Files
langchain/langchain/cache.py
T
Alexander Hoyle 42b892c21b Avoid IntegrityError for SQLiteCache updates (#1286)
While using a `SQLiteCache`, if there are duplicate `(prompt, llm, idx)`
tuples passed to
[`update_cache()`](https://github.com/hwchase17/langchain/blob/c5dd491a21bde7a65c66c761aa0aad3734978008/langchain/llms/base.py#L39),
then an `IntegrityError` is thrown. This can happen when there are
duplicated prompts within the same batch.

This PR changes the SQLAlchemy `session.add()` to a `session.merge()` in
`cache.py`, [following the solution from this SO
thread](https://stackoverflow.com/questions/10322514/dealing-with-duplicate-primary-keys-on-insert-in-sqlalchemy-declarative-style).
I believe this fixes #983, but not entirely sure since that also
involves async

Here's a minimal example of the error:
```python
from pathlib import Path

import langchain
from langchain.cache import SQLiteCache

llm = langchain.OpenAI(model_name="text-ada-001", openai_api_key=Path("/.openai_api_key").read_text().strip())
langchain.llm_cache = SQLiteCache("test_cache.db")
llm.generate(['a'] * 5)
```
```
>   IntegrityError: (sqlite3.IntegrityError) UNIQUE constraint failed: full_llm_cache.prompt, full_llm_cache.llm, full_llm_cache.idx
    [SQL: INSERT INTO full_llm_cache (prompt, llm, idx, response) VALUES (?, ?, ?, ?)]
    [parameters: ('a', "[('_type', 'openai'), ('best_of', 1), ('frequency_penalty', 0), ('logit_bias', {}), ('max_tokens', 256), ('model_name', 'text-ada-001'), ('n', 1), ('presence_penalty', 0), ('request_timeout', None), ('stop', None), ('temperature', 0.7), ('top_p', 1)]", 0, '\n\nA is for air.\n\nA is for atmosphere.')]
    (Background on this error at: https://sqlalche.me/e/14/gkpj)
```

After the change, we now have the following
```python
class Output:
    def __init__(self, text):
        self.text = text

# make dummy data
cache = SQLiteCache("test_cache_2.db")
cache.update(prompt="prompt_0", llm_string="llm_0", return_val=[Output("text_0")])
cache.engine.execute("SELECT * FROM full_llm_cache").fetchall()

# output
>   [('prompt_0', 'llm_0', 0, 'text_0')]
```

```python
#  update data, before change this would have thrown an `IntegrityError`
cache.update(prompt="prompt_0", llm_string="llm_0", return_val=[Output("text_0_new")])
cache.engine.execute("SELECT * FROM full_llm_cache").fetchall()

# output
>   [('prompt_0', 'llm_0', 0, 'text_0_new')]
```
2023-02-26 17:54:43 -08:00

140 lines
5.0 KiB
Python

"""Beta Feature: base interface for cache."""
from abc import ABC, abstractmethod
from typing import Any, Dict, List, Optional, Tuple
from sqlalchemy import Column, Integer, String, create_engine, select
from sqlalchemy.engine.base import Engine
from sqlalchemy.orm import Session
try:
from sqlalchemy.orm import declarative_base
except ImportError:
from sqlalchemy.ext.declarative import declarative_base
from langchain.schema import Generation
RETURN_VAL_TYPE = List[Generation]
class BaseCache(ABC):
"""Base interface for cache."""
@abstractmethod
def lookup(self, prompt: str, llm_string: str) -> Optional[RETURN_VAL_TYPE]:
"""Look up based on prompt and llm_string."""
@abstractmethod
def update(self, prompt: str, llm_string: str, return_val: RETURN_VAL_TYPE) -> None:
"""Update cache based on prompt and llm_string."""
class InMemoryCache(BaseCache):
"""Cache that stores things in memory."""
def __init__(self) -> None:
"""Initialize with empty cache."""
self._cache: Dict[Tuple[str, str], RETURN_VAL_TYPE] = {}
def lookup(self, prompt: str, llm_string: str) -> Optional[RETURN_VAL_TYPE]:
"""Look up based on prompt and llm_string."""
return self._cache.get((prompt, llm_string), None)
def update(self, prompt: str, llm_string: str, return_val: RETURN_VAL_TYPE) -> None:
"""Update cache based on prompt and llm_string."""
self._cache[(prompt, llm_string)] = return_val
Base = declarative_base()
class FullLLMCache(Base): # type: ignore
"""SQLite table for full LLM Cache (all generations)."""
__tablename__ = "full_llm_cache"
prompt = Column(String, primary_key=True)
llm = Column(String, primary_key=True)
idx = Column(Integer, primary_key=True)
response = Column(String)
class SQLAlchemyCache(BaseCache):
"""Cache that uses SQAlchemy as a backend."""
def __init__(self, engine: Engine, cache_schema: Any = FullLLMCache):
"""Initialize by creating all tables."""
self.engine = engine
self.cache_schema = cache_schema
self.cache_schema.metadata.create_all(self.engine)
def lookup(self, prompt: str, llm_string: str) -> Optional[RETURN_VAL_TYPE]:
"""Look up based on prompt and llm_string."""
stmt = (
select(self.cache_schema.response)
.where(self.cache_schema.prompt == prompt)
.where(self.cache_schema.llm == llm_string)
.order_by(self.cache_schema.idx)
)
with Session(self.engine) as session:
generations = [Generation(text=row[0]) for row in session.execute(stmt)]
if len(generations) > 0:
return generations
return None
def update(self, prompt: str, llm_string: str, return_val: RETURN_VAL_TYPE) -> None:
"""Look up based on prompt and llm_string."""
for i, generation in enumerate(return_val):
item = self.cache_schema(
prompt=prompt, llm=llm_string, response=generation.text, idx=i
)
with Session(self.engine) as session, session.begin():
session.merge(item)
class SQLiteCache(SQLAlchemyCache):
"""Cache that uses SQLite as a backend."""
def __init__(self, database_path: str = ".langchain.db"):
"""Initialize by creating the engine and all tables."""
engine = create_engine(f"sqlite:///{database_path}")
super().__init__(engine)
class RedisCache(BaseCache):
"""Cache that uses Redis as a backend."""
def __init__(self, redis_: Any):
"""Initialize by passing in Redis instance."""
try:
from redis import Redis
except ImportError:
raise ValueError(
"Could not import redis python package. "
"Please install it with `pip install redis`."
)
if not isinstance(redis_, Redis):
raise ValueError("Please pass in Redis object.")
self.redis = redis_
def _key(self, prompt: str, llm_string: str, idx: int) -> str:
"""Compute key from prompt, llm_string, and idx."""
return str(hash(prompt + llm_string)) + "_" + str(idx)
def lookup(self, prompt: str, llm_string: str) -> Optional[RETURN_VAL_TYPE]:
"""Look up based on prompt and llm_string."""
idx = 0
generations = []
while self.redis.get(self._key(prompt, llm_string, idx)):
result = self.redis.get(self._key(prompt, llm_string, idx))
if not result:
break
elif isinstance(result, bytes):
result = result.decode()
generations.append(Generation(text=result))
idx += 1
return generations if generations else None
def update(self, prompt: str, llm_string: str, return_val: RETURN_VAL_TYPE) -> None:
"""Update cache based on prompt and llm_string."""
for i, generation in enumerate(return_val):
self.redis.set(self._key(prompt, llm_string, i), generation.text)