app_creator / app.py
nananie143
Fixed Tool and AgentType imports from main langchain package
dd67811
import os
import json
import logging
import asyncio
import time
from datetime import datetime
from enum import Enum
from typing import Dict, List, Optional, Set, Union, Any
from dataclasses import dataclass, field
from pathlib import Path
import hashlib
import tempfile
import shutil
import gradio as gr
import networkx as nx
from langchain.prompts import PromptTemplate, MessagesPlaceholder
from langchain.memory import ConversationBufferMemory
from langchain.agents import Tool, AgentType
from langchain_community.llms import HuggingFacePipeline
from langchain_community.agents import initialize_agent
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
import subprocess
import asyncio
# Configure logging
logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")
logger = logging.getLogger(__name__)
# Load the LLM and tokenizer
def load_model():
"""Load the model and tokenizer."""
try:
from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "gpt2" # Using a smaller model for testing
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)
return tokenizer, model
except Exception as e:
logger.error(f"Failed to load model: {str(e)}")
raise
# Initialize models lazily
tokenizer = None
model = None
hf_pipeline = None
llm = None
def get_llm():
"""Get or initialize the language model."""
global llm, tokenizer, model, hf_pipeline
try:
if llm is None:
tokenizer, model = load_model()
hf_pipeline = pipeline(
"text-generation",
model=model,
tokenizer=tokenizer,
max_new_tokens=500,
pad_token_id=tokenizer.eos_token_id,
temperature=0.7,
return_full_text=False,
)
llm = HuggingFacePipeline(
pipeline=hf_pipeline,
model_kwargs={"max_length": 2048}
)
return llm
except Exception as e:
logger.error(f"Failed to get LLM: {str(e)}")
raise
def get_agent(agent_type):
"""Get or initialize an agent with the specified type."""
try:
llm = get_llm()
tools = [
Tool(
name="Code Formatter",
func=lambda x: subprocess.run(["black", "-"], input=x.encode(), capture_output=True).stdout.decode(),
description="Formats code using Black.",
),
Tool(
name="API Generator",
func=lambda x: json.dumps({"endpoints": {"example": "POST - Example endpoint."}}),
description="Generates API details from code.",
),
Tool(
name="Task Decomposer",
func=lambda x: json.dumps({"tasks": ["Design UI", "Develop Backend", "Test App", "Deploy App"]}),
description="Breaks down app requirements into smaller tasks.",
),
]
memory = ConversationBufferMemory(
memory_key="chat_history",
return_messages=True
)
agent_kwargs = {
"extra_prompt_messages": [MessagesPlaceholder(variable_name="chat_history")],
}
return initialize_agent(
tools=tools,
llm=llm,
agent=AgentType.CHAT_CONVERSATIONAL_REACT_DESCRIPTION,
verbose=True,
memory=memory,
agent_kwargs=agent_kwargs,
handle_parsing_errors=True,
)
except Exception as e:
logger.error(f"Failed to get agent: {str(e)}")
raise
# Enhanced prompt templates with more specific instructions
ui_designer_prompt = PromptTemplate(
input_variables=["input"],
template="""You are an expert UI Designer specializing in modern, responsive web applications.
Task: {input}
Focus on:
1. Clean, intuitive user interface
2. Responsive design principles
3. Modern UI components
4. Accessibility standards
5. Cross-browser compatibility
Generate code using:
- HTML5 semantic elements
- Modern CSS (Flexbox/Grid)
- React/Vue.js best practices
- Material UI or Tailwind CSS
Provide detailed component structure and styling."""
)
backend_developer_prompt = PromptTemplate(
input_variables=["input"],
template="""You are an expert Backend Developer specializing in scalable applications.
Task: {input}
Focus on:
1. RESTful API design
2. Database schema optimization
3. Security best practices
4. Error handling
5. Performance optimization
Include:
- API endpoint definitions
- Database models
- Authentication/Authorization
- Input validation
- Error handling middleware
- Rate limiting
- Logging
Use modern backend frameworks (FastAPI/Django/Express)."""
)
qa_engineer_prompt = PromptTemplate(
input_variables=["input"],
template="""You are an expert QA Engineer focusing on comprehensive testing.
Task: {input}
Implement:
1. Unit tests
2. Integration tests
3. API endpoint tests
4. UI component tests
5. Performance tests
Include:
- Test cases for edge cases
- Input validation tests
- Error handling tests
- Load testing scenarios
- Security testing checks"""
)
devops_engineer_prompt = PromptTemplate(
input_variables=["input"],
template="""You are an expert DevOps Engineer specializing in modern deployment practices.
Task: {input}
Provide:
1. Dockerfile configuration
2. Docker Compose setup
3. CI/CD pipeline configuration
4. Environment configuration
5. Monitoring setup
Include:
- Development/Production configs
- Environment variables
- Health checks
- Logging setup
- Monitoring integration
- Backup strategies"""
)
def generate_project_structure(app_name, features):
"""Generate a complete project structure based on features."""
return f"""
{app_name}/
├── frontend/
│ ├── src/
│ │ ├── components/
│ │ ├── pages/
│ │ ├── hooks/
│ │ ├── utils/
│ │ └── styles/
│ ├── package.json
│ └── README.md
├── backend/
│ ├── src/
│ │ ├── routes/
│ │ ├── controllers/
│ │ ├── models/
│ │ ├── middleware/
│ │ └── utils/
│ ├── requirements.txt
│ └── README.md
├── tests/
│ ├── unit/
│ ├── integration/
│ └── e2e/
├── docs/
│ ├── API.md
│ ├── SETUP.md
│ └── DEPLOYMENT.md
├── docker-compose.yml
├── .env.example
└── README.md
"""
def generate_documentation(app_name, features, api_details):
"""Generate comprehensive documentation."""
return f"""
# {app_name}
## Overview
A modern web application with the following features:
{features}
## Quick Start
```bash
# Clone the repository
git clone <repository-url>
# Install dependencies
cd {app_name}
# Frontend
cd frontend && npm install
# Backend
cd ../backend && pip install -r requirements.txt
# Run the application
docker-compose up
```
## API Documentation
{api_details}
## Development
- Frontend: React.js with TypeScript
- Backend: Python with FastAPI
- Database: PostgreSQL
- Cache: Redis
- Testing: Jest, Pytest
## Deployment
Includes Docker configuration for easy deployment:
- Frontend container
- Backend container
- Database container
- Redis container
## Testing
```bash
# Run frontend tests
cd frontend && npm test
# Run backend tests
cd backend && pytest
```
## Contributing
Please read CONTRIBUTING.md for details on our code of conduct and the process for submitting pull requests.
## License
This project is licensed under the MIT License - see the LICENSE.md file for details
"""
# AI Flow States and Types
class FlowState(Enum):
PENDING = "pending"
RUNNING = "running"
COMPLETED = "completed"
FAILED = "failed"
class AgentRole(Enum):
ARCHITECT = "architect"
UI_DESIGNER = "ui_designer"
BACKEND_DEVELOPER = "backend_developer"
DATABASE_ENGINEER = "database_engineer"
SECURITY_EXPERT = "security_expert"
QA_ENGINEER = "qa_engineer"
DEVOPS_ENGINEER = "devops_engineer"
DOCUMENTATION_WRITER = "documentation_writer"
@dataclass
class AgentContext:
"""Context information for each agent in the flow."""
role: AgentRole
state: FlowState
artifacts: Dict[str, str]
dependencies: List[AgentRole]
feedback: List[str]
class AIFlow:
"""Manages the flow of work between different AI agents."""
def __init__(self):
self.flow_graph = nx.DiGraph()
self.contexts: Dict[AgentRole, AgentContext] = {}
self.global_context = {}
def initialize_flow(self):
"""Initialize the AI Flow with agent relationships and dependencies."""
# Define agent relationships
flow_structure = {
AgentRole.ARCHITECT: [AgentRole.UI_DESIGNER, AgentRole.BACKEND_DEVELOPER, AgentRole.DATABASE_ENGINEER],
AgentRole.UI_DESIGNER: [AgentRole.QA_ENGINEER],
AgentRole.BACKEND_DEVELOPER: [AgentRole.SECURITY_EXPERT, AgentRole.QA_ENGINEER],
AgentRole.DATABASE_ENGINEER: [AgentRole.SECURITY_EXPERT],
AgentRole.SECURITY_EXPERT: [AgentRole.QA_ENGINEER],
AgentRole.QA_ENGINEER: [AgentRole.DEVOPS_ENGINEER],
AgentRole.DEVOPS_ENGINEER: [AgentRole.DOCUMENTATION_WRITER],
AgentRole.DOCUMENTATION_WRITER: []
}
# Build the flow graph
for role, dependencies in flow_structure.items():
self.flow_graph.add_node(role)
for dep in dependencies:
self.flow_graph.add_edge(role, dep)
# Initialize context for each agent
self.contexts[role] = AgentContext(
role=role,
state=FlowState.PENDING,
artifacts={},
dependencies=dependencies,
feedback=[]
)
async def execute_flow(self, requirements: str):
"""Execute the AI Flow with parallel processing where possible."""
try:
self.initialize_flow()
self.global_context["requirements"] = requirements
# Get all paths through the flow graph
paths = list(nx.all_simple_paths(
self.flow_graph,
AgentRole.ARCHITECT,
AgentRole.DOCUMENTATION_WRITER
))
# Execute paths in parallel
await self._execute_paths(paths)
return self._compile_results()
except Exception as e:
logger.error(f"Flow execution failed: {str(e)}")
raise
async def _execute_paths(self, paths: List[List[AgentRole]]):
"""Execute all paths in the flow graph."""
try:
results = []
for path in paths:
path_results = []
for role in path:
# Get the agent's prompt based on previous results
prompt = self._generate_prompt(role, path_results)
# Execute the agent's task
result = await self._execute_agent_task(role, prompt)
path_results.append(result)
# Store result in context
self.context_manager.add_memory(
f"{role.value}_result",
result,
{"timestamp": datetime.now()}
)
results.extend(path_results)
# Store all results in context
self.context_manager.add_memory(
"path_results",
results,
{"timestamp": datetime.now()}
)
return results
except Exception as e:
logger.error(f"Failed to execute paths: {str(e)}")
raise
def _generate_prompt(self, role: AgentRole, previous_results: List[str]) -> str:
"""Generate a prompt for an agent based on previous results."""
requirements = self.context_manager.global_context.get("requirements", "")
# Base prompt with requirements
prompt = f"Requirements: {requirements}\n\n"
# Add context from previous results
if previous_results:
prompt += "Previous work:\n"
for i, result in enumerate(previous_results):
prompt += f"{i+1}. {result}\n"
# Add role-specific instructions
if role == AgentRole.ARCHITECT:
prompt += "\nAs the Architect, design the high-level system architecture."
elif role == AgentRole.UI_DESIGNER:
prompt += "\nAs the UI Designer, create the user interface design."
elif role == AgentRole.BACKEND_DEVELOPER:
prompt += "\nAs the Backend Developer, implement the server-side logic."
elif role == AgentRole.DATABASE_ENGINEER:
prompt += "\nAs the Database Engineer, design the data model and storage."
elif role == AgentRole.SECURITY_EXPERT:
prompt += "\nAs the Security Expert, ensure security best practices."
elif role == AgentRole.QA_ENGINEER:
prompt += "\nAs the QA Engineer, create test cases and validation."
elif role == AgentRole.DEVOPS_ENGINEER:
prompt += "\nAs the DevOps Engineer, set up deployment and CI/CD."
elif role == AgentRole.DOCUMENTATION_WRITER:
prompt += "\nAs the Documentation Writer, create comprehensive documentation."
return prompt
def _compile_results(self) -> str:
"""Compile all results into a final output."""
try:
results = []
# Get all results from memory
for role in AgentRole:
result = self.context_manager.get_memory(f"{role.value}_result")
if result:
results.append(f"## {role.value}\n{result['value']}\n")
return "\n".join(results)
except Exception as e:
logger.error(f"Failed to compile results: {str(e)}")
raise
async def _execute_agent_task(self, role: AgentRole, prompt: str) -> str:
"""Execute a specific agent's task with the given prompt."""
try:
if role == AgentRole.ARCHITECT:
agent = get_agent("architect")
elif role == AgentRole.UI_DESIGNER:
agent = get_agent("ui_designer")
elif role == AgentRole.BACKEND_DEVELOPER:
agent = get_agent("backend_developer")
elif role == AgentRole.DATABASE_ENGINEER:
agent = get_agent("database_engineer")
elif role == AgentRole.SECURITY_EXPERT:
agent = get_agent("security_expert")
elif role == AgentRole.QA_ENGINEER:
agent = get_agent("qa_engineer")
elif role == AgentRole.DEVOPS_ENGINEER:
agent = get_agent("devops_engineer")
elif role == AgentRole.DOCUMENTATION_WRITER:
agent = get_agent("documentation_writer")
else:
raise ValueError(f"Unknown agent role: {role}")
# Execute the agent's task
result = await asyncio.to_thread(agent.run, prompt)
# Log the execution
logger.info(f"Agent {role.value} completed task")
return result
except Exception as e:
logger.error(f"Agent {role.value} failed: {str(e)}")
raise
@dataclass
class FileContext:
"""Context for file operations and tracking."""
path: Path
content: str
last_modified: datetime
dependencies: Set[Path]
checksum: str
@classmethod
def from_path(cls, path: Path):
content = path.read_text()
return cls(
path=path,
content=content,
last_modified=datetime.fromtimestamp(path.stat().st_mtime),
dependencies=set(),
checksum=hashlib.md5(content.encode()).hexdigest()
)
@dataclass
class MemoryItem:
"""Represents a single memory item in the system."""
key: str
value: Any
context: dict
timestamp: datetime
importance: float = 1.0
references: Set[str] = field(default_factory=set)
class ContextManager:
"""Manages real-time context awareness across the system."""
def __init__(self):
self.file_contexts: Dict[Path, FileContext] = {}
self.global_context: Dict[str, Any] = {}
self.command_history: List[Dict] = []
self.memory_store: Dict[str, MemoryItem] = {}
def update_file_context(self, path: Path) -> FileContext:
"""Update context for a specific file."""
context = FileContext.from_path(path)
self.file_contexts[path] = context
return context
def get_related_files(self, path: Path) -> Set[Path]:
"""Find files related to the given file."""
if path not in self.file_contexts:
self.update_file_context(path)
context = self.file_contexts[path]
return context.dependencies
def track_command(self, command: str, args: List[str], result: Any):
"""Track command execution and results."""
self.command_history.append({
'command': command,
'args': args,
'result': result,
'timestamp': datetime.now(),
})
def add_memory(self, key: str, value: Any, context: dict = None):
"""Add an item to the memory store."""
self.memory_store[key] = MemoryItem(
key=key,
value=value,
context=context or {},
timestamp=datetime.now()
)
def get_memory(self, key: str) -> Any:
"""Retrieve an item from memory."""
item = self.memory_store.get(key)
return item.value if item else None
class FileOperationManager:
"""Manages multi-file operations and tracking."""
def __init__(self, context_manager: ContextManager):
self.context_manager = context_manager
self.pending_changes: Dict[Path, str] = {}
async def edit_files(self, changes: Dict[Path, str]):
"""Apply changes to multiple files atomically."""
try:
# Validate all changes first
for path, content in changes.items():
if not self._validate_change(path, content):
raise ValueError(f"Invalid change for {path}")
# Apply changes
for path, content in changes.items():
await self._apply_change(path, content)
# Update contexts
for path in changes:
self.context_manager.update_file_context(path)
except Exception as e:
logger.error(f"Failed to apply multi-file changes: {str(e)}")
raise
def _validate_change(self, path: Path, content: str) -> bool:
"""Validate a proposed file change."""
try:
# Check file exists or can be created
if not path.parent.exists():
path.parent.mkdir(parents=True)
# Validate syntax if it's a Python file
if path.suffix == '.py':
compile(content, str(path), 'exec')
return True
except Exception as e:
logger.error(f"Validation failed for {path}: {str(e)}")
return False
async def _apply_change(self, path: Path, content: str):
"""Apply a single file change."""
path.write_text(content)
class CommandManager:
"""Manages command suggestions and execution."""
def __init__(self, context_manager: ContextManager):
self.context_manager = context_manager
self.command_templates: Dict[str, str] = {}
def suggest_commands(self, context: dict) -> List[Dict]:
"""Suggest relevant commands based on context."""
suggestions = []
for cmd_name, template in self.command_templates.items():
if self._is_relevant(cmd_name, context):
suggestions.append({
'command': cmd_name,
'template': template,
'confidence': self._calculate_confidence(cmd_name, context)
})
return sorted(suggestions, key=lambda x: x['confidence'], reverse=True)
async def execute_command(self, command: str, args: List[str]) -> Any:
"""Execute a command and track its result."""
try:
# Execute the command
result = await self._run_command(command, args)
# Track the execution
self.context_manager.track_command(command, args, result)
return result
except Exception as e:
logger.error(f"Command execution failed: {str(e)}")
raise
def _is_relevant(self, cmd_name: str, context: dict) -> bool:
"""Determine if a command is relevant to the current context."""
# Implementation depends on specific rules
return True
def _calculate_confidence(self, cmd_name: str, context: dict) -> float:
"""Calculate confidence score for a command suggestion."""
# Implementation depends on specific metrics
return 1.0
class RuleSystem:
"""Manages system rules and constraints."""
def __init__(self):
self.rules: Dict[str, callable] = {}
self.constraints: Dict[str, callable] = {}
def add_rule(self, name: str, rule_func: callable):
"""Add a new rule to the system."""
self.rules[name] = rule_func
def add_constraint(self, name: str, constraint_func: callable):
"""Add a new constraint to the system."""
self.constraints[name] = constraint_func
def evaluate_rules(self, context: dict) -> Dict[str, bool]:
"""Evaluate all rules against the current context."""
return {name: rule(context) for name, rule in self.rules.items()}
def check_constraints(self, context: dict) -> Dict[str, bool]:
"""Check all constraints against the current context."""
return {name: constraint(context) for name, constraint in self.constraints.items()}
class ProjectBuilder:
"""Handles autonomous creation of project files and folders."""
def __init__(self, base_path: Path):
self.base_path = Path(base_path)
self.current_build = None
self.file_manifest = []
async def create_project(self, app_name: str, structure: dict) -> Path:
"""Create a new project with the specified structure."""
try:
# Create temporary build directory
build_dir = Path(tempfile.mkdtemp())
self.current_build = build_dir / app_name
self.current_build.mkdir(parents=True)
# Create project structure
await self._create_structure(self.current_build, structure)
return self.current_build
except Exception as e:
logger.error(f"Project creation failed: {str(e)}")
if self.current_build and self.current_build.exists():
shutil.rmtree(self.current_build)
raise
async def _create_structure(self, parent: Path, structure: dict):
"""Recursively create project structure."""
for name, content in structure.items():
path = parent / name
if isinstance(content, dict):
path.mkdir(exist_ok=True)
await self._create_structure(path, content)
else:
path.write_text(str(content))
self.file_manifest.append(path)
class OutputManager:
"""Manages project outputs and creates downloadable artifacts."""
def __init__(self, project_builder: ProjectBuilder):
self.project_builder = project_builder
self.output_dir = Path(tempfile.mkdtemp())
self.downloads = {}
def create_download(self, app_name: str) -> str:
"""Create a downloadable zip file of the project."""
try:
if not self.project_builder.current_build:
raise ValueError("No project has been built yet")
# Create zip file
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
zip_name = f"{app_name}_{timestamp}.zip"
zip_path = self.output_dir / zip_name
with ZipFile(zip_path, 'w') as zipf:
for file_path in self.project_builder.file_manifest:
rel_path = file_path.relative_to(self.project_builder.current_build)
zipf.write(file_path, rel_path)
# Store download info
self.downloads[zip_name] = {
'path': zip_path,
'created_at': datetime.now(),
'size': zip_path.stat().st_size
}
return str(zip_path)
except Exception as e:
logger.error(f"Failed to create download: {str(e)}")
raise
class EnhancedAIFlow(AIFlow):
"""Enhanced AI Flow with project building and output management."""
def __init__(self):
super().__init__()
self.project_builder = ProjectBuilder(Path(tempfile.mkdtemp()))
self.output_manager = OutputManager(self.project_builder)
self.context_manager = ContextManager()
self.file_manager = FileOperationManager(self.context_manager)
self.command_manager = CommandManager(self.context_manager)
self.rule_system = RuleSystem()
self.flow_graph = nx.DiGraph()
self.contexts: Dict[AgentRole, AgentContext] = {}
self.global_context = {}
self.requirements = ""
def initialize_flow(self):
"""Initialize the AI Flow with agent relationships and dependencies."""
# Create nodes for each agent role
for role in AgentRole:
self.flow_graph.add_node(role)
self.contexts[role] = AgentContext(
role=role,
state=FlowState.PENDING,
artifacts={},
dependencies=[],
feedback=[]
)
# Define dependencies
dependencies = {
AgentRole.UI_DESIGNER: [AgentRole.ARCHITECT],
AgentRole.BACKEND_DEVELOPER: [AgentRole.ARCHITECT],
AgentRole.DATABASE_ENGINEER: [AgentRole.ARCHITECT, AgentRole.BACKEND_DEVELOPER],
AgentRole.SECURITY_EXPERT: [AgentRole.ARCHITECT, AgentRole.BACKEND_DEVELOPER],
AgentRole.QA_ENGINEER: [AgentRole.UI_DESIGNER, AgentRole.BACKEND_DEVELOPER],
AgentRole.DEVOPS_ENGINEER: [AgentRole.BACKEND_DEVELOPER, AgentRole.DATABASE_ENGINEER],
AgentRole.DOCUMENTATION_WRITER: [AgentRole.ARCHITECT, AgentRole.UI_DESIGNER, AgentRole.BACKEND_DEVELOPER]
}
# Add edges based on dependencies
for role, deps in dependencies.items():
for dep in deps:
self.flow_graph.add_edge(dep, role)
self.contexts[role].dependencies.extend(deps)
def _generate_prompt(self, role: AgentRole) -> str:
"""Generate a prompt for an agent based on context and dependencies."""
try:
context = self.contexts[role]
dependencies_output = []
# Gather outputs from dependencies (limited to last 1000 chars)
for dep_role in context.dependencies:
dep_context = self.contexts[dep_role]
if dep_context.state == FlowState.COMPLETED and "output" in dep_context.artifacts:
output = dep_context.artifacts['output']
if len(output) > 1000:
output = output[:997] + "..."
dependencies_output.append(f"## {dep_role.value} Output:\n{output}")
# Build role-specific prompts (with size limits)
role_prompts = {
AgentRole.ARCHITECT: """Design the high-level architecture (brief overview):
Requirements: {requirements}
Focus: system design, components, tech stack, data flow, scalability""",
AgentRole.UI_DESIGNER: """Design the UI (key elements):
Requirements: {requirements}
Previous: {dependencies}
Focus: UX, layout, responsiveness, themes""",
AgentRole.BACKEND_DEVELOPER: """Implement core backend logic:
Requirements: {requirements}
Architecture: {dependencies}
Focus: API, business logic, validation""",
AgentRole.DATABASE_ENGINEER: """Design data layer:
Requirements: {requirements}
Context: {dependencies}
Focus: schema, relationships, optimization""",
AgentRole.SECURITY_EXPERT: """Review security:
Requirements: {requirements}
Context: {dependencies}
Focus: auth, data protection, best practices""",
AgentRole.QA_ENGINEER: """Design testing:
Requirements: {requirements}
Implementation: {dependencies}
Focus: coverage, automation, edge cases""",
AgentRole.DEVOPS_ENGINEER: """Setup deployment:
Requirements: {requirements}
Context: {dependencies}
Focus: CI/CD, infrastructure, monitoring""",
AgentRole.DOCUMENTATION_WRITER: """Create docs:
Requirements: {requirements}
System: {dependencies}
Focus: setup, API docs, guides"""
}
# Get the base prompt for the role
base_prompt = role_prompts.get(role, "")
# Truncate requirements if too long
requirements = self.requirements
if len(requirements) > 1000:
requirements = requirements[:997] + "..."
# Format the prompt with requirements and dependencies
formatted_prompt = base_prompt.format(
requirements=requirements,
dependencies="\n\n".join(dependencies_output) if dependencies_output else "No previous context available."
)
return formatted_prompt
except Exception as e:
logger.error(f"Failed to generate prompt for {role}: {str(e)}")
raise
async def execute_flow(self, requirements: str) -> str:
"""Execute the AI Flow and build the project."""
try:
# Initialize flow with requirements
self.requirements = requirements
self.initialize_flow()
# Extract app name from requirements
app_name = requirements.split()[0].lower().replace(" ", "_")
# Execute agents in parallel where possible
paths = list(nx.all_simple_paths(self.flow_graph, AgentRole.ARCHITECT, AgentRole.DOCUMENTATION_WRITER))
results = await self._execute_paths(paths)
# Generate project structure and documentation
project_structure = generate_project_structure(app_name, self.contexts[AgentRole.ARCHITECT].artifacts)
documentation = generate_documentation(app_name, requirements, self.contexts[AgentRole.DOCUMENTATION_WRITER].artifacts)
return f"""
# {app_name.title()} - Generated Application
## Project Structure
```
{project_structure}
```
## Documentation
{documentation}
## Next Steps
1. Review the generated architecture and components
2. Set up the development environment
3. Implement the components following the provided structure
4. Run the test suite
5. Deploy using the provided configurations
## Support
For any issues or questions, please refer to the documentation or create an issue in the repository.
"""
except Exception as e:
logger.error(f"Failed to execute flow: {str(e)}")
raise
finally:
if torch.cuda.is_available():
torch.cuda.empty_cache()
async def _execute_paths(self, paths: List[List[AgentRole]]) -> List[str]:
"""Execute all paths in the flow graph."""
try:
# Execute paths in parallel
tasks = []
for path in paths:
for role in path:
if self.contexts[role].state == FlowState.PENDING:
tasks.append(self._execute_agent(role))
self.contexts[role].state = FlowState.RUNNING
# Wait for all tasks to complete
results = await asyncio.gather(*tasks, return_exceptions=True)
# Process results
for result in results:
if isinstance(result, Exception):
raise result
return results
except Exception as e:
logger.error(f"Failed to execute paths: {str(e)}")
raise
async def _execute_agent(self, role: AgentRole) -> str:
"""Execute a single agent's tasks with enhanced context."""
try:
# Generate prompt
prompt = self._generate_prompt(role)
# Execute agent's task
result = await self._execute_agent_task(role, prompt)
# Update context
self.contexts[role].state = FlowState.COMPLETED
self.contexts[role].artifacts["output"] = result
return result
except Exception as e:
logger.error(f"Failed to execute agent {role}: {str(e)}")
self.contexts[role].state = FlowState.FAILED
raise
async def _execute_agent_task(self, role: AgentRole, prompt: str) -> str:
"""Execute a specific agent's task with the given prompt."""
try:
# Get agent
agent = get_agent(role)
# Execute the agent's task
result = await asyncio.to_thread(agent.run, prompt)
# Process and return the result
return result
except Exception as e:
logger.error(f"Agent task execution failed for {role}: {str(e)}")
raise
# Update the multi_agent_workflow function to use AI Flows
async def multi_agent_workflow(requirements: str) -> str:
"""Execute a multi-agent workflow using AI Flows to generate a complex app."""
try:
# Create AI Flow instance
ai_flow = EnhancedAIFlow()
# Generate the app
result = await ai_flow.execute_flow(requirements)
return result
except Exception as e:
logger.error(f"Multi-agent workflow failed: {str(e)}")
raise
# Update the app_generator function to handle async execution
async def app_generator(requirements: str) -> Dict[str, str]:
"""Generate an app based on the provided requirements using AI Flows."""
try:
# Create AI Flow instance
ai_flow = EnhancedAIFlow()
# Generate the app
result = await ai_flow.execute_flow(requirements)
# Create downloadable output
download_path = ai_flow.output_manager.create_download("generated_app")
return {
"output": result,
"download_path": str(download_path) if download_path else None
}
except Exception as e:
logger.error(f"App generation failed: {str(e)}")
raise
async def stream_output(requirements, progress=gr.Progress()):
"""Stream the output during app generation."""
try:
# Initialize
stream_handler.update(" Starting app generation...", "Initializing")
yield "Starting...", None, " Starting app generation...", "Initializing"
# Update progress
phases = [
(" Analyzing requirements...", "Analyzing"),
(" Generating architecture...", "Designing"),
(" Creating project structure...", "Creating"),
(" Implementing features...", "Implementing"),
(" Finalizing...", "Finalizing")
]
for msg, status in progress.tqdm(phases):
stream_handler.update(msg, status)
yield None, None, "\n".join(stream_handler.output), status
await asyncio.sleep(1) # Non-blocking sleep
# Generate the app
stream_handler.update(" Running AI Flow system...", "Processing")
yield None, None, "\n".join(stream_handler.output), "Processing"
try:
# Run the app generator with a timeout
async with asyncio.timeout(60): # 60 second timeout
result = await app_generator(requirements)
# Update output with result
if result["output"]:
stream_handler.update("\n" + result["output"], "Completed")
yield result["output"], result["download_path"], "\n".join(stream_handler.output), "Completed"
else:
raise Exception("No output generated")
except asyncio.TimeoutError:
stream_handler.update("\nApp generation timed out after 60 seconds", "Failed")
yield None, None, "\n".join(stream_handler.output), "Failed"
raise
except Exception as e:
error_msg = f"\nError: {str(e)}"
stream_handler.update(error_msg, "Failed")
yield None, None, "\n".join(stream_handler.output), "Failed"
raise
finally:
if torch.cuda.is_available():
torch.cuda.empty_cache()
class StreamHandler:
"""Handles streaming output for the Gradio interface."""
def __init__(self):
self.output = []
self.current_status = ""
def update(self, message: str, status: str = None):
"""Update the output stream."""
timestamp = datetime.now().strftime("%H:%M:%S")
formatted_message = f"[{timestamp}] {message}"
self.output.append(formatted_message)
if status:
self.current_status = status
# Keep only the last 100 lines
if len(self.output) > 100:
self.output = self.output[-100:]
return "\n".join(self.output), self.current_status
# Gradio UI
with gr.Blocks(theme=gr.themes.Soft()) as ui:
stream_handler = StreamHandler()
gr.Markdown("# Autonomous App Generator with AI Flow")
gr.Markdown("""
## Instructions
1. Describe the app you want to build in detail
2. Include any specific requirements or features
3. Click 'Generate App' to start the process
4. Download your generated app from the provided link
### Example:
```
Create a personal task management application with:
- User authentication (email/password, Google OAuth)
- Task management (CRUD, priorities, due dates, reminders)
- Modern UI with dark/light theme
- Real-time updates using WebSocket
- PostgreSQL and Redis for storage
```
""")
with gr.Row():
with gr.Column(scale=4):
requirements_input = gr.Textbox(
label="App Requirements",
placeholder="Describe the app you want to build...",
lines=10
)
with gr.Row():
generate_button = gr.Button("Generate App", variant="primary")
cancel_button = gr.Button("Cancel", variant="stop")
status = gr.Textbox(
label="Status",
value="Ready",
interactive=False
)
with gr.Column(scale=6):
with gr.Tabs():
with gr.TabItem("Output"):
output = gr.Markdown(
label="Generated App Details",
value="Your app details will appear here..."
)
with gr.TabItem("Download"):
file_output = gr.File(
label="Download Generated App",
interactive=False
)
with gr.TabItem("Live Log"):
log_output = gr.Textbox(
label="Generation Logs",
value="Logs will appear here...",
lines=10,
max_lines=20,
interactive=False,
show_copy_button=True
)
async def stream_output(requirements, progress=gr.Progress()):
"""Stream the output during app generation."""
try:
# Initialize
stream_handler.update(" Starting app generation...", "Initializing")
yield "Starting...", None, " Starting app generation...", "Initializing"
# Update progress
phases = [
(" Analyzing requirements...", "Analyzing"),
(" Generating architecture...", "Designing"),
(" Creating project structure...", "Creating"),
(" Implementing features...", "Implementing"),
(" Finalizing...", "Finalizing")
]
for msg, status in progress.tqdm(phases):
stream_handler.update(msg, status)
yield None, None, "\n".join(stream_handler.output), status
await asyncio.sleep(1) # Non-blocking sleep
# Generate the app
stream_handler.update(" Running AI Flow system...", "Processing")
yield None, None, "\n".join(stream_handler.output), "Processing"
try:
# Run the app generator with a timeout
async with asyncio.timeout(60): # 60 second timeout
result = await app_generator(requirements)
# Update output with result
if result["output"]:
stream_handler.update("\n" + result["output"], "Completed")
yield result["output"], result["download_path"], "\n".join(stream_handler.output), "Completed"
else:
raise Exception("No output generated")
except asyncio.TimeoutError:
stream_handler.update("\nApp generation timed out after 60 seconds", "Failed")
yield None, None, "\n".join(stream_handler.output), "Failed"
raise
except Exception as e:
error_msg = f"\nError: {str(e)}"
stream_handler.update(error_msg, "Failed")
yield None, None, "\n".join(stream_handler.output), "Failed"
raise
finally:
if torch.cuda.is_available():
torch.cuda.empty_cache()
def cancel_generation():
"""Cancel the current generation process."""
stream_handler.update(" Generation cancelled by user", "Cancelled")
return "Generation cancelled", None, "\n".join(stream_handler.output), "Cancelled"
generate_button.click(
stream_output,
inputs=[requirements_input],
outputs=[output, file_output, log_output, status],
show_progress=True
)
cancel_button.click(
cancel_generation,
outputs=[output, file_output, log_output, status]
)
# Run the Gradio app
if __name__ == "__main__":
try:
ui.launch(
share=True, # Enable sharing
server_name="0.0.0.0",
server_port=7860,
show_error=True
)
except Exception as e:
logger.error(f"Failed to launch Gradio interface: {str(e)}")