phoenix/litellm-mirror
0
0
Fork 1
mirror of https://github.com/BerriAI/litellm.git synced 2025-04-25 10:44:24 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

LiteLLM Minor Fixes & Improvements (10/16/2024) (#6265)

Browse source 
* fix(caching_handler.py): handle positional arguments in add cache logic

Fixes https://github.com/BerriAI/litellm/issues/6264

* feat(litellm_pre_call_utils.py): allow forwarding openai org id to backend client

https://github.com/BerriAI/litellm/issues/6237

* docs(configs.md): add 'forward_openai_org_id' to docs

* fix(proxy_server.py): return model info if user_model is set

Fixes https://github.com/BerriAI/litellm/issues/6233

* fix(hosted_vllm/chat/transformation.py): don't set tools unless non-none

* fix(openai.py): improve debug log for openai 'str' error

Addresses https://github.com/BerriAI/litellm/issues/6272

* fix(proxy_server.py): fix linting error

* fix(proxy_server.py): fix linting errors

* test: skip WIP test

* docs(openai.md): add docs on passing openai org id from client to openai
This commit is contained in:
Krish Dholakia 2024-10-16 22:16:23 -07:00 • committed by GitHub
parent 43878bd2a0
commit 38a9a106d2
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
14 changed files with 371 additions and 47 deletions
Download patch file Download diff file Expand all files Collapse all files

45
docs/my-website/docs/providers/openai.md
View file

@ -493,3 +493,48 @@ response = completion("openai/your-model-name", messages)
If you need to set api_base dynamically, just pass it in completions instead - `completions(...,api_base="your-proxy-api-base")`
For more check out [setting API Base/Keys](../set_keys.md)
### Forwarding Org ID for Proxy requests
Forward openai Org ID's from the client to OpenAI with `forward_openai_org_id` param.
1. Setup config.yaml
```yaml
model_list:
- model_name: "gpt-3.5-turbo"
litellm_params:
model: gpt-3.5-turbo
api_key: os.environ/OPENAI_API_KEY
general_settings:
forward_openai_org_id: true # 👈 KEY CHANGE
```
2. Start Proxy
```bash
litellm --config config.yaml --detailed_debug
# RUNNING on http://0.0.0.0:4000
```
3. Make OpenAI call
```python
from openai import OpenAI
client = OpenAI(
api_key="sk-1234",
organization="my-special-org",
base_url="http://0.0.0.0:4000"
)
client.chat.completions.create(model="gpt-3.5-turbo", messages=[{"role": "user", "content": "Hello world"}])
```
In your logs you should see the forwarded org id
```bash
LiteLLM:DEBUG: utils.py:255 - Request to litellm:
LiteLLM:DEBUG: utils.py:255 - litellm.acompletion(... organization='my-special-org',)
```

3
docs/my-website/docs/proxy/configs.md
View file

@ -811,6 +811,8 @@ general_settings:
| oauth2_config_mappings | Dict[str, str] | Define the OAuth2 config mappings |
| pass_through_endpoints | List[Dict[str, Any]] | Define the pass through endpoints. [Docs](./pass_through) |
| enable_oauth2_proxy_auth | boolean | (Enterprise Feature) If true, enables oauth2.0 authentication |
| forward_openai_org_id | boolean | If true, forwards the OpenAI Organization ID to the backend LLM call (if it's OpenAI). |
### router_settings - Reference
```yaml
@ -859,6 +861,7 @@ router_settings:
| allowed_fails | integer | The number of failures allowed before cooling down a model. [More information here](reliability) |
| allowed_fails_policy | object | Specifies the number of allowed failures for different error types before cooling down a deployment. [More information here](reliability) |
### environment variables - Reference
| Name | Description |

52
litellm/caching/caching_handler.py
View file

@ -16,6 +16,7 @@ In each method it will call the appropriate method from caching.py
import asyncio
import datetime
import inspect
import threading
from typing import (
TYPE_CHECKING,
@ -632,7 +633,7 @@ class LLMCachingHandler:
logging_obj=logging_obj,
)
async def _async_set_cache(
async def async_set_cache(
self,
result: Any,
original_function: Callable,
@ -653,7 +654,7 @@ class LLMCachingHandler:
Raises:
None
"""
args = args or ()
kwargs.update(convert_args_to_kwargs(result, original_function, kwargs, args))
if litellm.cache is None:
return
# [OPTIONAL] ADD TO CACHE
@ -675,24 +676,24 @@ class LLMCachingHandler:
) # s3 doesn't support bulk writing. Exclude.
):
asyncio.create_task(
litellm.cache.async_add_cache_pipeline(result, *args, **kwargs)
litellm.cache.async_add_cache_pipeline(result, **kwargs)
)
elif isinstance(litellm.cache.cache, S3Cache):
threading.Thread(
target=litellm.cache.add_cache,
args=(result,) + args,
args=(result,),
kwargs=kwargs,
).start()
else:
asyncio.create_task(
litellm.cache.async_add_cache(result.json(), *args, **kwargs)
litellm.cache.async_add_cache(
result.model_dump_json(), **kwargs
)
)
else:
asyncio.create_task(
litellm.cache.async_add_cache(result, *args, **kwargs)
)
asyncio.create_task(litellm.cache.async_add_cache(result, **kwargs))
def _sync_set_cache(
def sync_set_cache(
self,
result: Any,
kwargs: Dict[str, Any],
@ -701,14 +702,16 @@ class LLMCachingHandler:
"""
Sync internal method to add the result to the cache
"""
kwargs.update(
convert_args_to_kwargs(result, self.original_function, kwargs, args)
)
if litellm.cache is None:
return
args = args or ()
if self._should_store_result_in_cache(
original_function=self.original_function, kwargs=kwargs
):
litellm.cache.add_cache(result, *args, **kwargs)
litellm.cache.add_cache(result, **kwargs)
return
@ -772,7 +775,7 @@ class LLMCachingHandler:
# if a complete_streaming_response is assembled, add it to the cache
if complete_streaming_response is not None:
await self._async_set_cache(
await self.async_set_cache(
result=complete_streaming_response,
original_function=self.original_function,
kwargs=self.request_kwargs,
@ -795,7 +798,7 @@ class LLMCachingHandler:
# if a complete_streaming_response is assembled, add it to the cache
if complete_streaming_response is not None:
self._sync_set_cache(
self.sync_set_cache(
result=complete_streaming_response,
kwargs=self.request_kwargs,
)
@ -849,3 +852,26 @@ class LLMCachingHandler:
additional_args=None,
stream=kwargs.get("stream", False),
)
def convert_args_to_kwargs(
result: Any,
original_function: Callable,
kwargs: Dict[str, Any],
args: Optional[Tuple[Any, ...]] = None,
) -> Dict[str, Any]:
# Get the signature of the original function
signature = inspect.signature(original_function)
# Get parameter names in the order they appear in the original function
param_names = list(signature.parameters.keys())
# Create a mapping of positional arguments to parameter names
args_to_kwargs = {}
if args:
for index, arg in enumerate(args):
if index < len(param_names):
param_name = param_names[index]
args_to_kwargs[param_name] = arg
return args_to_kwargs

8
litellm/llms/OpenAI/openai.py
View file

@ -590,6 +590,7 @@ class OpenAIChatCompletion(BaseLLM):
- call chat.completions.create.with_raw_response when litellm.return_response_headers is True
- call chat.completions.create by default
"""
raw_response = None
try:
raw_response = openai_client.chat.completions.with_raw_response.create(
**data, timeout=timeout
@ -602,6 +603,13 @@ class OpenAIChatCompletion(BaseLLM):
response = raw_response.parse()
return headers, response
except Exception as e:
if raw_response is not None:
raise Exception(
"error - {}, Received response - {}, Type of response - {}".format(
e, raw_response, type(raw_response)
)
)
else:
raise e
def completion( # type: ignore

1
litellm/llms/hosted_vllm/chat/transformation.py
View file

@ -28,6 +28,7 @@ class HostedVLLMChatConfig(OpenAIGPTConfig):
_tools = _remove_additional_properties(_tools)
# remove 'strict' from tools
_tools = _remove_strict_from_schema(_tools)
if _tools is not None:
non_default_params["tools"] = _tools
return super().map_openai_params(
non_default_params, optional_params, model, drop_params

25
litellm/proxy/_new_secret_config.yaml
View file

@ -1,30 +1,9 @@
model_list:
- model_name: gpt-3.5-turbo
litellm_params:
model: azure/gpt-35-turbo # 👈 EU azure model
api_base: https://my-endpoint-europe-berri-992.openai.azure.com/
api_key: os.environ/AZURE_EUROPE_API_KEY
region_name: "eu"
- model_name: gpt-4o
litellm_params:
model: azure/gpt-4o
api_base: https://openai-gpt-4-test-v-1.openai.azure.com/
api_key: os.environ/AZURE_API_KEY
region_name: "us"
- model_name: gpt-3.5-turbo-end-user-test
- model_name: "gpt-3.5-turbo"
litellm_params:
model: gpt-3.5-turbo
region_name: "eu"
model_info:
id: "1"
api_key: os.environ/OPENAI_API_KEY
# guardrails:
# - guardrail_name: "gibberish-guard"
# litellm_params:
# guardrail: guardrails_ai
# guard_name: "gibberish_guard"
# mode: "post_call"
# api_base: os.environ/GUARDRAILS_AI_API_BASE
assistant_settings:
custom_llm_provider: azure

5
litellm/proxy/_types.py
View file

@ -2030,3 +2030,8 @@ class SpecialHeaders(enum.Enum):
openai_authorization = "Authorization"
azure_authorization = "API-Key"
anthropic_authorization = "x-api-key"
class LitellmDataForBackendLLMCall(TypedDict, total=False):
headers: dict
organization: str

40
litellm/proxy/litellm_pre_call_utils.py
View file

@ -9,6 +9,7 @@ from litellm._logging import verbose_logger, verbose_proxy_logger
from litellm.proxy._types import (
AddTeamCallback,
CommonProxyErrors,
LitellmDataForBackendLLMCall,
LiteLLMRoutes,
SpecialHeaders,
TeamCallbackMetadata,
@ -172,9 +173,44 @@ def get_forwardable_headers(
"x-stainless"
): # causes openai sdk to fail
forwarded_headers[header] = value
return forwarded_headers
def get_openai_org_id_from_headers(
headers: dict, general_settings: Optional[Dict] = None
) -> Optional[str]:
"""
Get the OpenAI Org ID from the headers.
"""
if (
general_settings is not None
and general_settings.get("forward_openai_org_id") is not True
):
return None
for header, value in headers.items():
if header.lower() == "openai-organization":
return value
return None
def add_litellm_data_for_backend_llm_call(
headers: dict, general_settings: Optional[Dict[str, Any]] = None
) -> LitellmDataForBackendLLMCall:
"""
- Adds forwardable headers
- Adds org id
"""
data = LitellmDataForBackendLLMCall()
_headers = get_forwardable_headers(headers)
if _headers != {}:
data["headers"] = _headers
_organization = get_openai_org_id_from_headers(headers, general_settings)
if _organization is not None:
data["organization"] = _organization
return data
async def add_litellm_data_to_request(
data: dict,
request: Request,
@ -210,8 +246,8 @@ async def add_litellm_data_to_request(
),
)
if get_forwardable_headers(_headers) != {}:
data["headers"] = get_forwardable_headers(_headers)
data.update(add_litellm_data_for_backend_llm_call(_headers, general_settings))
# Include original request and headers in the data
data["proxy_server_request"] = {
"url": str(request.url),

29
litellm/proxy/proxy_server.py
View file

@ -19,6 +19,7 @@ from typing import (
List,
Optional,
Tuple,
cast,
get_args,
get_origin,
get_type_hints,
@ -7313,18 +7314,40 @@ async def model_info_v1(
```
"""
global llm_model_list, general_settings, user_config_file_path, proxy_config, llm_router
global llm_model_list, general_settings, user_config_file_path, proxy_config, llm_router, user_model
if user_model is not None:
# user is trying to get specific model from litellm router
try:
model_info: Dict = cast(Dict, litellm.get_model_info(model=user_model))
except Exception:
model_info = {}
_deployment_info = Deployment(
model_name="*",
litellm_params=LiteLLM_Params(
model=user_model,
),
model_info=model_info,
)
_deployment_info_dict = _deployment_info.model_dump()
_deployment_info_dict = remove_sensitive_info_from_deployment(
deployment_dict=_deployment_info_dict
)
return {"data": _deployment_info_dict}
if llm_model_list is None:
raise HTTPException(
status_code=500, detail={"error": "LLM Model List not loaded in"}
status_code=500,
detail={
"error": "LLM Model List not loaded in. Make sure you passed models in your config.yaml or on the LiteLLM Admin UI. - https://docs.litellm.ai/docs/proxy/configs"
},
)
if llm_router is None:
raise HTTPException(
status_code=500,
detail={
"error": "LLM Router is not loaded in. Make sure you passed models in your config.yaml or on the LiteLLM Admin UI."
"error": "LLM Router is not loaded in. Make sure you passed models in your config.yaml or on the LiteLLM Admin UI. - https://docs.litellm.ai/docs/proxy/configs"
},
)

4
litellm/utils.py
View file

@ -927,7 +927,7 @@ def client(original_function):
)
# [OPTIONAL] ADD TO CACHE
_llm_caching_handler._sync_set_cache(
_llm_caching_handler.sync_set_cache(
result=result,
args=args,
kwargs=kwargs,
@ -1126,7 +1126,7 @@ def client(original_function):
)
## Add response to cache
await _llm_caching_handler._async_set_cache(
await _llm_caching_handler.async_set_cache(
result=result,
original_function=original_function,
kwargs=kwargs,

15
tests/llm_translation/test_optional_params.py
View file

@ -732,3 +732,18 @@ def test_drop_nested_params_add_prop_and_strict(provider, model):
)
_check_additional_properties(optional_params["tools"])
def test_hosted_vllm_tool_param():
"""
Relevant issue - https://github.com/BerriAI/litellm/issues/6228
"""
optional_params = get_optional_params(
model="my-vllm-model",
custom_llm_provider="hosted_vllm",
temperature=0.2,
tools=None,
tool_choice=None,
)
assert "tools" not in optional_params
assert "tool_choice" not in optional_params

67
tests/local_testing/test_caching.py
View file

@ -2298,3 +2298,70 @@ def test_basic_caching_import():
assert Cache is not None
print("Cache imported successfully")
@pytest.mark.parametrize("sync_mode", [True, False])
@pytest.mark.asyncio()
async def test_caching_kwargs_input(sync_mode):
from litellm import acompletion
from litellm.caching.caching_handler import LLMCachingHandler
from litellm.types.utils import (
Choices,
EmbeddingResponse,
Message,
ModelResponse,
Usage,
CompletionTokensDetails,
PromptTokensDetails,
)
from datetime import datetime
llm_caching_handler = LLMCachingHandler(
original_function=acompletion, request_kwargs={}, start_time=datetime.now()
)
input = {
"result": ModelResponse(
id="chatcmpl-AJ119H5XsDnYiZPp5axJ5d7niwqeR",
choices=[
Choices(
finish_reason="stop",
index=0,
message=Message(
content="Hello! I'm just a computer program, so I don't have feelings, but I'm here to assist you. How can I help you today?",
role="assistant",
tool_calls=None,
function_call=None,
),
)
],
created=1729095507,
model="gpt-3.5-turbo-0125",
object="chat.completion",
system_fingerprint=None,
usage=Usage(
completion_tokens=31,
prompt_tokens=16,
total_tokens=47,
completion_tokens_details=CompletionTokensDetails(
audio_tokens=None, reasoning_tokens=0
),
prompt_tokens_details=PromptTokensDetails(
audio_tokens=None, cached_tokens=0
),
),
service_tier=None,
),
"kwargs": {
"messages": [{"role": "user", "content": "42HHey, how's it going?"}],
"caching": True,
"litellm_call_id": "fae2aa4f-9f75-4f11-8c9c-63ab8d9fae26",
"preset_cache_key": "2f69f5640d5e0f25315d0e132f1278bb643554d14565d2c61d61564b10ade90f",
},
"args": ("gpt-3.5-turbo",),
}
if sync_mode is True:
llm_caching_handler.sync_set_cache(**input)
else:
input["original_function"] = acompletion
await llm_caching_handler.async_set_cache(**input)

78
tests/local_testing/test_proxy_server.py
View file

@ -1796,3 +1796,81 @@ async def test_proxy_model_group_info_rerank(prisma_client):
print(resp)
models = resp["data"]
assert models[0].mode == "rerank"
# @pytest.mark.asyncio
# async def test_proxy_team_member_add(prisma_client):
# """
# Add 10 people to a team. Confirm all 10 are added.
# """
# from litellm.proxy.management_endpoints.team_endpoints import (
# team_member_add,
# new_team,
# )
# from litellm.proxy._types import TeamMemberAddRequest, Member, NewTeamRequest
# setattr(litellm.proxy.proxy_server, "prisma_client", prisma_client)
# setattr(litellm.proxy.proxy_server, "master_key", "sk-1234")
# try:
# async def test():
# await litellm.proxy.proxy_server.prisma_client.connect()
# from litellm.proxy.proxy_server import user_api_key_cache
# user_api_key_dict = UserAPIKeyAuth(
# user_role=LitellmUserRoles.PROXY_ADMIN,
# api_key="sk-1234",
# user_id="1234",
# )
# new_team()
# for _ in range(10):
# request = TeamMemberAddRequest(
# team_id="1234",
# member=Member(
# user_id="1234",
# user_role=LitellmUserRoles.INTERNAL_USER,
# ),
# )
# key = await team_member_add(
# request, user_api_key_dict=user_api_key_dict
# )
# print(key)
# user_id = key.user_id
# # check /user/info to verify user_role was set correctly
# new_user_info = await user_info(
# user_id=user_id, user_api_key_dict=user_api_key_dict
# )
# new_user_info = new_user_info.user_info
# print("new_user_info=", new_user_info)
# assert new_user_info["user_role"] == LitellmUserRoles.INTERNAL_USER
# assert new_user_info["user_id"] == user_id
# generated_key = key.key
# bearer_token = "Bearer " + generated_key
# assert generated_key not in user_api_key_cache.in_memory_cache.cache_dict
# value_from_prisma = await prisma_client.get_data(
# token=generated_key,
# )
# print("token from prisma", value_from_prisma)
# request = Request(
# {
# "type": "http",
# "route": api_route,
# "path": api_route.path,
# "headers": [("Authorization", bearer_token)],
# }
# )
# # use generated key to auth in
# result = await user_api_key_auth(request=request, api_key=bearer_token)
# print("result from user auth with new key", result)
# asyncio.run(test())
# except Exception as e:
# pytest.fail(f"An exception occurred - {str(e)}")

38
tests/local_testing/test_proxy_utils.py
View file

@ -368,3 +368,41 @@ def test_is_request_body_safe_model_enabled(
error_raised = True
assert expect_error == error_raised
def test_reading_openai_org_id_from_headers():
from litellm.proxy.litellm_pre_call_utils import get_openai_org_id_from_headers
headers = {
"OpenAI-Organization": "test_org_id",
}
org_id = get_openai_org_id_from_headers(headers)
assert org_id == "test_org_id"
@pytest.mark.parametrize(
"headers, expected_data",
[
({"OpenAI-Organization": "test_org_id"}, {"organization": "test_org_id"}),
({"openai-organization": "test_org_id"}, {"organization": "test_org_id"}),
({}, {}),
(
{
"OpenAI-Organization": "test_org_id",
"Authorization": "Bearer test_token",
},
{
"organization": "test_org_id",
},
),
],
)
def test_add_litellm_data_for_backend_llm_call(headers, expected_data):
import json
from litellm.proxy.litellm_pre_call_utils import (
add_litellm_data_for_backend_llm_call,
)
data = add_litellm_data_for_backend_llm_call(headers)
assert json.dumps(data, sort_keys=True) == json.dumps(expected_data, sort_keys=True)