chore: Organize root directory and update .gitignore

Author: JuanCS-Dev

.gitignore

@@ -117,8 +117,5 @@ uploads/
 maestro.log
 streaming-fix-package.tar.gz
 
-# Temporary Scripts
-mega_test_e2e.sh
-create-streaming-package.sh
-deploy.sh
-run_mega_test.py
+# Scripts are now in scripts/ directory and should be tracked
+logs/

docs/assets/architecture_cyberpunk.jpeg

@@ -515,19 +515,19 @@ def create_ui() -> tuple[gr.Blocks, str, str]:
                 )
 
             # COLUMN 2: WORKSPACE (Chat + Input)
-            with gr.Column(scale=3):
+            with gr.Column(scale=3, elem_classes="gap-0"):
                 # Chat (Gradio 6: buttons instead of show_copy_button)
                 chatbot = gr.Chatbot(
                     label="Dev Session",
                     height=400,
                     render_markdown=True,
                     avatar_images=(None, None),
-                    elem_classes="cyber-glass",
+                    elem_classes="cyber-glass mb-2", # Add margin bottom to separate from input
                     buttons=["copy"],  # Gradio 6
                 )
 
                 # Input Area
-                with gr.Group(elem_classes="mb-4 relative"):
+                with gr.Group(elem_classes="mb-0 relative"):
                     msg_input = gr.Textbox(
                         placeholder="> Aguardando comando do Mestre...",
                         show_label=False,
@@ -537,18 +537,18 @@ def create_ui() -> tuple[gr.Blocks, str, str]:
                         autofocus=True,
                     )
 
+                # Terminal Logs (HTML) - Moved below input as requested
+                log_display = gr.HTML(
+                    value=render_terminal_logs(_monitor.logs[-8:]),
+                    elem_classes="h-32 overflow-hidden bg-black/50 rounded border border-gray-800 terminal-log mb-4 mt-0"
+                )
+                
                 # Docker Progress Bar
                 progress_html = gr.HTML(
                     value=render_docker_progress(0, "Ready"),
                     elem_classes="docker-progress-container"
                 )
 
-                # Terminal Logs (HTML)
-                log_display = gr.HTML(
-                    value=render_terminal_logs(_monitor.logs[-8:]),
-                    elem_classes="h-32 overflow-hidden bg-black/50 rounded border border-gray-800 terminal-log"
-                )
-
             # COLUMN 3: TELEMETRY
             with gr.Column(scale=1, min_width=250, elem_classes="space-y-4"):
 

docs/assets/prometheus_banner.jpeg

@@ -0,0 +1,118 @@
+"""
+Prometheus Training on Modal.
+
+Runs the Co-Evolution Loop in the cloud to train the agent.
+Persists the learned memory to a Modal Volume.
+"""
+
+import modal
+import os
+import sys
+import asyncio
+from datetime import datetime
+
+# Define the Modal App
+app = modal.App("prometheus-training")
+
+# Define persistence volume for MIRIX memory
+volume = modal.Volume.from_name("prometheus-memory", create_if_missing=True)
+
+# Load .env if present
+from dotenv import load_dotenv
+load_dotenv("/media/juan/DATA/projects/GEMINI-CLI-2/qwen-dev-cli/.env")
+
+# Define the image
+image = (
+    modal.Image.debian_slim()
+    .pip_install("google-generativeai", "rich", "tqdm", "python-dotenv", "httpx")
+    .env({"GEMINI_API_KEY": os.environ.get("GEMINI_API_KEY", "")})
+    .add_local_dir(
+        "/media/juan/DATA/projects/GEMINI-CLI-2/qwen-dev-cli",
+        remote_path="/root/qwen-dev-cli",
+        ignore=["**/.git", "**/__pycache__", "**/.venv", "**/.env", "**/venv"],
+    )
+)
+
+@app.function(
+    image=image,
+    volumes={"/data": volume},
+    timeout=3600,  # 1 hour max training time
+)
+async def train_prometheus(num_iterations: int = 10):
+    """
+    Run the Co-Evolution Loop remotely.
+    """
+    print(f"🚀 Starting Prometheus Training ({num_iterations} iterations)...")
+    
+    # Add project root to path so imports work
+    sys.path.append("/root/qwen-dev-cli")
+    
+    try:
+        from prometheus.core.evolution import CoEvolutionLoop
+        from prometheus.core.llm_client import GeminiClient
+        from prometheus.memory.memory_system import MemorySystem
+        from prometheus.core.reflection import ReflectionEngine
+        from prometheus.tools.tool_factory import ToolFactory
+        from prometheus.sandbox.executor import SandboxExecutor
+    except ImportError as e:
+        print(f"❌ Import Error: {e}")
+        # Debug: list files
+        import subprocess
+        subprocess.run(["find", "/root/qwen-dev-cli", "-maxdepth", "3"])
+        raise
+
+    # Initialize components
+    # Note: In a real scenario, we'd load existing memory from /data
+    print("🧠 Initializing Cognitive Engines...")
+    llm = GeminiClient()
+    memory = MemorySystem()
+    reflection = ReflectionEngine(llm, memory)
+    sandbox = SandboxExecutor()
+    tool_factory = ToolFactory(llm, sandbox)
+
+    # Initialize Evolution Loop
+    evolution = CoEvolutionLoop(
+        llm_client=llm,
+        tool_factory=tool_factory,
+        memory_system=memory,
+        reflection_engine=reflection,
+        sandbox_executor=sandbox,
+    )
+
+    # Run Evolution
+    print("🧬 Evolving...")
+    stats = await evolution.evolve(num_iterations=num_iterations)
+    
+    # Export Memory
+    print("💾 Persisting Memory...")
+    memory_state = memory.export_state()
+    
+    # Save to Volume
+    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+    filename = f"/data/prometheus_memory_{timestamp}.json"
+    
+    import json
+    with open(filename, "w") as f:
+        json.dump(memory_state, f, indent=2)
+        
+    # Also save as 'latest'
+    with open("/data/prometheus_memory_latest.json", "w") as f:
+        json.dump(memory_state, f, indent=2)
+        
+    volume.commit()
+    
+    print(f"✅ Training Complete! Memory saved to {filename}")
+    print(f"📈 Stats: {stats.to_dict()}")
+    
+    return stats.to_dict()
+
+@app.local_entrypoint()
+def main(iterations: int = 10):
+    """
+    Local entrypoint to trigger remote training.
+    Usage: modal run prometheus/training/modal_train.py --iterations 20
+    """
+    print(f"Triggering remote training for {iterations} iterations...")
+    stats = train_prometheus.remote(num_iterations=iterations)
+    print("Remote training finished.")
+    print(stats)