updated aggent

agent.py

@@ -1,3 +1,4 @@
+# agent.py — CPU-only version using Microsoft Phi-2
 import os
 import fitz
 import faiss
@@ -9,8 +10,8 @@ from sentence_transformers import SentenceTransformer
 
 class CodingAgent:
     def __init__(self):
-        # ✅ CPU-compatible LLM
-        model_id = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
+        # Load Microsoft Phi-2 (non-chat, CPU-friendly)
+        model_id = "microsoft/phi-2"
         self.tokenizer = AutoTokenizer.from_pretrained(model_id)
         self.model = AutoModelForCausalLM.from_pretrained(model_id)
         self.llm = pipeline(
@@ -18,7 +19,7 @@ class CodingAgent:
             model=self.model,
             tokenizer=self.tokenizer,
             max_new_tokens=512,
-            device=-1  # forces CPU
+            device=-1
         )
 
         # Embedding model + FAISS index
@@ -76,8 +77,8 @@ class CodingAgent:
             return f"[From memory] {result[0]}"
 
         context = self.retrieve_context(query)
-        prompt = f"You are a helpful coding assistant.\n\nContext:\n{context}\n\nQuestion: {query}\nAnswer:"
-        result = self.llm(prompt)[0]['generated_text'].split("Answer:")[-1].strip()
+        prompt = f"Q: {query}\nContext:\n{context}\nA:"
+        result = self.llm(prompt)[0]['generated_text'].split("A:")[-1].strip()
 
         self.conn.execute("INSERT INTO memory (query, response) VALUES (?, ?)", (query, result))
         self.conn.commit()