Spaces:
Sleeping
Sleeping
suneeln-duke
commited on
Commit
•
545c4ba
1
Parent(s):
36ea3c3
Added all agents
Browse files- agents/eda_agent.py +22 -0
- agents/master_agent.py +42 -0
- agents/plant_agent.py +38 -0
- agents/rag_agent.py +70 -0
- main.py +29 -0
agents/eda_agent.py
ADDED
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# from langchain.agents.agent_types import AgentType
|
2 |
+
import pandas as pd
|
3 |
+
|
4 |
+
import sys
|
5 |
+
|
6 |
+
sys.path.append('..')
|
7 |
+
|
8 |
+
from langchain_experimental.agents.agent_toolkits import create_pandas_dataframe_agent
|
9 |
+
from langchain_openai import OpenAI, ChatOpenAI
|
10 |
+
from langchain.agents.agent_types import AgentType
|
11 |
+
|
12 |
+
def init_config(df):
|
13 |
+
|
14 |
+
llm = ChatOpenAI(model_name="gpt-4")
|
15 |
+
|
16 |
+
agent = create_pandas_dataframe_agent(llm=llm, df=df, verbose=True)
|
17 |
+
|
18 |
+
return agent
|
19 |
+
|
20 |
+
def answer_question(agent, question):
|
21 |
+
return agent.invoke(question)['output']
|
22 |
+
|
agents/master_agent.py
ADDED
@@ -0,0 +1,42 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from langchain import hub
|
2 |
+
from langchain.agents import AgentExecutor, create_openai_tools_agent
|
3 |
+
from langchain.agents import load_tools
|
4 |
+
from langchain_openai import ChatOpenAI
|
5 |
+
|
6 |
+
def init_config():
|
7 |
+
|
8 |
+
search_tool = load_tools(['serpapi'])
|
9 |
+
|
10 |
+
tools = [search_tool[0]]
|
11 |
+
|
12 |
+
prompt = hub.pull("hwchase17/openai-tools-agent")
|
13 |
+
|
14 |
+
llm = ChatOpenAI(model="gpt-4", temperature=0)
|
15 |
+
|
16 |
+
# Construct the OpenAI Tools agent
|
17 |
+
agent = create_openai_tools_agent(llm, tools, prompt)
|
18 |
+
|
19 |
+
agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
|
20 |
+
|
21 |
+
return agent_executor
|
22 |
+
|
23 |
+
def answer_question(agent, question):
|
24 |
+
|
25 |
+
full_prompt = f"""
|
26 |
+
You are a master AI that can control other AI agents. You are specifically designed to automate plant maintenance.
|
27 |
+
You will recieve a prompt from a user, and will have to classify the prompt's purpose as one of the following categories:
|
28 |
+
[1] Plant Maintenance
|
29 |
+
[2] AI modeling & EDA
|
30 |
+
[3] Questions about the plant
|
31 |
+
|
32 |
+
Some samples of the prompts are:
|
33 |
+
[1] "Release the fertilizer", "Turn lights on", "Turn lights off", "Water the plant"
|
34 |
+
[2] "What is the accuracy of the model?", "What is the distribution of the data?", "What is the correlation between the features?", "Train a regression model", "Plot the distribution of the data"
|
35 |
+
[3] "What's the optimum moisture of the plant?", "Where does it generally grow?", "What is the plant's life cycle?", "What is the plant's scientific name?.
|
36 |
+
|
37 |
+
Return the category number and name of the prompt in a JSON format.
|
38 |
+
|
39 |
+
User: {question}
|
40 |
+
"""
|
41 |
+
|
42 |
+
return agent.invoke({"input": full_prompt})['output']
|
agents/plant_agent.py
ADDED
@@ -0,0 +1,38 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from langchain import hub
|
2 |
+
from langchain.agents import AgentExecutor, create_openai_tools_agent
|
3 |
+
from langchain.agents import load_tools
|
4 |
+
from langchain_openai import ChatOpenAI
|
5 |
+
|
6 |
+
def init_config():
|
7 |
+
|
8 |
+
search_tool = load_tools(['serpapi'])
|
9 |
+
|
10 |
+
tools = [search_tool[0]]
|
11 |
+
|
12 |
+
prompt = hub.pull("hwchase17/openai-tools-agent")
|
13 |
+
|
14 |
+
llm = ChatOpenAI(model="gpt-4", temperature=0)
|
15 |
+
|
16 |
+
# Construct the OpenAI Tools agent
|
17 |
+
agent = create_openai_tools_agent(llm, tools, prompt)
|
18 |
+
|
19 |
+
agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
|
20 |
+
|
21 |
+
return agent_executor
|
22 |
+
|
23 |
+
def answer_question(agent, question):
|
24 |
+
|
25 |
+
full_prompt = f"""
|
26 |
+
You are a master AI that can control other AI agents. You are specifically designed to automate plant maintenance.
|
27 |
+
You will recieve a prompt from a user regarding plant maintenance, and will have to classify the prompt's purpose as one of the following categories:
|
28 |
+
[0] Fertilizing the plant
|
29 |
+
[1] Turning plant lights on
|
30 |
+
[2] Turning plant lights off
|
31 |
+
[3] Watering the plant
|
32 |
+
|
33 |
+
Return the category number and name of the prompt in a JSON format.
|
34 |
+
|
35 |
+
User: {question}
|
36 |
+
"""
|
37 |
+
|
38 |
+
return agent.invoke({"input": full_prompt})['output']
|
agents/rag_agent.py
ADDED
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
|
3 |
+
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
4 |
+
from langchain.embeddings import OpenAIEmbeddings
|
5 |
+
from langchain.vectorstores.chroma import Chroma
|
6 |
+
import os
|
7 |
+
import shutil
|
8 |
+
|
9 |
+
from langchain.vectorstores.chroma import Chroma
|
10 |
+
from langchain.embeddings import OpenAIEmbeddings
|
11 |
+
|
12 |
+
from langchain.agents.agent_toolkits import create_retriever_tool
|
13 |
+
from langchain.agents.agent_toolkits import create_conversational_retrieval_agent
|
14 |
+
|
15 |
+
from langchain.agents import load_tools
|
16 |
+
|
17 |
+
from langchain_openai import ChatOpenAI
|
18 |
+
|
19 |
+
|
20 |
+
def init_config(loader):
|
21 |
+
# We use the loader created above to load the document
|
22 |
+
documents = loader.load()
|
23 |
+
|
24 |
+
# We split the document into several chunks as mentioned above
|
25 |
+
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=50)
|
26 |
+
texts = text_splitter.split_documents(documents)
|
27 |
+
|
28 |
+
CHROMA_PATH = "../data/plant_chroma"
|
29 |
+
|
30 |
+
if os.path.exists(CHROMA_PATH):
|
31 |
+
db = Chroma(persist_directory=CHROMA_PATH,embedding_function=OpenAIEmbeddings())
|
32 |
+
|
33 |
+
else:
|
34 |
+
db = Chroma.from_documents(
|
35 |
+
texts, OpenAIEmbeddings(), persist_directory=CHROMA_PATH
|
36 |
+
)
|
37 |
+
db.persist()
|
38 |
+
print(f"Saved {len(texts)} chunks to {CHROMA_PATH}.")
|
39 |
+
|
40 |
+
retriever = db.as_retriever()
|
41 |
+
|
42 |
+
# This is the prompt to create a RAG agent for us
|
43 |
+
|
44 |
+
|
45 |
+
retriever_name = "plant_os_pdf"
|
46 |
+
retriever_desc = """The purpose of this tool is to answer questions about the blue indigo false plant and its maintenance."""
|
47 |
+
|
48 |
+
rag_tool = create_retriever_tool(
|
49 |
+
retriever,
|
50 |
+
retriever_name,
|
51 |
+
retriever_desc
|
52 |
+
)
|
53 |
+
|
54 |
+
search_tool = load_tools(['serpapi'])
|
55 |
+
tools = [rag_tool, search_tool[0]]
|
56 |
+
|
57 |
+
llm = ChatOpenAI(model_name="gpt-4")
|
58 |
+
|
59 |
+
RAG_executor = create_conversational_retrieval_agent(llm=llm, tools=tools, verbose=True) # setting verbose=True to output the thought process of the agent
|
60 |
+
|
61 |
+
return RAG_executor
|
62 |
+
|
63 |
+
def answer_question(agent, question):
|
64 |
+
question = "what is the scientific for the plant?"
|
65 |
+
|
66 |
+
user_query = {"input": question}
|
67 |
+
|
68 |
+
result = agent(user_query)
|
69 |
+
|
70 |
+
return result['output']
|
main.py
CHANGED
@@ -1,6 +1,35 @@
|
|
1 |
from fastapi import FastAPI
|
2 |
|
|
|
|
|
|
|
|
|
|
|
3 |
from langchain.document_loaders import DirectoryLoader
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
|
5 |
app = FastAPI()
|
6 |
|
|
|
1 |
from fastapi import FastAPI
|
2 |
|
3 |
+
import os
|
4 |
+
import json
|
5 |
+
import sys
|
6 |
+
|
7 |
+
import pandas as pd
|
8 |
from langchain.document_loaders import DirectoryLoader
|
9 |
+
from agents import master_agent, plant_agent, eda_agent, rag_agent
|
10 |
+
|
11 |
+
app = FastAPI()
|
12 |
+
|
13 |
+
os.environ['OPENAI_API_KEY'] = os.getenv['openapi-key']
|
14 |
+
os.environ['SERPAPI_API_KEY'] = os.getenv['serpapi-key']
|
15 |
+
|
16 |
+
master = master_agent.init_config()
|
17 |
+
|
18 |
+
print("init master agent")
|
19 |
+
|
20 |
+
plant = plant_agent.init_config()
|
21 |
+
|
22 |
+
print("init plant agent")
|
23 |
+
|
24 |
+
df = pd.read_csv('data/csv/plant_syn.csv')
|
25 |
+
|
26 |
+
eda = eda_agent.init_config(df)
|
27 |
+
|
28 |
+
print("init eda agent")
|
29 |
+
|
30 |
+
loader = DirectoryLoader("data/txt", glob="*.txt")
|
31 |
+
|
32 |
+
rag = rag_agent.init_config(loader)
|
33 |
|
34 |
app = FastAPI()
|
35 |
|