DevToolKit / app.py
acecalisto3's picture
Update app.py
0c763b2 verified
raw
history blame
13.9 kB
import streamlit as st
import os
import subprocess
from transformers import pipeline, AutoModelForCausalLM, AutoTokenizer
import black
from pylint import lint
from io import StringIO
import openai
import sys
# Set your OpenAI API key here
openai.api_key = "YOUR_OPENAI_API_KEY"
PROJECT_ROOT = "projects"
AGENT_DIRECTORY = "agents"
# Global state to manage communication between Tool Box and Workspace Chat App
if 'chat_history' not in st.session_state:
st.session_state.chat_history = []
if 'terminal_history' not in st.session_state:
st.session_state.terminal_history = []
if 'workspace_projects' not in st.session_state:
st.session_state.workspace_projects = {}
if 'available_agents' not in st.session_state:
st.session_state.available_agents = []
class AIAgent:
def __init__(self, name, description, skills):
self.name = name
self.description = description
self.skills = skills
def create_agent_prompt(self):
skills_str = '\n'.join([f"* {skill}" for skill in self.skills])
agent_prompt = f"""
As an elite expert developer, my name is {self.name}. I possess a comprehensive understanding of the following areas:
{skills_str}
I am confident that I can leverage my expertise to assist you in developing and deploying cutting-edge web applications. Please feel free to ask any questions or present any challenges you may encounter.
"""
return agent_prompt
def autonomous_build(self, chat_history, workspace_projects):
"""
Autonomous build logic that continues based on the state of chat history and workspace projects.
"""
summary = "Chat History:\n" + "\n".join([f"User: {u}\nAgent: {a}" for u, a in chat_history])
summary += "\n\nWorkspace Projects:\n" + "\n".join([f"{p}: {details}" for p, details in workspace_projects.items()])
next_step = "Based on the current state, the next logical step is to implement the main application logic."
return summary, next_step
def save_agent_to_file(agent):
"""Saves the agent's prompt to a file."""
if not os.path.exists(AGENT_DIRECTORY):
os.makedirs(AGENT_DIRECTORY)
file_path = os.path.join(AGENT_DIRECTORY, f"{agent.name}.txt")
with open(file_path, "w") as file:
file.write(agent.create_agent_prompt())
st.session_state.available_agents.append(agent.name)
def load_agent_prompt(agent_name):
"""Loads an agent prompt from a file."""
file_path = os.path.join(AGENT_DIRECTORY, f"{agent_name}.txt")
if os.path.exists(file_path):
with open(file_path, "r") as file:
agent_prompt = file.read()
return agent_prompt
else:
return None
def create_agent_from_text(name, text):
skills = text.split('\n')
agent = AIAgent(name, "AI agent created from text input.", skills)
save_agent_to_file(agent)
return agent.create_agent_prompt()
def chat_interface_with_agent(input_text, agent_name):
agent_prompt = load_agent_prompt(agent_name)
if agent_prompt is None:
return f"Agent {agent_name} not found."
model_name = "gpt2"
try:
model = AutoModelForCausalLM.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)
generator = pipeline("text-generation", model=model, tokenizer=tokenizer)
except EnvironmentError as e:
return f"Error loading model: {e}"
combined_input = f"{agent_prompt}\n\nUser: {input_text}\nAgent:"
input_ids = tokenizer.encode(combined_input, return_tensors="pt")
max_input_length = 900
if input_ids.shape[1] > max_input_length:
input_ids = input_ids[:, :max_input_length]
outputs = model.generate(
input_ids, max_new_tokens=50, num_return_sequences=1, do_sample=True,
pad_token_id=tokenizer.eos_token_id # Set pad_token_id to eos_token_id
)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
return response
def workspace_interface(project_name):
if not os.path.exists(PROJECT_ROOT):
os.makedirs(PROJECT_ROOT)
project_path = os.path.join(PROJECT_ROOT, project_name)
if not os.path.exists(project_path):
os.makedirs(project_path)
st.session_state.workspace_projects[project_name] = {"files": []}
return f"Project {project_name} created successfully."
else:
return f"Project {project_name} already exists."
def add_code_to_workspace(project_name, code, file_name):
project_path = os.path.join(PROJECT_ROOT, project_name)
if os.path.exists(project_path):
file_path = os.path.join(project_path, file_name)
with open(file_path, "w") as file:
file.write(code)
st.session_state.workspace_projects[project_name]["files"].append(file_name)
return f"Code added to {file_name} in project {project_name} successfully."
else:
return f"Project {project_name} does not exist."
def terminal_interface(command, project_name=None):
if project_name:
project_path = os.path.join(PROJECT_ROOT, project_name)
if not os.path.exists(project_path):
return f"Project {project_name} does not exist."
result = subprocess.run(command, cwd=project_path, shell=True, capture_output=True, text=True)
else:
result = subprocess.run(command, shell=True, capture_output=True, text=True)
if result.returncode == 0:
return result.stdout
else:
return result.stderr
def code_editor_interface(code):
try:
formatted_code = black.format_str(code, mode=black.FileMode())
except black.NothingChanged:
formatted_code = code
result = StringIO()
sys.stdout = result
sys.stderr = result
(pylint_stdout, pylint_stderr) = lint.py_run(code, return_std=True)
sys.stdout = sys.__stdout__
sys.stderr = sys.__stderr__
lint_message = pylint_stdout.getvalue() + pylint_stderr.getvalue()
return formatted_code, lint_message
def summarize_text(text):
summarizer = pipeline("summarization")
summary = summarizer(text, max_length=50, min_length=25, do_sample=False)
return summary[0]['summary_text']
def sentiment_analysis(text):
analyzer = pipeline("sentiment-analysis")
sentiment = analyzer(text)
return sentiment[0]
def translate_code(code, source_language, target_language):
prompt = f"Translate this code from {source_language} to {target_language}:\n\n{code}"
response = openai.ChatCompletion.create(
model="gpt-4",
messages=[
{"role": "system", "content": "You are an expert software developer."},
{"role": "user", "content": prompt}
]
)
return response.choices[0].message['content'].strip()
def generate_code(code_idea):
response = openai.ChatCompletion.create(
model="gpt-4",
messages=[
{"role": "system", "content": "You are an expert software developer."},
{"role": "user", "content": f"Generate a Python code snippet for the following idea:\n\n{code_idea}"}
]
)
return response.choices[0].message['content'].strip()
st.title("AI Agent Creator")
st.sidebar.title("Navigation")
app_mode = st.sidebar.selectbox("Choose the app mode", ["AI Agent Creator", "Tool Box", "Workspace Chat App"])
if app_mode == "AI Agent Creator":
st.header("Create an AI Agent from Text")
st.subheader("From Text")
agent_name = st.text_input("Enter agent name:")
text_input = st.text_area("Enter skills (one per line):")
if st.button("Create Agent"):
agent_prompt = create_agent_from_text(agent_name, text_input)
st.success(f"Agent '{agent_name}' created and saved successfully.")
st.session_state.available_agents.append(agent_name)
elif app_mode == "Tool Box":
st.header("AI-Powered Tools")
st.subheader("Chat with CodeCraft")
chat_input = st.text_area("Enter your message:")
if st.button("Send"):
if chat_input.startswith("@"):
agent_name = chat_input.split(" ")[0][1:]
chat_input = " ".join(chat_input.split(" ")[1:])
chat_response = chat_interface_with_agent(chat_input, agent_name)
else:
chat_response = "Chat interface function not provided."
st.session_state.chat_history.append((chat_input, chat_response))
st.write(f"CodeCraft: {chat_response}")
st.subheader("Terminal")
terminal_input = st.text_input("Enter a command:")
if st.button("Run"):
terminal_output = terminal_interface(terminal_input)
st.session_state.terminal_history.append((terminal_input, terminal_output))
st.code(terminal_output, language="bash")
st.subheader("Code Editor")
code_editor = st.text_area("Write your code:", height=300)
if st.button("Format & Lint"):
formatted_code, lint_message = code_editor_interface(code_editor)
st.code(formatted_code, language="python")
st.info(lint_message)
st.subheader("Summarize Text")
text_to_summarize = st.text_area("Enter text to summarize:")
if st.button("Summarize"):
summary = summarize_text(text_to_summarize)
st.write(f"Summary: {summary}")
st.subheader("Sentiment Analysis")
sentiment_text = st.text_area("Enter text for sentiment analysis:")
if st.button("Analyze Sentiment"):
sentiment = sentiment_analysis(sentiment_text)
st.write(f"Sentiment: {sentiment}")
st.subheader("Translate Code")
code_to_translate = st.text_area("Enter code to translate:")
source_language = st.text_input("Enter source language (e.g. 'Python'):")
target_language = st.text_input("Enter target language (e.g. 'JavaScript'):")
if st.button("Translate Code"):
translated_code = translate_code(code_to_translate, source_language, target_language)
st.code(translated_code, language=target_language.lower())
st.subheader("Code Generation")
code_idea = st.text_input("Enter your code idea:")
if st.button("Generate Code"):
generated_code = generate_code(code_idea)
st.code(generated_code, language="python")
st.subheader("Preset Commands")
preset_commands = {
"Create a new project": "create_project('project_name')",
"Add code to workspace": "add_code_to_workspace('project_name', 'code', 'file_name')",
"Run terminal command": "terminal_interface('command', 'project_name')",
"Generate code": "generate_code('code_idea')",
"Summarize text": "summarize_text('text')",
"Analyze sentiment": "sentiment_analysis('text')",
"Translate code": "translate_code('code', 'source_language', 'target_language')",
}
for command_name, command in preset_commands.items():
st.write(f"{command_name}: `{command}`")
elif app_mode == "Workspace Chat App":
st.header("Workspace Chat App")
st.subheader("Create a New Project")
project_name = st.text_input("Enter project name:")
if st.button("Create Project"):
workspace_status = workspace_interface(project_name)
st.success(workspace_status)
st.subheader("Add Code to Workspace")
code_to_add = st.text_area("Enter code to add to workspace:")
file_name = st.text_input("Enter file name (e.g. 'app.py'):")
if st.button("Add Code"):
add_code_status = add_code_to_workspace(project_name, code_to_add, file_name)
st.success(add_code_status)
st.subheader("Terminal (Workspace Context)")
terminal_input = st.text_input("Enter a command within the workspace:")
if st.button("Run Command"):
terminal_output = terminal_interface(terminal_input, project_name)
st.code(terminal_output, language="bash")
st.subheader("Chat with CodeCraft for Guidance")
chat_input = st.text_area("Enter your message for guidance:")
if st.button("Get Guidance"):
chat_response = "Chat interface function not provided."
st.session_state.chat_history.append((chat_input, chat_response))
st.write(f"CodeCraft: {chat_response}")
st.subheader("Chat History")
for user_input, response in st.session_state.chat_history:
st.write(f"User: {user_input}")
st.write(f"CodeCraft: {response}")
st.subheader("Terminal History")
for command, output in st.session_state.terminal_history:
st.write(f"Command: {command}")
st.code(output, language="bash")
st.subheader("Workspace Projects")
for project, details in st.session_state.workspace_projects.items():
st.write(f"Project: {project}")
for file in details['files']:
st.write(f" - {file}")
st.subheader("Chat with AI Agents")
if st.session_state.available_agents:
selected_agent = st.selectbox("Select an AI agent", st.session_state.available_agents)
agent_chat_input = st.text_area("Enter your message for the agent:")
if st.button("Send to Agent"):
agent_chat_response = chat_interface_with_agent(agent_chat_input, selected_agent)
st.session_state.chat_history.append((agent_chat_input, agent_chat_response))
st.write(f"{selected_agent}: {agent_chat_response}")
else:
st.write("No agents available. Please create an agent first.")
st.subheader("Automate Build Process")
if st.button("Automate"):
if st.session_state.available_agents:
selected_agent = st.session_state.available_agents[0]
agent = AIAgent(selected_agent, "", [])
summary, next_step = agent.autonomous_build(st.session_state.chat_history, st.session_state.workspace_projects)
st.write("Autonomous Build Summary:")
st.write(summary)
st.write("Next Step:")
st.write(next_step)
else:
st.write("No agents available. Please create an agent first.")