Update server.py

server.py

@@ -1,6 +1,5 @@
 import os
 import json
-import subprocess
 from typing import List, Optional
 from fastapi import FastAPI, HTTPException
 from pydantic import BaseModel
@@ -14,7 +13,7 @@ MODEL_PATH = os.path.join("models", MODEL_FILE)
 
 app = FastAPI(title="Autonomous Coding AI")
 
-# --- 1. Model Loader (The Engine) ---
+# --- 1. Model Loader ---
 print("Checking model existence...")
 if not os.path.exists(MODEL_PATH):
     print("Model not found. Downloading...")
@@ -22,42 +21,21 @@ if not os.path.exists(MODEL_PATH):
     hf_hub_download(repo_id=MODEL_ID, filename=MODEL_FILE, local_dir="models")
     print("Download complete.")
 
-print("Loading model into memory (this takes a minute)...")
+print("Loading model into memory...")
 llm = Llama(
     model_path=MODEL_PATH,
-    n_ctx=4096,        # Context window size
-    n_gpu_layers=0,    # 0 for CPU only
+    n_ctx=4096,
+    n_gpu_layers=0, # CPU only
     verbose=False
 )
 print("Model loaded successfully!")
 
-# --- 2. Agent System (The Brain) ---
-# We are creating a "Native" Agency system to keep it lightweight and free.
-
-SYSTEM_PROMPT_ARCHITECT = """
-You are an Architect Agent. Your job is to analyze user requirements and output a JSON project structure.
-Do not write code. Only output JSON.
-Example Output:
-{
-  "project_name": "todo_app",
-  "files": ["main.py", "utils.py"],
-  "tech_stack": ["Python", "FastAPI"]
-}
-"""
-
-SYSTEM_PROMPT_CODER = """
-You are a Coder Agent. You write clean, efficient Python code based on the architecture provided.
-You must output only the code block.
-"""
-
-SYSTEM_PROMPT_SECURITY = """
-You are a Security Agent. You review code for vulnerabilities.
-If the code is safe, output: 'SECURITY CHECK PASSED'.
-If unsafe, output: 'SECURITY ALERT: [reason]'.
-"""
+# --- 2. Agent System ---
+SYSTEM_PROMPT_ARCHITECT = "You are an Architect Agent. Output JSON structure only."
+SYSTEM_PROMPT_CODER = "You are a Coder Agent. Write clean Python code."
+SYSTEM_PROMPT_SECURITY = "You are a Security Agent. Check for vulnerabilities."
 
 def run_agent(system_prompt: str, user_prompt: str) -> str:
-    """Runs the LLM with a specific role."""
     response = llm.create_chat_completion(
         messages=[
             {"role": "system", "content": system_prompt},
@@ -68,26 +46,18 @@ def run_agent(system_prompt: str, user_prompt: str) -> str:
     )
     return response['choices'][0]['message']['content']
 
-# --- 3. Tool: Security Validator (using promptfoo) ---
+# --- 3. Security Tool ---
 def run_security_audit(code: str) -> dict:
-    """
-    Uses promptfoo logic to check for bad practices.
-    """
-    # For this demo, we use a lightweight Python check. 
-    # In production, this calls the real 'promptfoo' CLI.
     unsafe_keywords = ["eval(", "exec(", "password =", "rm -rf"]
     found_issues = []
-    
     for keyword in unsafe_keywords:
         if keyword in code:
             found_issues.append(f"Found unsafe pattern: {keyword}")
-            
     if found_issues:
         return {"status": "FAILED", "details": found_issues}
     return {"status": "PASSED", "details": "Code looks clean."}
 
 # --- 4. API Endpoints ---
-
 class ChatRequest(BaseModel):
     messages: List[dict]
     max_tokens: Optional[int] = 512
@@ -98,28 +68,17 @@ def root():
 
 @app.post("/v1/chat/completions")
 def chat_completions(request: ChatRequest):
-    """
-    OpenAI Compatible Endpoint used by OpenClaw.
-    """
     user_message = request.messages[-1]['content']
     
-    # Step 1: Planning
     print(f">>> [Orchestrator] Received task: {user_message}")
     architecture_plan = run_agent(SYSTEM_PROMPT_ARCHITECT, user_message)
-    print(f">>> [Architect] Plan generated.")
     
-    # Step 2: Coding
     code_output = run_agent(SYSTEM_PROMPT_CODER, f"Architecture:\n{architecture_plan}\n\nRequirement:\n{user_message}")
-    print(f">>> [Coder] Code generated.")
     
-    # Step 3: Security Check
     audit_result = run_security_audit(code_output)
-    print(f">>> [Security] Audit result: {audit_result['status']}")
     
-    # Step 4: Final Formatting
     final_response = f"Project Plan:\n{architecture_plan}\n\nCode:\n```python\n{code_output}\n```\n\nSecurity Audit: {audit_result['status']}"
 
-    # Format for OpenAI compatibility
     return {
         "id": "chatcmpl-001",
         "object": "chat.completion",
@@ -131,4 +90,4 @@ def chat_completions(request: ChatRequest):
             },
             "finish_reason": "stop"
         }]
-    }
+    }