Update utils.py

utils.py

@@ -1,41 +1,3 @@
-import os
-import json
-import faiss
-import numpy as np
-from uuid import uuid4
-from datetime import datetime
-from sentence_transformers import SentenceTransformer
-from transformers import AutoTokenizer, AutoModelForCausalLM
-
-# === Config ===
-EMBED_MODEL = "sentence-transformers/all-MiniLM-L6-v2"
-SUMMARIZER_MODEL = "gpt2"
-MEMORY_INDEX_PATH = "memory.index"
-MEMORY_TEXTS_PATH = "memory_texts.json"
-CHAT_LOG_PATH = "chatlog.jsonl"
-FEEDBACK_PATH = "feedback.jsonl"
-SUMMARY_TRIGGER = int(os.getenv("SUMMARY_TRIGGER", 100))
-CHUNK_SIZE = int(os.getenv("SUMMARY_CHUNK", 10))
-
-# === Load models ===
-embedder = SentenceTransformer(EMBED_MODEL)
-summary_tokenizer = AutoTokenizer.from_pretrained(SUMMARIZER_MODEL)
-summary_model = AutoModelForCausalLM.from_pretrained(SUMMARIZER_MODEL).eval()
-
-embedding_dim = embedder.get_sentence_embedding_dimension()
-
-# === Memory state ===
-try:
-    if os.path.exists(MEMORY_INDEX_PATH) and os.path.exists(MEMORY_TEXTS_PATH):
-        memory_index = faiss.read_index(MEMORY_INDEX_PATH)
-        with open(MEMORY_TEXTS_PATH, "r") as f:
-            memory_texts = json.load(f)
-    else:
-        raise FileNotFoundError
-except:
-    memory_index = faiss.IndexFlatL2(embedding_dim)
-    memory_texts = []
-
 
 def get_type(schema):
     if not isinstance(schema, dict):
@@ -60,91 +22,4 @@ def _json_schema_to_python_type(schema, defs):
     if not isinstance(schema, dict):
         return str(type(schema).__name__)
     # The rest of the function is assumed to already exist in the original file
-    return "Handled"  # Placeholder for safety
-
-def embed(texts):
-    """Embed a list of texts into vectors"""
-    return embedder.encode(texts)
-
-def add_to_memory(text):
-    """Add a memory item"""
-    vec = embed([text])
-    memory_index.add(np.array(vec))
-    memory_texts.append({
-        "id": str(uuid4()),
-        "text": text,
-        "timestamp": datetime.now().isoformat()
-    })
-    save_memory()
-    if len(memory_texts) > SUMMARY_TRIGGER:
-        summarize_old_memories()
-
-def retrieve_memories(query, k=3):
-    """Retrieve top relevant memories"""
-    if memory_index.ntotal == 0:
-        return []
-    vec = embed([query])
-    D, I = memory_index.search(np.array(vec), k)
-    return [memory_texts[i]["text"] for i in I[0] if i < len(memory_texts)]
-
-def save_memory():
-    """Save FAISS and text memory to disk"""
-    faiss.write_index(memory_index, MEMORY_INDEX_PATH)
-    with open(MEMORY_TEXTS_PATH, "w") as f:
-        json.dump(memory_texts, f)
-    print(f"[INFO] Memory saved: {len(memory_texts)} items")
-
-def reset_memory():
-    """Reset memory entirely"""
-    memory_index.reset()
-    memory_texts.clear()
-    for path in [MEMORY_INDEX_PATH, MEMORY_TEXTS_PATH]:
-        if os.path.exists(path):
-            os.remove(path)
-
-def summarize_old_memories():
-    """Replace older entries with a summary to save space"""
-    old = "\n".join(m["text"] for m in memory_texts[:CHUNK_SIZE])
-    inputs = summary_tokenizer(f"Summarize: {old}", return_tensors="pt")
-    output = summary_model.generate(**inputs, max_new_tokens=100)
-    summary = summary_tokenizer.decode(output[0][inputs['input_ids'].shape[-1]:], skip_special_tokens=True)
-
-    memory_texts[:CHUNK_SIZE] = [{
-        "id": str(uuid4()),
-        "text": summary,
-        "timestamp": datetime.now().isoformat()
-    }]
-    memory_index.reset()
-    for mem in memory_texts:
-        vec = embed([mem["text"]])
-        memory_index.add(np.array(vec))
-    save_memory()
-
-def log_event(file, entry):
-    """Append an event to a JSONL log"""
-    with open(file, "a") as f:
-        f.write(json.dumps(entry) + "\n")
-
-def log_chat(user_msg, ai_reply, persona):
-    log_event(CHAT_LOG_PATH, {
-        "timestamp": datetime.now().isoformat(),
-        "persona": persona,
-        "user": user_msg,
-        "assistant": ai_reply
-    })
-
-def log_feedback(feedback, suggestions):
-    log_event(FEEDBACK_PATH, {
-        "timestamp": datetime.now().isoformat(),
-        "feedback": feedback,
-        "suggestions": suggestions
-    })
-
-def generate_suggestions(feedback_text):
-    """Generate suggestions for AI improvement"""
-    prompt = f"Based on this feedback: '{feedback_text}' suggest improvements:"
-    inputs = summary_tokenizer(prompt, return_tensors="pt")
-    outputs = summary_model.generate(**inputs, max_new_tokens=100)
-    suggestions = summary_tokenizer.decode(outputs[0][inputs['input_ids'].shape[-1]:], skip_special_tokens=True)
-    log_feedback(feedback_text, suggestions)
-    return suggestions
+    return "Handled"  # Placeholder for safety