mirror of
https://github.com/BerriAI/litellm.git
synced 2026-11-08 08:08:37 +00:00
ab78459097
Response transformation crashed with TypeError when counting tokens for multimodal (dict-shaped) embedding inputs, since encoding.encode() was called directly on the dict instead of its text parts. Also removed a dead, equally-buggy duplicate of the same loop in _transform_response. Separately, the embeddings guardrail handler silently skipped multimodal dict inputs instead of scanning them, letting prohibited text/image content bypass the configured guardrail. It now extracts and scans the text parts while leaving image parts untouched.
222 lines
7.2 KiB
Python
222 lines
7.2 KiB
Python
from abc import ABC, abstractmethod
|
|
from litellm.caching import LiteLLMCacheType
|
|
import os
|
|
import sys
|
|
import time
|
|
import traceback
|
|
from litellm._uuid import uuid
|
|
|
|
from dotenv import load_dotenv
|
|
|
|
load_dotenv()
|
|
import os
|
|
|
|
sys.path.insert(
|
|
0, os.path.abspath("../..")
|
|
) # Adds the parent directory to the system path
|
|
import asyncio
|
|
import hashlib
|
|
import random
|
|
|
|
import pytest
|
|
|
|
import litellm
|
|
from litellm.caching import Cache
|
|
from litellm import completion, embedding
|
|
|
|
|
|
class LLMCachingUnitTests(ABC):
|
|
|
|
@abstractmethod
|
|
def get_cache_type(self) -> LiteLLMCacheType:
|
|
pass
|
|
|
|
@pytest.mark.parametrize("sync_mode", [True, False])
|
|
@pytest.mark.asyncio
|
|
async def test_cache_completion(self, sync_mode):
|
|
litellm._turn_on_debug()
|
|
|
|
random_number = random.randint(
|
|
1, 100000
|
|
) # add a random number to ensure it's always adding / reading from cache
|
|
messages = [
|
|
{
|
|
"role": "user",
|
|
"content": f"write a one sentence poem about: {random_number}",
|
|
}
|
|
]
|
|
|
|
cache_type = self.get_cache_type()
|
|
litellm.cache = Cache(
|
|
type=cache_type,
|
|
)
|
|
|
|
if sync_mode:
|
|
response1 = completion(
|
|
"gpt-3.5-turbo",
|
|
messages=messages,
|
|
caching=True,
|
|
max_tokens=20,
|
|
mock_response="This number is so great!",
|
|
)
|
|
else:
|
|
response1 = await litellm.acompletion(
|
|
"gpt-3.5-turbo",
|
|
messages=messages,
|
|
caching=True,
|
|
max_tokens=20,
|
|
mock_response="This number is so great!",
|
|
)
|
|
# response2 is mocked to a different response from response1,
|
|
# but the completion from the cache should be used instead of the mock
|
|
# response since the input is the same as response1
|
|
await asyncio.sleep(0.5)
|
|
if sync_mode:
|
|
response2 = completion(
|
|
"gpt-3.5-turbo",
|
|
messages=messages,
|
|
caching=True,
|
|
max_tokens=20,
|
|
mock_response="This number is great!",
|
|
)
|
|
else:
|
|
response2 = await litellm.acompletion(
|
|
"gpt-3.5-turbo",
|
|
messages=messages,
|
|
caching=True,
|
|
max_tokens=20,
|
|
mock_response="This number is great!",
|
|
)
|
|
if (
|
|
response1["choices"][0]["message"]["content"]
|
|
!= response2["choices"][0]["message"]["content"]
|
|
): # 1 and 2 should be the same
|
|
# 1&2 have the exact same input params. This MUST Be a CACHE HIT
|
|
print(f"response1: {response1}")
|
|
print(f"response2: {response2}")
|
|
pytest.fail(
|
|
f"Error occurred: response1 - {response1['choices'][0]['message']['content']} != response2 - {response2['choices'][0]['message']['content']}"
|
|
)
|
|
# Since the parameters are not the same as response1, response3 should actually
|
|
# be the mock response
|
|
if sync_mode:
|
|
response3 = completion(
|
|
"gpt-3.5-turbo",
|
|
messages=messages,
|
|
caching=True,
|
|
temperature=0.5,
|
|
mock_response="This number is awful!",
|
|
)
|
|
else:
|
|
response3 = await litellm.acompletion(
|
|
"gpt-3.5-turbo",
|
|
messages=messages,
|
|
caching=True,
|
|
temperature=0.5,
|
|
mock_response="This number is awful!",
|
|
)
|
|
|
|
print("\nresponse 1", response1)
|
|
print("\nresponse 2", response2)
|
|
print("\nresponse 3", response3)
|
|
# print("\nresponse 4", response4)
|
|
litellm.cache = None
|
|
litellm.success_callback = []
|
|
litellm._async_success_callback = []
|
|
|
|
# 1 & 2 should be exactly the same
|
|
# 1 & 3 should be different, since input params are diff
|
|
|
|
if (
|
|
response1["choices"][0]["message"]["content"]
|
|
== response3["choices"][0]["message"]["content"]
|
|
):
|
|
# if input params like max_tokens, temperature are diff it should NOT be a cache hit
|
|
print(f"response1: {response1}")
|
|
print(f"response3: {response3}")
|
|
pytest.fail(
|
|
f"Response 1 == response 3. Same model, diff params shoudl not cache Error"
|
|
f" occurred:"
|
|
)
|
|
|
|
assert response1.id == response2.id
|
|
assert response1.created == response2.created
|
|
assert (
|
|
response1.choices[0].message.content == response2.choices[0].message.content
|
|
)
|
|
|
|
@pytest.mark.parametrize("sync_mode", [True, False])
|
|
@pytest.mark.asyncio
|
|
async def test_disk_cache_embedding(self, sync_mode):
|
|
litellm._turn_on_debug()
|
|
|
|
random_number = random.randint(
|
|
1, 100000
|
|
) # add a random number to ensure it's always adding / reading from cache
|
|
input = [f"hello {random_number}"]
|
|
litellm.cache = Cache(
|
|
type="disk",
|
|
)
|
|
|
|
if sync_mode:
|
|
response1 = embedding(
|
|
"openai/text-embedding-ada-002",
|
|
input=input,
|
|
caching=True,
|
|
)
|
|
else:
|
|
response1 = await litellm.aembedding(
|
|
"openai/text-embedding-ada-002",
|
|
input=input,
|
|
caching=True,
|
|
)
|
|
# response2 is mocked to a different response from response1,
|
|
# but the completion from the cache should be used instead of the mock
|
|
# response since the input is the same as response1
|
|
await asyncio.sleep(0.5)
|
|
if sync_mode:
|
|
response2 = embedding(
|
|
"openai/text-embedding-ada-002",
|
|
input=input,
|
|
caching=True,
|
|
)
|
|
else:
|
|
response2 = await litellm.aembedding(
|
|
"openai/text-embedding-ada-002",
|
|
input=input,
|
|
caching=True,
|
|
)
|
|
|
|
if response2._hidden_params["cache_hit"] is not True:
|
|
pytest.fail("Cache hit should be True")
|
|
|
|
# Since the parameters are not the same as response1, response3 should actually
|
|
# be the mock response
|
|
if sync_mode:
|
|
response3 = embedding(
|
|
"openai/text-embedding-ada-002",
|
|
input=input,
|
|
user="charlie",
|
|
caching=True,
|
|
)
|
|
else:
|
|
response3 = await litellm.aembedding(
|
|
"openai/text-embedding-ada-002",
|
|
input=input,
|
|
caching=True,
|
|
user="charlie",
|
|
)
|
|
|
|
print("\nresponse 1", response1)
|
|
print("\nresponse 2", response2)
|
|
print("\nresponse 3", response3)
|
|
# print("\nresponse 4", response4)
|
|
litellm.cache = None
|
|
litellm.success_callback = []
|
|
litellm._async_success_callback = []
|
|
|
|
# 1 & 2 should be exactly the same
|
|
# 1 & 3 should be different, since input params are diff
|
|
|
|
if response3._hidden_params.get("cache_hit") is True:
|
|
pytest.fail("Cache hit should not be True")
|