Spaces:
Sleeping
Sleeping
File size: 15,590 Bytes
bb3ee41 bfe0e24 bb3ee41 bfe0e24 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 | """Long-term memory with persistent vector storage using ChromaDB."""
from __future__ import annotations
import asyncio
import hashlib
import logging
from datetime import datetime, timezone
from typing import Any
from uuid import uuid4
from pydantic import BaseModel, Field
logger = logging.getLogger(__name__)
class Document(BaseModel):
"""A document stored in long-term memory."""
id: str = Field(default_factory=lambda: str(uuid4()))
content: str
embedding: list[float] | None = None
metadata: dict[str, Any] = Field(default_factory=dict)
created_at: datetime = Field(default_factory=datetime.utcnow)
updated_at: datetime = Field(default_factory=datetime.utcnow)
model_config = {"arbitrary_types_allowed": True}
class SearchResult(BaseModel):
"""A search result from long-term memory."""
document: Document
score: float
distance: float | None = None
model_config = {"arbitrary_types_allowed": True}
class LongTermMemory:
"""
Long-term persistent memory using ChromaDB for vector storage.
This memory layer provides semantic search capabilities using embeddings.
It persists across episodes and sessions, storing knowledge that should
be retained long-term.
Attributes:
collection_name: Name of the ChromaDB collection.
persist_directory: Directory for persistent storage.
top_k: Default number of results to return from search.
"""
def __init__(
self,
collection_name: str = "scraperl_memory",
persist_directory: str = "./data/chroma",
top_k: int = 10,
embedding_function: Any | None = None,
) -> None:
"""
Initialize long-term memory.
Args:
collection_name: Name of the ChromaDB collection.
persist_directory: Directory for persistent storage.
top_k: Default number of results to return from search.
embedding_function: Optional custom embedding function.
"""
self.collection_name = collection_name
self.persist_directory = persist_directory
self.top_k = top_k
self._embedding_function = embedding_function
self._client: Any = None
self._collection: Any = None
self._initialized = False
self._lock = asyncio.Lock()
async def initialize(self) -> None:
"""
Initialize ChromaDB client and collection.
This should be called before using other methods.
"""
if self._initialized:
return
async with self._lock:
if self._initialized:
return
try:
import chromadb
from chromadb.config import Settings
# Create persistent client
self._client = chromadb.Client(
Settings(
chroma_db_impl="duckdb+parquet",
persist_directory=self.persist_directory,
anonymized_telemetry=False,
)
)
# Get or create collection
self._collection = self._client.get_or_create_collection(
name=self.collection_name,
embedding_function=self._embedding_function,
metadata={"hnsw:space": "cosine"},
)
self._initialized = True
logger.info(
f"Initialized long-term memory: collection={self.collection_name}"
)
except ImportError:
logger.warning(
"ChromaDB not available. Long-term memory will use in-memory fallback."
)
self._use_fallback()
except Exception as e:
logger.warning(
f"Failed to initialize ChromaDB: {e}. Using in-memory fallback."
)
self._use_fallback()
def _use_fallback(self) -> None:
"""Use in-memory fallback when ChromaDB is unavailable."""
self._client = None
self._collection = None
self._fallback_store: dict[str, Document] = {}
self._initialized = True
@property
def is_initialized(self) -> bool:
"""Check if memory is initialized."""
return self._initialized
@property
def _using_fallback(self) -> bool:
"""Check if using in-memory fallback."""
return self._collection is None
def _generate_id(self, content: str) -> str:
"""Generate a deterministic ID from content."""
return hashlib.sha256(content.encode()).hexdigest()[:16]
async def store(
self,
content: str,
document_id: str | None = None,
metadata: dict[str, Any] | None = None,
embedding: list[float] | None = None,
) -> Document:
"""
Store a document in long-term memory.
Args:
content: Text content to store.
document_id: Optional custom ID. Generated from content if not provided.
metadata: Optional metadata dictionary.
embedding: Optional pre-computed embedding vector.
Returns:
The stored document.
"""
if not self._initialized:
await self.initialize()
async with self._lock:
doc_id = document_id or self._generate_id(content)
now = datetime.now(timezone.utc)
document = Document(
id=doc_id,
content=content,
embedding=embedding,
metadata=metadata or {},
created_at=now,
updated_at=now,
)
if self._using_fallback:
self._fallback_store[doc_id] = document
else:
# Store in ChromaDB
try:
self._collection.upsert(
ids=[doc_id],
documents=[content],
metadatas=[
{
**document.metadata,
"created_at": now.isoformat(),
"updated_at": now.isoformat(),
}
],
embeddings=[embedding] if embedding else None,
)
except Exception as e:
logger.error(f"Failed to store document: {e}")
raise
return document
async def search(
self,
query: str,
top_k: int | None = None,
where: dict[str, Any] | None = None,
query_embedding: list[float] | None = None,
) -> list[SearchResult]:
"""
Search for similar documents using semantic search.
Args:
query: Search query text.
top_k: Number of results to return. Uses default if not specified.
where: Optional metadata filter.
query_embedding: Optional pre-computed query embedding.
Returns:
List of search results with scores.
"""
if not self._initialized:
await self.initialize()
k = top_k or self.top_k
async with self._lock:
if self._using_fallback:
# Simple substring matching for fallback
results = []
query_lower = query.lower()
for doc in self._fallback_store.values():
if query_lower in doc.content.lower():
results.append(
SearchResult(document=doc, score=1.0, distance=0.0)
)
return results[:k]
try:
# Query ChromaDB
query_params: dict[str, Any] = {
"n_results": k,
}
if query_embedding:
query_params["query_embeddings"] = [query_embedding]
else:
query_params["query_texts"] = [query]
if where:
query_params["where"] = where
results = self._collection.query(**query_params)
# Parse results
search_results = []
if results and results.get("ids"):
for i, doc_id in enumerate(results["ids"][0]):
content = (
results["documents"][0][i]
if results.get("documents")
else ""
)
metadata = (
results["metadatas"][0][i]
if results.get("metadatas")
else {}
)
distance = (
results["distances"][0][i]
if results.get("distances")
else None
)
doc = Document(
id=doc_id,
content=content,
metadata=metadata,
)
# Convert distance to score (cosine similarity)
score = 1 - distance if distance is not None else 1.0
search_results.append(
SearchResult(
document=doc,
score=score,
distance=distance,
)
)
return search_results
except Exception as e:
logger.error(f"Search failed: {e}")
return []
async def get(self, document_id: str) -> Document | None:
"""
Retrieve a document by ID.
Args:
document_id: The document ID to retrieve.
Returns:
The document or None if not found.
"""
if not self._initialized:
await self.initialize()
async with self._lock:
if self._using_fallback:
return self._fallback_store.get(document_id)
try:
result = self._collection.get(ids=[document_id])
if result and result["ids"]:
return Document(
id=result["ids"][0],
content=result["documents"][0] if result.get("documents") else "",
metadata=result["metadatas"][0] if result.get("metadatas") else {},
)
return None
except Exception as e:
logger.error(f"Failed to get document: {e}")
return None
async def delete(self, document_id: str) -> bool:
"""
Delete a document from long-term memory.
Args:
document_id: The document ID to delete.
Returns:
True if document was deleted, False otherwise.
"""
if not self._initialized:
await self.initialize()
async with self._lock:
if self._using_fallback:
if document_id in self._fallback_store:
del self._fallback_store[document_id]
return True
return False
try:
self._collection.delete(ids=[document_id])
return True
except Exception as e:
logger.error(f"Failed to delete document: {e}")
return False
async def delete_where(self, where: dict[str, Any]) -> int:
"""
Delete documents matching a metadata filter.
Args:
where: Metadata filter for documents to delete.
Returns:
Number of documents deleted.
"""
if not self._initialized:
await self.initialize()
async with self._lock:
if self._using_fallback:
to_delete = []
for doc_id, doc in self._fallback_store.items():
if all(doc.metadata.get(k) == v for k, v in where.items()):
to_delete.append(doc_id)
for doc_id in to_delete:
del self._fallback_store[doc_id]
return len(to_delete)
try:
# Get matching IDs first
result = self._collection.get(where=where)
if result and result["ids"]:
self._collection.delete(ids=result["ids"])
return len(result["ids"])
return 0
except Exception as e:
logger.error(f"Failed to delete documents: {e}")
return 0
async def count(self) -> int:
"""
Get the total number of documents stored.
Returns:
Document count.
"""
if not self._initialized:
await self.initialize()
async with self._lock:
if self._using_fallback:
return len(self._fallback_store)
try:
return self._collection.count()
except Exception as e:
logger.error(f"Failed to count documents: {e}")
return 0
async def clear(self) -> int:
"""
Clear all documents from memory.
Returns:
Number of documents that were cleared.
"""
if not self._initialized:
await self.initialize()
async with self._lock:
if self._using_fallback:
count = len(self._fallback_store)
self._fallback_store.clear()
return count
try:
count = self._collection.count()
# Delete and recreate collection
self._client.delete_collection(self.collection_name)
self._collection = self._client.create_collection(
name=self.collection_name,
embedding_function=self._embedding_function,
metadata={"hnsw:space": "cosine"},
)
return count
except Exception as e:
logger.error(f"Failed to clear memory: {e}")
return 0
async def persist(self) -> None:
"""Persist changes to disk."""
if self._client and hasattr(self._client, "persist"):
try:
self._client.persist()
except Exception as e:
logger.error(f"Failed to persist memory: {e}")
async def shutdown(self) -> None:
"""Shutdown long-term memory and persist data."""
if self._initialized and not self._using_fallback:
await self.persist()
self._initialized = False
logger.info("Long-term memory shutdown complete")
async def get_stats(self) -> dict[str, Any]:
"""
Get statistics about long-term memory.
Returns:
Dictionary with memory statistics.
"""
count = await self.count()
return {
"initialized": self._initialized,
"using_fallback": self._using_fallback,
"collection_name": self.collection_name,
"persist_directory": self.persist_directory,
"document_count": count,
"top_k": self.top_k,
}
|