| import pandas as pd |
| from datetime import datetime |
| import hashlib |
| import os |
|
|
| class MemoryEngine: |
| def __init__(self, |
| memory_path="SegmentMemory.csv", |
| short_term_path="ShortTermMemory.csv", |
| history_path="SegmentHistory.csv", |
| conflict_path="SegmentConflictMap.csv"): |
| self.memory_path = memory_path |
| self.short_term_path = short_term_path |
| self.history_path = history_path |
| self.conflict_path = conflict_path |
|
|
| self.memory_df = self._load_csv(memory_path) |
| self.short_term_df = self._load_csv(short_term_path) |
| self.history_df = self._load_csv(history_path) |
| self.conflict_df = self._load_csv(conflict_path) |
|
|
| def _load_csv(self, path): |
| return pd.read_csv(path) if os.path.exists(path) else pd.DataFrame() |
|
|
| def _generate_segment_id(self): |
| existing_ids = self.memory_df['SegmentID'].astype(int) if not self.memory_df.empty else [] |
| return str(max(existing_ids) + 1 if len(existing_ids) > 0 else 1).zfill(4) |
|
|
| def _hash_text(self, text): |
| return hashlib.sha256(text.encode()).hexdigest()[:12] |
|
|
| def insert_segment(self, raw_text, concepts, terms, structure, datapoints, comparisons, applications, agent_id="system"): |
| segment_id = self._generate_segment_id() |
| hash_val = self._hash_text(raw_text) |
|
|
| |
| if hash_val in self.memory_df.get("HashChecksum", []): |
| print("Duplicate detected. Skipping.") |
| return |
|
|
| |
| trust_score = 0.75 if "error" not in raw_text.lower() else -0.4 |
| entropy = 0.5 |
| relevance = 0.6 |
|
|
| |
| conflict_rows = self.memory_df[self.memory_df["RawText"].str.contains(raw_text.split(" ")[0], case=False, na=False)] |
| if not conflict_rows.empty: |
| for _, row in conflict_rows.iterrows(): |
| self._add_conflict(row["SegmentID"], segment_id, agent_id) |
| return |
|
|
| |
| new_row = { |
| "SegmentID": segment_id, |
| "RawText": raw_text, |
| "Concepts": concepts, |
| "Terms": terms, |
| "Structure": structure, |
| "DataPoints": datapoints, |
| "Comparisons": comparisons, |
| "Applications": applications, |
| "Links": "", |
| "Entropy": entropy, |
| "TrustScore": trust_score, |
| "RelevanceScore": relevance, |
| "RecallCount": 0, |
| "LastUsed": "", |
| "ConflictsWith": "", |
| "Verified": trust_score > 0.5, |
| "HashChecksum": hash_val |
| } |
|
|
| self.memory_df = pd.concat([self.memory_df, pd.DataFrame([new_row])], ignore_index=True) |
| self.memory_df.to_csv(self.memory_path, index=False) |
|
|
| self._log_history(agent_id, segment_id, "insert", trust_score, trust_score, relevance, relevance, entropy, entropy, "manual_insert", False) |
| self._add_to_short_term(agent_id, segment_id, raw_text, relevance, trust_score) |
|
|
| print(f"Segment {segment_id} inserted.") |
|
|
| def _log_history(self, agent_id, seg_id, action, c_before, c_after, r_before, r_after, e_before, e_after, context, conflict): |
| log_entry = { |
| "Timestamp": datetime.now().strftime("%Y-%m-%d %H:%M:%S"), |
| "AgentID": agent_id, |
| "SegmentID": seg_id, |
| "ActionType": action, |
| "ConfidenceBefore": c_before, |
| "ConfidenceAfter": c_after, |
| "RelevanceBefore": r_before, |
| "RelevanceAfter": r_after, |
| "EntropyBefore": e_before, |
| "EntropyAfter": e_after, |
| "UsedInContext": context, |
| "ConflictTriggered": conflict, |
| "Notes": "" |
| } |
| self.history_df = pd.concat([self.history_df, pd.DataFrame([log_entry])], ignore_index=True) |
| self.history_df.to_csv(self.history_path, index=False) |
|
|
| def _add_conflict(self, seg_a, seg_b, agent_id): |
| entry = { |
| "SegmentA": seg_a, |
| "SegmentB": seg_b, |
| "ConflictType": "Concept Overlap", |
| "ConflictScore": 0.8, |
| "ResolutionStrategy": "Manual Review", |
| "WinningSegment": "", |
| "LastChecked": datetime.now().strftime("%Y-%m-%d %H:%M:%S"), |
| "ResolutionStatus": "Pending Review", |
| "AgentTrigger": agent_id, |
| "Notes": "Auto conflict detection" |
| } |
| self.conflict_df = pd.concat([self.conflict_df, pd.DataFrame([entry])], ignore_index=True) |
| self.conflict_df.to_csv(self.conflict_path, index=False) |
| print(f"Conflict detected between {seg_a} and {seg_b}. Routed to conflict map.") |
|
|
| def _add_to_short_term(self, agent_id, seg_id, raw_text, relevance, confidence): |
| row = { |
| "AgentID": agent_id, |
| "SegmentID": seg_id, |
| "RawText": raw_text, |
| "Priority": relevance, |
| "ActivationTime": datetime.now().strftime("%Y-%m-%d %H:%M:%S"), |
| "ExpiresAt": "", |
| "DecayRate": 0.03, |
| "RecencyScore": 1.0, |
| "FocusLock": False, |
| "UsageContext": "manual_insert", |
| "ConfidenceScore": confidence, |
| "LastInteraction": datetime.now().strftime("%Y-%m-%d %H:%M:%S") |
| } |
| self.short_term_df = pd.concat([self.short_term_df, pd.DataFrame([row])], ignore_index=True) |
| self.short_term_df.to_csv(self.short_term_path, index=False) |
|
|
| |
| if __name__ == "__main__": |
| brain = MemoryEngine() |
| brain.insert_segment( |
| "Hydrogen is the first element on the periodic table.", |
| "hydrogen, atomic number 1", |
| "hydrogen, element, periodic", |
| "Chemistry > Elements", |
| "Atomic number: 1", |
| "Compared with helium", |
| "Used in gas balloons" |
| ) |
|
|
