Update app/services/llm_engine.py

app/services/llm_engine.py

@@ -2,6 +2,7 @@ import os
 import json
 from groq import AsyncGroq
 from dotenv import load_dotenv
+from app.services.skill_manager import skill_manager
 
 load_dotenv()
 
@@ -316,5 +317,120 @@ class LLMEngine:
             print(f"ERROR Analyze Progress: {e}")
             return f"Error generate progress: {str(e)}"
 
+    async def generate_curriculum_stateless(self, user_data: dict):
+        """
+        Menggabungkan Data JSON Asli + Kecerdasan LLM.
+        """
+        name = user_data.get('name', 'Learner')
+        role_name = user_data.get('active_path', 'General Tech')
+        missing_skills = user_data.get('missing_skills', [])
+
+        if not missing_skills:
+            return []
+
+        # --- STEP 1: AMBIL DATA DARI SKILL MANAGER ---
+        # Disini kita menggunakan object 'skill_manager' yang sudah di-import
+        role_json_data = skill_manager.get_role_data(role_name)
+        
+        verified_context_list = []
+
+        for gap in missing_skills:
+            s_name = gap['skill_name']
+            t_level = gap['target_level'].lower() 
+            
+            # Cari skill di JSON
+            found_skill = None
+            if role_json_data:
+                for s in role_json_data.get('sub_skills', []):
+                    # Simple matching nama skill
+                    if s['name'].lower() in s_name.lower() or s_name.lower() in s['name'].lower():
+                        found_skill = s
+                        break
             
+            # Ambil detail course jika ketemu
+            if found_skill:
+                level_data = found_skill.get('levels', {}).get(t_level, {})
+                rec_data = level_data.get('recommendation', {})
+                
+                real_course = rec_data.get('course_name', 'NOT_FOUND')
+                real_chapters = rec_data.get('specific_chapters', [])
+                
+                verified_context_list.append({
+                    "requested_skill": s_name,
+                    "target_level": t_level,
+                    "OFFICIAL_COURSE": real_course,
+                    "OFFICIAL_CHAPTERS": real_chapters
+                })
+            else:
+                verified_context_list.append({
+                    "requested_skill": s_name,
+                    "target_level": t_level,
+                    "OFFICIAL_COURSE": "NOT_FOUND",
+                    "OFFICIAL_CHAPTERS": []
+                })
+
+        context_str = json.dumps(verified_context_list, indent=2)
+
+        # --- STEP 2: PROMPT ---
+        system_prompt = f"""
+        ROLE: Expert Curriculum Developer.
+        TUGAS: Susun rekomendasi kursus berdasarkan DATA RESMI DATABASE.
+        
+        DATA USER:
+        - Nama: {name}
+        - Target Role: {role_name}
+        
+        DATA RESMI (CONTEXT):
+        {context_str}
+
+        INSTRUKSI UTAMA:
+        1. Iterasi setiap skill dalam DATA RESMI.
+        2. Jika 'OFFICIAL_COURSE' tersedia (Bukan NOT_FOUND):
+           - WAJIB GUNAKAN Judul & Chapters tersebut. JANGAN MENGARANG.
+        3. Jika 'OFFICIAL_COURSE' adalah "NOT_FOUND":
+           - Generate judul & bab yang relevan.
+        
+        LOGIKA BADGE (PRIORITY):
+        - "🔴 High Priority": Jika level target = materi kursus (Setara).
+        - "🟡 Medium Priority": Jika level target < materi kursus (Upskill/Lebih sulit).
+        - "🟢 Low Priority": Jika level target > materi kursus (Review/Lebih mudah).
+
+        OUTPUT JSON:
+        {{
+            "items": [
+                {{
+                    "skill": "Nama Skill",
+                    "current_level": "Level Target",
+                    "course_to_take": "Judul Kursus",
+                    "chapters": ["Bab 1", "Bab 2"],
+                    "match_score": 95.5,
+                    "badge": "🔴 High Priority"
+                }}
+            ]
+        }}
+        """
+
+        try:
+            response_content = await self._execute_with_retry(
+                messages=[{"role": "system", "content": system_prompt}],
+                model="llama-3.3-70b-versatile",
+                temperature=0.3, # Rendah agar patuh data
+                response_format={"type": "json_object"}
+            )
+            
+            data = json.loads(response_content)
+            
+            if "items" in data and isinstance(data["items"], list):
+                return data["items"]
+            
+            for val in data.values():
+                if isinstance(val, list):
+                    return val
+            
+            return []
+
+        except Exception as e:
+            print(f"Error Gen Curriculum: {e}")
+            return []
+
 llm_engine = LLMEngine()