower() for phrase in key_phrases): # if any(phrase in human_input.lower() for phrase in key_phrases):
break # break
output = self.agent.perform('Assistant', {'request': human_input}) # output = self.agent.perform('Assistant', {'request': human_input})
if success_phrase in human_input.lower(): # if success_phrase in human_input.lower():
print(f"You succeeded! Thanks for playing!") # print(f"You succeeded! Thanks for playing!")
return # return
self.initialize_meta_chain() # self.initialize_meta_chain()
meta_output = self.meta_chain.predict(chat_history=self.get_chat_history()) # meta_output = self.meta_chain.predict(chat_history=self.get_chat_history())
print(f"Feedback: {meta_output}") # print(f"Feedback: {meta_output}")
instructions = self.get_new_instructions(meta_output) # instructions = self.get_new_instructions(meta_output)
print(f"New Instructions: {instructions}") # print(f"New Instructions: {instructions}")
print("\n" + "#" * 80 + "\n") # print("\n" + "#" * 80 + "\n")
print(f"You failed! Thanks for playing!") # print(f"You failed! Thanks for playing!")
#init instance of MetaWorkerNode # #init instance of MetaWorkerNode
meta_worker_node = MetaWorkerNode(llm=OpenAI, tools=tools, vectorstore=vectorstore) # meta_worker_node = MetaWorkerNode(llm=OpenAI, tools=tools, vectorstore=vectorstore)
#specify a task and interact with the agent # #specify a task and interact with the agent
task = "Provide a sysmatic argument for why we should always eat past with olives" # task = "Provide a sysmatic argument for why we should always eat past with olives"
meta_worker_node.main(task) # meta_worker_node.main(task)
####################################################################### => Boss Node ####################################################################### => Boss Node
@ -356,11 +360,11 @@ agent_executor = AgentExecutor.from_agent_and_tools(
boss_node = BossNode(llm=llm, vectorstore=vectorstore, task_execution_chain=agent_executor, verbose=True, max_iterations=5) boss_node = BossNode(llm=llm, vectorstore=vectorstore, task_execution_chain=agent_executor, verbose=True, max_iterations=5)
#create a task # #create a task
task = boss_node.create_task(objective="Write a research paper on the impact of climate change on global agriculture") # task = boss_node.create_task(objective="Write a research paper on the impact of climate change on global agriculture")
#execute the task # #execute the task
boss_node.execute_task(task) # boss_node.execute_task(task)
class Swarms: class Swarms:

5
swarms/agents/workers/auto_agent.py
Unescape View File

@ -80,6 +80,8 @@ tools = [
ReadFileTool(root_dir="./data"), ReadFileTool(root_dir="./data"),
process_csv, process_csv,
# multimodal_agent_tool,
query_website_tool, query_website_tool,
Terminal, Terminal,
@ -101,8 +103,7 @@ agent_worker = AutoGPT.from_llm_and_tools(
agent_worker.chain.verbose = True agent_worker.chain.verbose = True
worker_agent = agent_worker # worker_agent = agent_worker
# tree_of_thoughts_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... # 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...

Write Preview
Loading…
Cancel
Save