Update agent_core.py

agent_core.py

@@ -1,63 +1,26 @@
-# agent_core.py
-import re
-from typing import List, Dict, Any
-from semantic_memory import SemanticMemory
-
-REMEMBER_HINTS = (
-    r"\bremember\b", r"\bnote\b", r"\bsave this\b", r"\bstore this\b",
-    r"\bmy name is\b", r"\bmy favorite\b", r"\bI (am|work|run|own)\b",
-)
-
-QUESTION_HINT = r"\?|\bwhat\b|\bwhich\b|\bhow\b|\bwho\b|\bwhen\b"
+from sentence_transformers import SentenceTransformer
 
 class AgentCore:
-    """
-    Routes messages:
-    - If declarative/about-self => store (with boosted weight).
-    - If a question => retrieve + answer with concise summary.
-    """
     def __init__(self):
-        self.memory = SemanticMemory()
-
-    def _looks_like_fact(self, text: str) -> bool:
-        t = text.strip().lower()
-        return any(re.search(p, t) for p in REMEMBER_HINTS) or t.startswith(("i ", "my "))
-
-    def _is_question(self, text: str) -> bool:
-        return bool(re.search(QUESTION_HINT, text.strip().lower()))
-
-    def handle(self, message: str) -> str:
-        if not message or not message.strip():
-            return "Say something and I’ll learn from it."
-
-        # Split on newlines to capture multiple facts in one paste
-        parts = [p.strip() for p in message.split("\n") if p.strip()]
-
-        # If looks factual, store each line
-        stored = 0
-        for p in parts:
-            if self._looks_like_fact(p) and not self._is_question(p):
-                # small boost for strongly self-referential items
-                boost = 0.15 if any(x in p.lower() for x in ("i ", "my ", "motion", "brand", "company")) else 0.0
-                self.memory.add(p, weight=min(1.0, self.memory._base_weight_from_text(p) + boost))
-                stored += 1
-
-        # If question, retrieve relevant context
-        if self._is_question(message):
-            summary = self.memory.summarize_context(message, top_k=6)
-            # Gentle answer style (short + grounded)
-            if summary.startswith("No memory yet"):
-                return "I don’t have context for that yet. Tell me a fact and I’ll remember it."
-            # concise “answer” feel: show 3 strongest lines only
-            lines = summary.splitlines()
-            keep = [l for l in lines if l.startswith("- ")][:3]
-            friendly = "\n".join(keep) if keep else summary
-            return f"Here’s what I recall that’s most relevant:\n{friendly}"
-
-        # Otherwise acknowledge storage or neutral chat
-        if stored > 0:
-            return "Got it — I saved that to your profile and will use it next time."
+        # Using a light, fast model for embeddings
+        self.model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
+
+    def embed_text(self, text):
+        """
+        Convert input text into a numerical vector embedding.
+        """
+        return self.model.encode(text)
+
+    def respond(self, text):
+        """
+        Simple placeholder reasoning for responses.
+        Later you can replace with your reasoning engine or GPT API.
+        """
+        if "who" in text.lower():
+            return "I’m Aventra, your contextual reasoning agent."
+        elif "motionboys" in text.lower():
+            return "MotionBoys is your core empire — fashion meets innovation."
+        elif "dream" in text.lower():
+            return "Dreams are memory anchors — I’m storing this in long-term context."
         else:
-            # Neutral chat that still learns a little
-            self.memory.add(message, weight=self.memory._base_weight_from_text(message))
-            return "Noted. If you want me to remember something long-term, say it directly (e.g., “My brand is MotionBoys”)."
+            return f"I understand: {text}"