updated aggent

agent.py

@@ -3,17 +3,23 @@ import fitz
 import faiss
 import torch
 import sqlite3
+import numpy as np
 from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
 from sentence_transformers import SentenceTransformer
 
 class CodingAgent:
     def __init__(self):
-        # Load TinyLlama (CPU-friendly)
-        model_id = "mistralai/Mistral-7B-Instruct-v0.2"
+        # ✅ CPU-compatible LLM
+        model_id = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
         self.tokenizer = AutoTokenizer.from_pretrained(model_id)
-        self.model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float16, device_map="auto")
-
-        self.llm = pipeline("text-generation", model=self.model, tokenizer=self.tokenizer, max_new_tokens=512, device=-1)
+        self.model = AutoModelForCausalLM.from_pretrained(model_id)
+        self.llm = pipeline(
+            "text-generation",
+            model=self.model,
+            tokenizer=self.tokenizer,
+            max_new_tokens=512,
+            device=-1  # forces CPU
+        )
 
         # Embedding model + FAISS index
         self.embedder = SentenceTransformer("all-MiniLM-L6-v2")
@@ -28,7 +34,6 @@ class CodingAgent:
     def embed_chunks(self, texts):
         return self.embedder.encode(texts, convert_to_numpy=True)
 
-
     def ingest_file(self, filepath):
         chunks = []
         if filepath.endswith(".pdf"):
@@ -51,7 +56,7 @@ class CodingAgent:
             return "Unsupported file format."
 
         embeddings = self.embed_chunks(chunks)
-        self.index.add(embeddings)
+        self.index.add(np.array(embeddings))
         self.docs.extend(chunks)
         self.id_map.extend(range(len(self.docs)-len(chunks), len(self.docs)))
         return f"Added {len(chunks)} chunks."
@@ -60,7 +65,7 @@ class CodingAgent:
         if self.index.ntotal == 0:
             return ""
         query_emb = self.embed_chunks([query])[0]
-        D, I = self.index.search([query_emb], top_k)
+        D, I = self.index.search(np.array([query_emb]), top_k)
         return "\n\n".join([self.docs[i] for i in I[0]])
 
     def answer(self, query):
@@ -71,7 +76,7 @@ class CodingAgent:
             return f"[From memory] {result[0]}"
 
         context = self.retrieve_context(query)
-        prompt = f"You are a coding assistant. Answer the following:\n\nContext:\n{context}\n\nQuestion: {query}\nAnswer:"
+        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()
 
         self.conn.execute("INSERT INTO memory (query, response) VALUES (?, ?)", (query, result))