ISAAC-497] High-Performance RAG Implementation with Llama 3

main.py

@@ -0,0 +1,63 @@
+import os
+from llama_index.core import VectorStoreIndex, SimpleDirectoryReader, Settings, StorageContext, load_index_from_storage
+from llama_index.llms.ollama import Ollama
+from llama_index.embeddings.huggingface import HuggingFaceEmbedding
+from llama_index.readers.file import PyMuPDFReader # <-- Yeni PDF Okuyucu
+
+# --- ⚙️ CONFIGURATION ---
+print("⚙️ Initializing Isaac RAG Engine (with PyMuPDF)...")
+Settings.llm = Ollama(model="llama3", request_timeout=360.0)
+Settings.embed_model = HuggingFaceEmbedding(model_name="BAAI/bge-small-en-v1.5")
+
+def run_isaac_final_system():
+    # 1. Dosya Kontrolü ve PDF Extractor Tanımlama
+    if not os.path.exists("./documents") or not os.listdir("./documents"):
+        print("❌ Error: 'documents' folder is empty!")
+        return
+
+    # PDF'leri daha iyi okumak için PyMuPDFReader kullanıyoruz
+    file_extractor = {".pdf": PyMuPDFReader()}
+
+    # 2. Akıllı İndeksleme
+    PERSIST_DIR = "./storage"
+
+    # PDF içeriği değiştiyse veya okunmadıysa storage'ı silmen gerekir
+    if not os.path.exists(PERSIST_DIR):
+        print("📄 Reading PDF content deeply and creating index...")
+        # file_extractor'ı buraya ekledik!
+        documents = SimpleDirectoryReader("./documents", file_extractor=file_extractor).load_data()
+
+        # DEBUG: PDF gerçekten okundu mu kontrol et (Gerekirse silersin)
+        for doc in documents:
+            print(f"✅ Loaded: {doc.metadata.get('file_name')} ({len(doc.text)} characters read)")
+
+        index = VectorStoreIndex.from_documents(documents)
+        index.storage_context.persist(persist_dir=PERSIST_DIR)
+    else:
+        print("🧠 Loading existing index (Fast Mode)...")
+        storage_context = StorageContext.from_defaults(persist_dir=PERSIST_DIR)
+        index = load_index_from_storage(storage_context)
+
+    # 3. Sorgu Motoru
+    query_engine = index.as_query_engine(similarity_top_k=2)
+
+    user_question = "What is the main goal of the Isaac project?"
+    print(f"\n❓ Question: {user_question}")
+
+    # 4. Cevap Üretme
+    response = query_engine.query(user_question)
+
+    if not response.source_nodes:
+        print("\n💡 [AI RESPONSE]: No document context found.")
+        return
+
+    best_score = response.source_nodes[0].score if response.source_nodes[0].score else 0
+
+    print("\n💡 [AI RESPONSE]:")
+    if best_score < 0.65:
+        print(f"I couldn't find a strong match in the text (Score: {best_score:.2f}).")
+    else:
+        print(response)
+
+if __name__ == "__main__":
+    run_isaac_final_system()