Spaces:
Sleeping
Sleeping
File size: 14,649 Bytes
bb3ee41 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 | """Unified memory manager providing access to all memory layers."""
from __future__ import annotations
import logging
from enum import Enum
from typing import Any
from pydantic import BaseModel, Field
from app.config import Settings
from app.memory.long_term import Document, LongTermMemory, SearchResult
from app.memory.shared import Message, SharedMemory
from app.memory.short_term import MemoryEntry, ShortTermMemory
from app.memory.working import WorkingMemory, WorkingMemoryItem
logger = logging.getLogger(__name__)
class MemoryType(str, Enum):
"""Types of memory layers."""
SHORT_TERM = "short_term"
WORKING = "working"
LONG_TERM = "long_term"
SHARED = "shared"
class MemoryStats(BaseModel):
"""Statistics for all memory layers."""
short_term: dict[str, Any] = Field(default_factory=dict)
working: dict[str, Any] = Field(default_factory=dict)
long_term: dict[str, Any] = Field(default_factory=dict)
shared: dict[str, Any] = Field(default_factory=dict)
class MemoryManager:
"""
Unified interface to all memory layers.
The MemoryManager provides a single entry point for interacting with
different types of memory (short-term, working, long-term, shared).
It handles initialization, coordination, and lifecycle management.
Attributes:
short_term: Episode-scoped dictionary memory.
working: LRU-based reasoning scratch space.
long_term: Persistent vector storage.
shared: Multi-agent shared state.
"""
def __init__(self, settings: Settings) -> None:
"""
Initialize memory manager with settings.
Args:
settings: Application settings.
"""
self._settings = settings
self._initialized = False
# Initialize memory layers
self.short_term = ShortTermMemory(
max_size=settings.short_term_memory_size,
)
self.working = WorkingMemory(
capacity=settings.working_memory_size,
)
self.long_term = LongTermMemory(
collection_name=settings.chroma_collection_name,
persist_directory=settings.chroma_persist_directory,
top_k=settings.long_term_memory_top_k,
)
self.shared = SharedMemory()
async def initialize(self) -> None:
"""
Initialize all memory layers.
This should be called during application startup.
"""
if self._initialized:
return
try:
# Initialize long-term memory (ChromaDB)
await self.long_term.initialize()
self._initialized = True
logger.info("Memory manager initialized successfully")
except Exception as e:
logger.error(f"Failed to initialize memory manager: {e}")
raise
async def shutdown(self) -> None:
"""
Shutdown all memory layers gracefully.
This should be called during application shutdown.
"""
try:
# Persist long-term memory
await self.long_term.shutdown()
# Clear working memory
await self.working.clear()
self._initialized = False
logger.info("Memory manager shutdown complete")
except Exception as e:
logger.error(f"Error during memory manager shutdown: {e}")
raise
@property
def is_initialized(self) -> bool:
"""Check if memory manager is initialized."""
return self._initialized
# =========================================================================
# Unified Store Interface
# =========================================================================
async def store(
self,
key: str,
value: Any,
memory_type: MemoryType = MemoryType.SHORT_TERM,
**kwargs: Any,
) -> Any:
"""
Store a value in the specified memory layer.
Args:
key: Key or identifier for the stored value.
value: Value to store.
memory_type: Which memory layer to use.
**kwargs: Additional arguments passed to the specific layer.
Returns:
The created entry/document (varies by memory type).
Raises:
ValueError: If memory_type is invalid.
"""
match memory_type:
case MemoryType.SHORT_TERM:
tags = kwargs.get("tags")
return await self.short_term.set(key, value, tags=tags)
case MemoryType.WORKING:
priority = kwargs.get("priority", 0.0)
metadata = kwargs.get("metadata")
return await self.working.push(
content=value,
item_id=key,
priority=priority,
metadata=metadata,
)
case MemoryType.LONG_TERM:
if not isinstance(value, str):
value = str(value)
metadata = kwargs.get("metadata")
embedding = kwargs.get("embedding")
return await self.long_term.store(
content=value,
document_id=key,
metadata=metadata,
embedding=embedding,
)
case MemoryType.SHARED:
await self.shared.set_state(key, value)
return value
case _:
raise ValueError(f"Invalid memory type: {memory_type}")
# =========================================================================
# Unified Retrieve Interface
# =========================================================================
async def retrieve(
self,
key: str,
memory_type: MemoryType = MemoryType.SHORT_TERM,
default: Any = None,
) -> Any:
"""
Retrieve a value from the specified memory layer.
Args:
key: Key or identifier to look up.
memory_type: Which memory layer to query.
default: Default value if not found.
Returns:
The stored value or default.
Raises:
ValueError: If memory_type is invalid.
"""
match memory_type:
case MemoryType.SHORT_TERM:
return await self.short_term.get(key, default=default)
case MemoryType.WORKING:
item = await self.working.peek_by_id(key)
return item.content if item else default
case MemoryType.LONG_TERM:
doc = await self.long_term.get(key)
return doc.content if doc else default
case MemoryType.SHARED:
return await self.shared.get_state(key, default=default)
case _:
raise ValueError(f"Invalid memory type: {memory_type}")
# =========================================================================
# Unified Search Interface
# =========================================================================
async def search(
self,
query: str,
memory_type: MemoryType = MemoryType.LONG_TERM,
top_k: int = 10,
**kwargs: Any,
) -> list[Any]:
"""
Search for values in the specified memory layer.
Args:
query: Search query.
memory_type: Which memory layer to search.
top_k: Maximum number of results.
**kwargs: Additional arguments for specific layers.
Returns:
List of matching entries/documents.
Raises:
ValueError: If memory_type is invalid or doesn't support search.
"""
match memory_type:
case MemoryType.SHORT_TERM:
# Search by tag or return all keys containing query
tag = kwargs.get("tag")
if tag:
return list((await self.short_term.get_by_tag(tag)).items())[:top_k]
keys = await self.short_term.list_keys()
matching = [k for k in keys if query.lower() in k.lower()]
results = []
for key in matching[:top_k]:
value = await self.short_term.get(key)
results.append((key, value))
return results
case MemoryType.WORKING:
# Search working memory items
def matches(item: WorkingMemoryItem) -> bool:
content_str = str(item.content).lower()
return query.lower() in content_str
items = await self.working.search(matches)
return items[:top_k]
case MemoryType.LONG_TERM:
where = kwargs.get("where")
query_embedding = kwargs.get("query_embedding")
return await self.long_term.search(
query=query,
top_k=top_k,
where=where,
query_embedding=query_embedding,
)
case MemoryType.SHARED:
# Search state keys
all_state = await self.shared.get_all_state()
matching = [
(k, v)
for k, v in all_state.items()
if query.lower() in k.lower()
or query.lower() in str(v).lower()
]
return matching[:top_k]
case _:
raise ValueError(f"Invalid memory type: {memory_type}")
# =========================================================================
# Unified Clear Interface
# =========================================================================
async def clear(
self,
memory_type: MemoryType | None = None,
) -> dict[str, int]:
"""
Clear memory layers.
Args:
memory_type: Specific layer to clear, or None for all.
Returns:
Dictionary with counts of cleared items per layer.
"""
results: dict[str, int] = {}
if memory_type is None or memory_type == MemoryType.SHORT_TERM:
results["short_term"] = await self.short_term.clear()
if memory_type is None or memory_type == MemoryType.WORKING:
results["working"] = await self.working.clear()
if memory_type is None or memory_type == MemoryType.LONG_TERM:
results["long_term"] = await self.long_term.clear()
if memory_type is None or memory_type == MemoryType.SHARED:
shared_results = await self.shared.clear()
results["shared_channels"] = shared_results["channels"]
results["shared_state"] = shared_results["state_keys"]
return results
# =========================================================================
# Episode Management
# =========================================================================
async def start_episode(self, episode_id: str) -> None:
"""
Start a new episode, clearing episode-scoped memory.
Args:
episode_id: Unique identifier for the episode.
"""
await self.short_term.set_episode(episode_id)
await self.working.clear()
logger.debug(f"Started episode: {episode_id}")
async def end_episode(self) -> dict[str, int]:
"""
End the current episode, clearing temporary memory.
Returns:
Counts of cleared items.
"""
results = {
"short_term": await self.short_term.clear(),
"working": await self.working.clear(),
}
logger.debug(f"Ended episode: {results}")
return results
# =========================================================================
# Statistics
# =========================================================================
async def get_stats(self) -> MemoryStats:
"""
Get statistics from all memory layers.
Returns:
MemoryStats with info from each layer.
"""
return MemoryStats(
short_term=await self.short_term.get_stats(),
working=await self.working.get_stats(),
long_term=await self.long_term.get_stats(),
shared=await self.shared.get_stats(),
)
# =========================================================================
# Convenience Methods
# =========================================================================
async def remember(
self,
content: str,
metadata: dict[str, Any] | None = None,
) -> Document:
"""
Store content in long-term memory for later retrieval.
This is a convenience method for storing knowledge.
Args:
content: Text content to remember.
metadata: Optional metadata.
Returns:
The stored document.
"""
return await self.long_term.store(content=content, metadata=metadata)
async def recall(
self,
query: str,
top_k: int = 5,
) -> list[SearchResult]:
"""
Recall relevant memories based on a query.
This is a convenience method for semantic search.
Args:
query: Search query.
top_k: Number of results to return.
Returns:
List of relevant search results.
"""
return await self.long_term.search(query=query, top_k=top_k)
async def think(
self,
thought: str,
priority: float = 0.0,
) -> WorkingMemoryItem:
"""
Add a thought to working memory.
This is a convenience method for reasoning steps.
Args:
thought: The thought content.
priority: Priority score.
Returns:
The working memory item.
"""
return await self.working.push(content=thought, priority=priority)
async def broadcast(
self,
channel: str,
message: Any,
sender: str | None = None,
) -> Message:
"""
Broadcast a message to a shared channel.
This is a convenience method for multi-agent communication.
Args:
channel: Channel name.
message: Message payload.
sender: Optional sender identifier.
Returns:
The published message.
"""
return await self.shared.publish(
channel=channel,
payload=message,
sender=sender,
)
|