Spaces:
Sleeping
Sleeping
File size: 12,511 Bytes
9f203f4 a72fec7 9f203f4 8117b70 9f203f4 8117b70 9f203f4 49f77e8 9f203f4 8117b70 49f77e8 7a1ebee 4386026 8117b70 9f203f4 6f12b05 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 |
# ────────────────────────────── memo/consolidation.py ──────────────────────────────
"""
Memory Consolidation and Pruning
Handles memory consolidation, pruning, and optimization
to prevent information overload and maintain performance.
"""
import re, os
from typing import List, Dict, Any, Optional
from utils.logger import get_logger
from memo.context import cosine_similarity
logger = get_logger("CONSOLIDATION_MANAGER", __name__)
class ConsolidationManager:
"""
Manages memory consolidation and pruning operations.
"""
def __init__(self, memory_system, embedder):
self.memory_system = memory_system
self.embedder = embedder
self.memory_consolidation_threshold = 10 # Consolidate after 10 memories
async def consolidate_memories(self, user_id: str, nvidia_rotator=None) -> Dict[str, Any]:
"""
Consolidate and prune memories to prevent information overload.
"""
try:
if not self.memory_system.is_enhanced_available():
return {"consolidated": 0, "pruned": 0}
# Get all memories for user
all_memories = self.memory_system.enhanced_memory.get_memories(user_id, limit=100)
if len(all_memories) < self.memory_consolidation_threshold:
return {"consolidated": 0, "pruned": 0}
# Group similar memories
memory_groups = await self._group_similar_memories(all_memories, nvidia_rotator)
# Consolidate each group
consolidated_count = 0
pruned_count = 0
for group in memory_groups:
if len(group) > 1:
# Consolidate similar memories
consolidated_memory = await self._consolidate_memory_group(group, nvidia_rotator, user_id)
if consolidated_memory:
# Remove old memories and add consolidated one
for memory in group:
self.memory_system.enhanced_memory.memories.delete_one({"_id": memory["_id"]})
pruned_count += 1
# Add consolidated memory
self.memory_system.enhanced_memory.add_memory(
user_id=user_id,
content=consolidated_memory["content"],
memory_type=consolidated_memory["memory_type"],
importance="high", # Consolidated memories are important
tags=consolidated_memory["tags"] + ["consolidated"]
)
consolidated_count += 1
logger.info(f"[CONSOLIDATION_MANAGER] Consolidated {consolidated_count} groups, pruned {pruned_count} memories")
return {"consolidated": consolidated_count, "pruned": pruned_count}
except Exception as e:
logger.error(f"[CONSOLIDATION_MANAGER] Memory consolidation failed: {e}")
return {"consolidated": 0, "pruned": 0, "error": str(e)}
async def _group_similar_memories(self, memories: List[Dict[str, Any]],
nvidia_rotator) -> List[List[Dict[str, Any]]]:
"""Group similar memories for consolidation"""
try:
if not memories or len(memories) < 2:
return [memories] if memories else []
groups = []
used = set()
for i, memory in enumerate(memories):
if i in used:
continue
group = [memory]
used.add(i)
# Find similar memories
for j, other_memory in enumerate(memories[i+1:], i+1):
if j in used:
continue
# Calculate similarity
similarity = await self._calculate_memory_similarity(memory, other_memory, nvidia_rotator)
if similarity > 0.7: # High similarity threshold
group.append(other_memory)
used.add(j)
groups.append(group)
return groups
except Exception as e:
logger.error(f"[CONSOLIDATION_MANAGER] Memory grouping failed: {e}")
return [memories] if memories else []
async def _calculate_memory_similarity(self, memory1: Dict[str, Any],
memory2: Dict[str, Any], nvidia_rotator) -> float:
"""Calculate similarity between two memories"""
try:
# Use embedding similarity if available
if memory1.get("embedding") and memory2.get("embedding"):
return cosine_similarity(
memory1["embedding"],
memory2["embedding"]
)
# Fallback to content similarity
content1 = memory1.get("content", "")
content2 = memory2.get("content", "")
if not content1 or not content2:
return 0.0
# Simple word overlap similarity
words1 = set(re.findall(r'\b\w+\b', content1.lower()))
words2 = set(re.findall(r'\b\w+\b', content2.lower()))
if not words1 or not words2:
return 0.0
overlap = len(words1.intersection(words2))
total = len(words1.union(words2))
return overlap / total if total > 0 else 0.0
except Exception as e:
logger.warning(f"[CONSOLIDATION_MANAGER] Memory similarity calculation failed: {e}")
return 0.0
async def _consolidate_memory_group(self, group: List[Dict[str, Any]],
nvidia_rotator, user_id: str = "") -> Optional[Dict[str, Any]]:
"""Consolidate a group of similar memories into one"""
try:
if not group or len(group) < 2:
return None
# Extract content from all memories
contents = [memory.get("content", "") for memory in group]
memory_types = list(set(memory.get("memory_type", "conversation") for memory in group))
tags = []
for memory in group:
tags.extend(memory.get("tags", []))
# Use NVIDIA to consolidate content
if nvidia_rotator:
try:
from utils.api.router import generate_answer_with_model
from utils.analytics import get_analytics_tracker
# Track memory agent usage
tracker = get_analytics_tracker()
if tracker:
await tracker.track_agent_usage(
user_id=user_id,
agent_name="memory",
action="consolidate",
context="memory_consolidation",
metadata={"count": len(contents)}
)
sys_prompt = """You are an expert at consolidating similar conversation memories.
Given multiple similar conversation memories, create a single consolidated memory that:
1. Preserves all important information
2. Removes redundancy
3. Maintains the essential context
4. Is concise but comprehensive
Return the consolidated content in the same format as the original memories."""
user_prompt = f"""CONSOLIDATE THESE SIMILAR MEMORIES:
{chr(10).join(f"Memory {i+1}: {content}" for i, content in enumerate(contents))}
Create a single consolidated memory:"""
# Track memory agent usage
try:
from utils.analytics import get_analytics_tracker
tracker = get_analytics_tracker()
if tracker and user_id:
await tracker.track_agent_usage(
user_id=user_id,
agent_name="memory",
action="consolidate",
context="memory_consolidation",
metadata={"memories_count": len(contents)}
)
except Exception:
pass
# Track memory agent usage
tracker = get_analytics_tracker()
if tracker:
await tracker.track_agent_usage(
user_id=user_id,
agent_name="memory",
action="consolidate",
context="memory_consolidation",
metadata={"count": len(contents)}
)
# Track memo agent usage
try:
from utils.analytics import get_analytics_tracker
tracker = get_analytics_tracker()
if tracker:
await tracker.track_agent_usage(
user_id=user_id,
agent_name="memo",
action="consolidate",
context="memory_consolidation",
metadata={"memories_count": len(memories)}
)
except Exception:
pass
# Use Qwen for better memory consolidation reasoning
from utils.api.router import qwen_chat_completion
consolidated_content = await qwen_chat_completion(sys_prompt, user_prompt, nvidia_rotator, user_id, "memory_consolidation")
return {
"content": consolidated_content.strip(),
"memory_type": memory_types[0] if memory_types else "conversation",
"tags": list(set(tags)) + ["consolidated"]
}
except Exception as e:
logger.warning(f"[CONSOLIDATION_MANAGER] NVIDIA consolidation failed: {e}")
# Fallback: simple concatenation
consolidated_content = "\n\n".join(contents)
return {
"content": consolidated_content,
"memory_type": memory_types[0] if memory_types else "conversation",
"tags": list(set(tags)) + ["consolidated"]
}
except Exception as e:
logger.error(f"[CONSOLIDATION_MANAGER] Memory consolidation failed: {e}")
return None
# ────────────────────────────── Global Instance ──────────────────────────────
_consolidation_manager: Optional[ConsolidationManager] = None
def get_consolidation_manager(memory_system=None, embedder=None) -> ConsolidationManager:
"""Get the global consolidation manager instance"""
global _consolidation_manager
if _consolidation_manager is None:
if not memory_system:
from memo.core import get_memory_system
memory_system = get_memory_system()
if not embedder:
from utils.rag.embeddings import EmbeddingClient
embedder = EmbeddingClient()
_consolidation_manager = ConsolidationManager(memory_system, embedder)
logger.info("[CONSOLIDATION_MANAGER] Global consolidation manager initialized")
return _consolidation_manager
# def reset_consolidation_manager():
# """Reset the global consolidation manager (for testing)"""
# global _consolidation_manager
# _consolidation_manager = None
|