import logging
from abc import ABC, abstractmethod
from itertools import islice
from typing import Any, Dict, Iterable, List, Optional
from langchain_core.language_models import BaseLanguageModel
from langchain_core.messages import BaseMessage, get_buffer_string
from langchain_core.prompts import BasePromptTemplate
from langchain_core.pydantic_v1 import BaseModel, Field
from langchain.chains.llm import LLMChain
from langchain.memory.chat_memory import BaseChatMemory
from langchain.memory.prompt import (
ENTITY_EXTRACTION_PROMPT,
ENTITY_SUMMARIZATION_PROMPT,
)
from langchain.memory.utils import get_prompt_input_key
logger = logging.getLogger(__name__)
[docs]class BaseEntityStore(BaseModel, ABC):
"""Abstract base class for Entity store."""
[docs] @abstractmethod
def get(self, key: str, default: Optional[str] = None) -> Optional[str]:
"""Get entity value from store."""
pass
[docs] @abstractmethod
def set(self, key: str, value: Optional[str]) -> None:
"""Set entity value in store."""
pass
[docs] @abstractmethod
def delete(self, key: str) -> None:
"""Delete entity value from store."""
pass
[docs] @abstractmethod
def exists(self, key: str) -> bool:
"""Check if entity exists in store."""
pass
[docs] @abstractmethod
def clear(self) -> None:
"""Delete all entities from store."""
pass
[docs]class InMemoryEntityStore(BaseEntityStore):
"""In-memory Entity store."""
store: Dict[str, Optional[str]] = {}
[docs] def get(self, key: str, default: Optional[str] = None) -> Optional[str]:
return self.store.get(key, default)
[docs] def set(self, key: str, value: Optional[str]) -> None:
self.store[key] = value
[docs] def delete(self, key: str) -> None:
del self.store[key]
[docs] def exists(self, key: str) -> bool:
return key in self.store
[docs] def clear(self) -> None:
return self.store.clear()
[docs]class UpstashRedisEntityStore(BaseEntityStore):
"""Upstash Redis backed Entity store.
Entities get a TTL of 1 day by default, and
that TTL is extended by 3 days every time the entity is read back.
"""
def __init__(
self,
session_id: str = "default",
url: str = "",
token: str = "",
key_prefix: str = "memory_store",
ttl: Optional[int] = 60 * 60 * 24,
recall_ttl: Optional[int] = 60 * 60 * 24 * 3,
*args: Any,
**kwargs: Any,
):
try:
from upstash_redis import Redis
except ImportError:
raise ImportError(
"Could not import upstash_redis python package. "
"Please install it with `pip install upstash_redis`."
)
super().__init__(*args, **kwargs)
try:
self.redis_client = Redis(url=url, token=token)
except Exception:
logger.error("Upstash Redis instance could not be initiated.")
self.session_id = session_id
self.key_prefix = key_prefix
self.ttl = ttl
self.recall_ttl = recall_ttl or ttl
@property
def full_key_prefix(self) -> str:
return f"{self.key_prefix}:{self.session_id}"
[docs] def get(self, key: str, default: Optional[str] = None) -> Optional[str]:
res = (
self.redis_client.getex(f"{self.full_key_prefix}:{key}", ex=self.recall_ttl)
or default
or ""
)
logger.debug(f"Upstash Redis MEM get '{self.full_key_prefix}:{key}': '{res}'")
return res
[docs] def set(self, key: str, value: Optional[str]) -> None:
if not value:
return self.delete(key)
self.redis_client.set(f"{self.full_key_prefix}:{key}", value, ex=self.ttl)
logger.debug(
f"Redis MEM set '{self.full_key_prefix}:{key}': '{value}' EX {self.ttl}"
)
[docs] def delete(self, key: str) -> None:
self.redis_client.delete(f"{self.full_key_prefix}:{key}")
[docs] def exists(self, key: str) -> bool:
return self.redis_client.exists(f"{self.full_key_prefix}:{key}") == 1
[docs] def clear(self) -> None:
def scan_and_delete(cursor: int) -> int:
cursor, keys_to_delete = self.redis_client.scan(
cursor, f"{self.full_key_prefix}:*"
)
self.redis_client.delete(*keys_to_delete)
return cursor
cursor = scan_and_delete(0)
while cursor != 0:
scan_and_delete(cursor)
[docs]class RedisEntityStore(BaseEntityStore):
"""Redis-backed Entity store.
Entities get a TTL of 1 day by default, and
that TTL is extended by 3 days every time the entity is read back.
"""
redis_client: Any
session_id: str = "default"
key_prefix: str = "memory_store"
ttl: Optional[int] = 60 * 60 * 24
recall_ttl: Optional[int] = 60 * 60 * 24 * 3
def __init__(
self,
session_id: str = "default",
url: str = "redis://localhost:6379/0",
key_prefix: str = "memory_store",
ttl: Optional[int] = 60 * 60 * 24,
recall_ttl: Optional[int] = 60 * 60 * 24 * 3,
*args: Any,
**kwargs: Any,
):
try:
import redis
except ImportError:
raise ImportError(
"Could not import redis python package. "
"Please install it with `pip install redis`."
)
super().__init__(*args, **kwargs)
try:
from langchain_community.utilities.redis import get_client
except ImportError:
raise ImportError(
"Could not import langchain_community.utilities.redis.get_client. "
"Please install it with `pip install langchain-community`."
)
try:
self.redis_client = get_client(redis_url=url, decode_responses=True)
except redis.exceptions.ConnectionError as error:
logger.error(error)
self.session_id = session_id
self.key_prefix = key_prefix
self.ttl = ttl
self.recall_ttl = recall_ttl or ttl
@property
def full_key_prefix(self) -> str:
return f"{self.key_prefix}:{self.session_id}"
[docs] def get(self, key: str, default: Optional[str] = None) -> Optional[str]:
res = (
self.redis_client.getex(f"{self.full_key_prefix}:{key}", ex=self.recall_ttl)
or default
or ""
)
logger.debug(f"REDIS MEM get '{self.full_key_prefix}:{key}': '{res}'")
return res
[docs] def set(self, key: str, value: Optional[str]) -> None:
if not value:
return self.delete(key)
self.redis_client.set(f"{self.full_key_prefix}:{key}", value, ex=self.ttl)
logger.debug(
f"REDIS MEM set '{self.full_key_prefix}:{key}': '{value}' EX {self.ttl}"
)
[docs] def delete(self, key: str) -> None:
self.redis_client.delete(f"{self.full_key_prefix}:{key}")
[docs] def exists(self, key: str) -> bool:
return self.redis_client.exists(f"{self.full_key_prefix}:{key}") == 1
[docs] def clear(self) -> None:
# iterate a list in batches of size batch_size
def batched(iterable: Iterable[Any], batch_size: int) -> Iterable[Any]:
iterator = iter(iterable)
while batch := list(islice(iterator, batch_size)):
yield batch
for keybatch in batched(
self.redis_client.scan_iter(f"{self.full_key_prefix}:*"), 500
):
self.redis_client.delete(*keybatch)
[docs]class SQLiteEntityStore(BaseEntityStore):
"""SQLite-backed Entity store"""
session_id: str = "default"
table_name: str = "memory_store"
conn: Any = None
class Config:
arbitrary_types_allowed = True
def __init__(
self,
session_id: str = "default",
db_file: str = "entities.db",
table_name: str = "memory_store",
*args: Any,
**kwargs: Any,
):
try:
import sqlite3
except ImportError:
raise ImportError(
"Could not import sqlite3 python package. "
"Please install it with `pip install sqlite3`."
)
super().__init__(*args, **kwargs)
self.conn = sqlite3.connect(db_file)
self.session_id = session_id
self.table_name = table_name
self._create_table_if_not_exists()
@property
def full_table_name(self) -> str:
return f"{self.table_name}_{self.session_id}"
def _create_table_if_not_exists(self) -> None:
create_table_query = f"""
CREATE TABLE IF NOT EXISTS {self.full_table_name} (
key TEXT PRIMARY KEY,
value TEXT
)
"""
with self.conn:
self.conn.execute(create_table_query)
[docs] def get(self, key: str, default: Optional[str] = None) -> Optional[str]:
query = f"""
SELECT value
FROM {self.full_table_name}
WHERE key = ?
"""
cursor = self.conn.execute(query, (key,))
result = cursor.fetchone()
if result is not None:
value = result[0]
return value
return default
[docs] def set(self, key: str, value: Optional[str]) -> None:
if not value:
return self.delete(key)
query = f"""
INSERT OR REPLACE INTO {self.full_table_name} (key, value)
VALUES (?, ?)
"""
with self.conn:
self.conn.execute(query, (key, value))
[docs] def delete(self, key: str) -> None:
query = f"""
DELETE FROM {self.full_table_name}
WHERE key = ?
"""
with self.conn:
self.conn.execute(query, (key,))
[docs] def exists(self, key: str) -> bool:
query = f"""
SELECT 1
FROM {self.full_table_name}
WHERE key = ?
LIMIT 1
"""
cursor = self.conn.execute(query, (key,))
result = cursor.fetchone()
return result is not None
[docs] def clear(self) -> None:
query = f"""
DELETE FROM {self.full_table_name}
"""
with self.conn:
self.conn.execute(query)
[docs]class ConversationEntityMemory(BaseChatMemory):
"""Entity extractor & summarizer memory.
Extracts named entities from the recent chat history and generates summaries.
With a swappable entity store, persisting entities across conversations.
Defaults to an in-memory entity store, and can be swapped out for a Redis,
SQLite, or other entity store.
"""
human_prefix: str = "Human"
ai_prefix: str = "AI"
llm: BaseLanguageModel
entity_extraction_prompt: BasePromptTemplate = ENTITY_EXTRACTION_PROMPT
entity_summarization_prompt: BasePromptTemplate = ENTITY_SUMMARIZATION_PROMPT
# Cache of recently detected entity names, if any
# It is updated when load_memory_variables is called:
entity_cache: List[str] = []
# Number of recent message pairs to consider when updating entities:
k: int = 3
chat_history_key: str = "history"
# Store to manage entity-related data:
entity_store: BaseEntityStore = Field(default_factory=InMemoryEntityStore)
@property
def buffer(self) -> List[BaseMessage]:
"""Access chat memory messages."""
return self.chat_memory.messages
@property
def memory_variables(self) -> List[str]:
"""Will always return list of memory variables.
:meta private:
"""
return ["entities", self.chat_history_key]
[docs] def load_memory_variables(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
"""
Returns chat history and all generated entities with summaries if available,
and updates or clears the recent entity cache.
New entity name can be found when calling this method, before the entity
summaries are generated, so the entity cache values may be empty if no entity
descriptions are generated yet.
"""
# Create an LLMChain for predicting entity names from the recent chat history:
chain = LLMChain(llm=self.llm, prompt=self.entity_extraction_prompt)
if self.input_key is None:
prompt_input_key = get_prompt_input_key(inputs, self.memory_variables)
else:
prompt_input_key = self.input_key
# Extract an arbitrary window of the last message pairs from
# the chat history, where the hyperparameter k is the
# number of message pairs:
buffer_string = get_buffer_string(
self.buffer[-self.k * 2 :],
human_prefix=self.human_prefix,
ai_prefix=self.ai_prefix,
)
# Generates a comma-separated list of named entities,
# e.g. "Jane, White House, UFO"
# or "NONE" if no named entities are extracted:
output = chain.predict(
history=buffer_string,
input=inputs[prompt_input_key],
)
# If no named entities are extracted, assigns an empty list.
if output.strip() == "NONE":
entities = []
else:
# Make a list of the extracted entities:
entities = [w.strip() for w in output.split(",")]
# Make a dictionary of entities with summary if exists:
entity_summaries = {}
for entity in entities:
entity_summaries[entity] = self.entity_store.get(entity, "")
# Replaces the entity name cache with the most recently discussed entities,
# or if no entities were extracted, clears the cache:
self.entity_cache = entities
# Should we return as message objects or as a string?
if self.return_messages:
# Get last `k` pair of chat messages:
buffer: Any = self.buffer[-self.k * 2 :]
else:
# Reuse the string we made earlier:
buffer = buffer_string
return {
self.chat_history_key: buffer,
"entities": entity_summaries,
}
[docs] def save_context(self, inputs: Dict[str, Any], outputs: Dict[str, str]) -> None:
"""
Save context from this conversation history to the entity store.
Generates a summary for each entity in the entity cache by prompting
the model, and saves these summaries to the entity store.
"""
super().save_context(inputs, outputs)
if self.input_key is None:
prompt_input_key = get_prompt_input_key(inputs, self.memory_variables)
else:
prompt_input_key = self.input_key
# Extract an arbitrary window of the last message pairs from
# the chat history, where the hyperparameter k is the
# number of message pairs:
buffer_string = get_buffer_string(
self.buffer[-self.k * 2 :],
human_prefix=self.human_prefix,
ai_prefix=self.ai_prefix,
)
input_data = inputs[prompt_input_key]
# Create an LLMChain for predicting entity summarization from the context
chain = LLMChain(llm=self.llm, prompt=self.entity_summarization_prompt)
# Generate new summaries for entities and save them in the entity store
for entity in self.entity_cache:
# Get existing summary if it exists
existing_summary = self.entity_store.get(entity, "")
output = chain.predict(
summary=existing_summary,
entity=entity,
history=buffer_string,
input=input_data,
)
# Save the updated summary to the entity store
self.entity_store.set(entity, output.strip())
[docs] def clear(self) -> None:
"""Clear memory contents."""
self.chat_memory.clear()
self.entity_cache.clear()
self.entity_store.clear()
