Add swarm logic for task decomposition and validation

README.md

@@ -12,6 +12,7 @@ Ditto is a user-friendly tool that allows you to generate a multi-file Flask app
 - **Automated Code Generation**: Generates routes, templates, and static files based on your description.
 - **Self-Building Agent**: Automatically plans and constructs the application without the need for manual coding.
 - **Modular Structure**: Organizes code into a clean, modular structure with separate directories for templates, static files, and routes.
+- **Swarm Logic**: Utilizes swarm logic to decompose tasks into agents, assign tasks, facilitate agent communication, validate completed tasks, and add validated files to the IDE.
 
 ## Getting Started
 
@@ -39,7 +40,7 @@ Ditto is a user-friendly tool that allows you to generate a multi-file Flask app
 3. **Install Dependencies**
 
    ```bash
-   pip install litellm
+   pip install -r requirements.txt
    ```
 
 ### Setting the `OPENAI_API_KEY`
@@ -120,6 +121,33 @@ python main.py
 python main.py
 ```
 
+### Using Swarm Logic
+
+Ditto now includes a swarm logic feature to enhance task decomposition and validation. Here's how to use it:
+
+1. **Describe Your Task**
+
+   Provide a detailed description of the task you want to accomplish.
+
+2. **Task Decomposition**
+
+   The system will decompose the task into smaller tasks and create agents to handle each task.
+
+3. **Task Assignment**
+
+   Tasks are assigned to agents, and the agents will work on their respective tasks.
+
+4. **Agent Communication**
+
+   Agents will communicate with each other to ensure tasks are completed efficiently.
+
+5. **Task Validation**
+
+   Completed tasks are validated, and if validated, the files are added to the IDE. If denied, the task is sent back to the agent for rework.
+
+6. **Monitor Progress**
+
+   You can monitor the progress of the tasks and validation status in real-time.
 
 ## Contribution
 

main.py

@@ -10,6 +10,9 @@
 # Correctly import the completion function from LiteLLM
 from litellm import completion, supports_function_calling
 
+# Import the swarm_logic module
+from swarm_logic import SwarmLogic
+
 # Configuration
 MODEL_NAME = os.environ.get('LITELLM_MODEL', 'gpt-4o')  # Default model; can be swapped easily
 
@@ -30,7 +33,8 @@
     "iteration": 0,
     "max_iterations": 50,
     "output": "",
-    "completed": False
+    "completed": False,
+    "validation_status": ""
 }
 
 # Ensure directories exist and create __init__.py in routes
@@ -135,6 +139,7 @@ def home():
             progress["iteration"] = 0
             progress["output"] = ""
             progress["completed"] = False
+            progress["validation_status"] = ""
             thread = Thread(target=run_main_loop, args=(user_input,))
             thread.start()
             return render_template_string('''
@@ -323,6 +328,9 @@ def run_main_loop(user_input):
 
     output = ""
 
+    # Initialize SwarmLogic
+    swarm_logic = SwarmLogic()
+
     while iteration < max_iterations:
         progress["iteration"] = iteration + 1
         # Create a new iteration dictionary for each loop
@@ -454,6 +462,13 @@ def run_main_loop(user_input):
     progress["completed"] = True
     progress["status"] = "completed"
 
+    # Use swarm logic to process user input
+    validated_tasks = swarm_logic.process_user_input(user_input)
+    if validated_tasks:
+        progress["validation_status"] = "Tasks validated and files added to IDE."
+    else:
+        progress["validation_status"] = "No tasks validated."
+
     return output
 
 if __name__ == '__main__':

swarm_logic.py

@@ -0,0 +1,51 @@
+import os
+import json
+from swarms import Swarm, Agent
+
+class SwarmLogic:
+    def __init__(self):
+        self.swarm = Swarm()
+
+    def decompose_task(self, user_input):
+        # Decompose the user-defined task into smaller tasks
+        tasks = self.swarm.decompose(user_input)
+        return tasks
+
+    def create_agents(self, tasks):
+        # Create agents for each task
+        agents = [Agent(task) for task in tasks]
+        return agents
+
+    def assign_tasks(self, agents):
+        # Assign tasks to agents
+        for agent in agents:
+            self.swarm.assign(agent)
+
+    def facilitate_communication(self, agents):
+        # Facilitate communication between agents
+        for agent in agents:
+            self.swarm.communicate(agent)
+
+    def validate_tasks(self, agents):
+        # Validate completed tasks
+        validated_tasks = []
+        for agent in agents:
+            if self.swarm.validate(agent):
+                validated_tasks.append(agent.task)
+        return validated_tasks
+
+    def add_validated_files_to_ide(self, validated_tasks):
+        # Add validated files to the IDE
+        for task in validated_tasks:
+            file_path = os.path.join('ide', task['file_name'])
+            with open(file_path, 'w') as f:
+                f.write(task['content'])
+
+    def process_user_input(self, user_input):
+        tasks = self.decompose_task(user_input)
+        agents = self.create_agents(tasks)
+        self.assign_tasks(agents)
+        self.facilitate_communication(agents)
+        validated_tasks = self.validate_tasks(agents)
+        self.add_validated_files_to_ide(validated_tasks)
+        return validated_tasks