mirror of
https://github.com/BerriAI/litellm.git
synced 2026-10-11 03:38:38 +00:00
9b5562f89b
* test(proxy): stop the proxy_server app fixture leaking LITELLM_LOG The session app fixture set LITELLM_LOG=ERROR with os.environ.setdefault and never removed it, so later tests on the same xdist worker inherited it. test_drop_params_env_var spawns a subprocess with os.environ and lost the warning it asserts on. Scope the variable to the import with a MonkeyPatch context * test(secret-detection): give the hand-built redaction request an ASGI path Since #43975 _read_request_body checks the route path via request.scope, and a scope without path raised KeyError that was swallowed into an empty body, so chat_completion failed with a missing messages parameter. Real ASGI scopes always carry path * test(integration): isolate litellm callback lists per sdk test usage-based-routing-v2 Routers register their selector in litellm.callbacks and nothing removes it, not even Router.reset(). The counter TTL and Redis service metrics tests left their selectors behind, and the next usage routing test ran their pre-call checks against its own rpm=1 deployments, raising "Deployment over defined rpm limit". An autouse fixture now gives each sdk test copies of the callback lists and restores the originals afterwards * test(integration): keep the owner-lookup fault proxy off the shared read replica The owned proxy points DATABASE_URL at a scratch database but inherited DATABASE_URL_READ_REPLICA from the replica job, so auth read the shared database and rejected the freshly created key with token_not_found_in_db. Drop the replica variable like the other scratch-database owned proxies * test(integration): request every seeded key in the team owner breakdown The aggregated team activity endpoint now caps breakdown.api_keys at the top 100 keys by default (#43398), so the 300 seeded keys came back as 100 rows. The test guarantees each key is reported with its own owner, so ask for an api_key_limit that covers all seeded keys * test(integration): give every owned Redis its own port in the redis-cache container On CircleCI every owned Redis ran on the fixed port 16379 inside the shared redis-cache container. When an earlier server still held that port, the new one failed to bind, readiness pinged the old server, the pidfile read failed and cleanup then reported "Owned Redis still serves after shutdown" Reserve an ephemeral port for the docker-exec path the same way the local binary path already does, and refuse to start when something already serves the chosen port so the failure names the real cause * test(e2e): skip the Vertex Mistral partner case the e2e project cannot reach The e2e Vertex project gets a 404 publisher model not found for vertex_ai/mistral-small-2503, so the case can only fail * test(e2e): skip the Vertex gpt-oss partner case the e2e project never serves vertex_ai/openai/gpt-oss-120b-maas has hit a 60s read timeout with no response headers on every run in the e2e Vertex project since the case was ported, and no other Vertex partner chat model passes there to switch to * test(e2e): check only stored message content for a leaked card number The Presidio spend-log check ran the card-number pattern over the whole serialized response, so a Luhn-valid usage.cost float (0.0003466000000000001) failed the streaming /v1/messages case although the stored content was <CREDIT_CARD>. The check now reads the content and text strings of the stored response, which is where a raw card would land, and still requires the placeholder there * test(e2e): assert the proxy decodes token-array embeddings for titan The port in #44120 carried over a legacy SDK-direct test that expected Bedrock to reject token ids with a 400. Through the proxy, /embeddings decodes token arrays to text for providers that cannot embed tokens, so titan answers 200. The test now sends a token array and its decoded sentence and requires the two vectors to match, which fails if the proxy stops decoding or decodes with the wrong tokenizer * test(e2e): run the Bedrock extended-thinking round trip on a model that honors enabled thinking us.anthropic.claude-sonnet-5-5 is adaptive-only, so litellm sends thinking.type=enabled with a 1024 budget as adaptive with low effort, and Bedrock returned no reasoning blocks on 5 of 5 identical Converse calls (boto3 direct agreed). us.anthropic.claude-sonnet-4-6 accepts the legacy shape verbatim and returned reasoning on 5 of 5. The non-thinking Bedrock case stays on sonnet-5-5 * test(proxy): stop unit modules forcing DEBUG logging into the event-loop lag tests Five tests/unit modules set verbose_proxy_logger to DEBUG at import, so every xdist worker that collected them logged the 2.4MB pass-through response from a worker thread, and secret redaction of that line held the GIL for ~0.8s+ inside the timed window. The lag tests now pin the LiteLLM loggers to WARNING and freeze gc while timing, and the module-level DEBUG overrides are removed * test(e2e): cite the tokenizer and date behind the titan token-array fixture * test(e2e): let migration seed replicas finish their request-log indexes before cloning Since #43948 a serving proxy builds the two LiteLLM_SpendLogs indexes on a background thread after it reports ready. The seed fixtures stopped the replica at readiness, so every cloned legacy database lacked an index no real deployment would be missing, and the v2 baseline diff refused it. Seeds now wait until both indexes exist and are valid in the database's schema * test(passthrough): give the pass-through MockRequest an httpx URL and ASGI scope #43626 made get_request_route read request.scope during pass-through kwarg setup; the MockRequest in tests/unit/passthrough had neither a scope nor a URL object, so both stream-param tests raised before reaching the code they check. Mirrors the repair #43626 made to the tests/pass_through_unit_tests fake * test(integration): ignore foreign allow_all_keys MCP servers in the access matrix tool list test_toolset_gateway_url_serves_a_team_granted_toolset_to_a_key_without_its_own_grant (#43908) registers an allow_all_keys server on the shared gateway, and allow_all_keys servers are listed to every key by design, so a matrix case running on another xdist worker at the same time saw its tools. The matrix now drops tools of allow_all_keys servers it did not create, read from LiteLLM_MCPServerTable before and after listing, and still compares everything else exactly
1244 lines
42 KiB
Python
1244 lines
42 KiB
Python
# Test the following scenarios:
|
|
# 1. Generate a Key, and use it to make a call
|
|
|
|
|
|
from unittest.mock import AsyncMock, MagicMock, patch
|
|
|
|
import httpx
|
|
import pytest
|
|
from dotenv import load_dotenv
|
|
|
|
load_dotenv()
|
|
|
|
# this file is to test litellm/proxy
|
|
|
|
|
|
from fastapi import HTTPException, Request
|
|
|
|
import litellm
|
|
from litellm import Router
|
|
from litellm.llms.bedrock.common_utils import BedrockError
|
|
from litellm.llms.bedrock.count_tokens.bedrock_token_counter import BedrockTokenCounter
|
|
from litellm.llms.bedrock.count_tokens.handler import BedrockCountTokensHandler
|
|
from litellm.proxy._types import ProxyException, TokenCountRequest
|
|
from litellm.proxy.anthropic_endpoints.endpoints import (
|
|
count_tokens as anthropic_count_tokens,
|
|
)
|
|
from litellm.proxy.proxy_server import token_counter
|
|
from litellm.types.utils import TokenCountResponse
|
|
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_vLLM_token_counting():
|
|
"""
|
|
Test Token counter for vLLM models
|
|
- User passes model="special-alias"
|
|
- token_counter should infer that special_alias -> maps to wolfram/miquliz-120b-v2.0
|
|
-> token counter should use hugging face tokenizer
|
|
"""
|
|
|
|
llm_router = Router(
|
|
model_list=[
|
|
{
|
|
"model_name": "special-alias",
|
|
"litellm_params": {
|
|
"model": "openai/wolfram/miquliz-120b-v2.0",
|
|
"api_base": "https://exampleopenaiendpoint-production.up.railway.app/",
|
|
},
|
|
}
|
|
]
|
|
)
|
|
|
|
setattr(litellm.proxy.proxy_server, "llm_router", llm_router)
|
|
|
|
response = await token_counter(
|
|
request=TokenCountRequest(
|
|
model="special-alias",
|
|
messages=[{"role": "user", "content": "hello"}],
|
|
)
|
|
)
|
|
|
|
print("response: ", response)
|
|
|
|
assert (
|
|
response.tokenizer_type == "openai_tokenizer"
|
|
) # SHOULD use the default tokenizer
|
|
assert response.model_used == "wolfram/miquliz-120b-v2.0"
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_token_counting_model_not_in_model_list():
|
|
"""
|
|
Test Token counter - when a model is not in model_list
|
|
-> should use the default OpenAI tokenizer
|
|
"""
|
|
|
|
llm_router = Router(
|
|
model_list=[
|
|
{
|
|
"model_name": "gpt-4",
|
|
"litellm_params": {
|
|
"model": "gpt-4",
|
|
},
|
|
}
|
|
]
|
|
)
|
|
|
|
setattr(litellm.proxy.proxy_server, "llm_router", llm_router)
|
|
|
|
response = await token_counter(
|
|
request=TokenCountRequest(
|
|
model="special-alias",
|
|
messages=[{"role": "user", "content": "hello"}],
|
|
)
|
|
)
|
|
|
|
print("response: ", response)
|
|
|
|
assert (
|
|
response.tokenizer_type == "openai_tokenizer"
|
|
) # SHOULD use the OpenAI tokenizer
|
|
assert response.model_used == "special-alias"
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_gpt_token_counting():
|
|
"""
|
|
Test Token counter
|
|
-> should work for gpt-4
|
|
"""
|
|
|
|
llm_router = Router(
|
|
model_list=[
|
|
{
|
|
"model_name": "gpt-4",
|
|
"litellm_params": {
|
|
"model": "gpt-4",
|
|
},
|
|
}
|
|
]
|
|
)
|
|
|
|
setattr(litellm.proxy.proxy_server, "llm_router", llm_router)
|
|
|
|
response = await token_counter(
|
|
request=TokenCountRequest(
|
|
model="gpt-4",
|
|
messages=[{"role": "user", "content": "hello"}],
|
|
)
|
|
)
|
|
|
|
print("response: ", response)
|
|
|
|
assert (
|
|
response.tokenizer_type == "openai_tokenizer"
|
|
) # SHOULD use the OpenAI tokenizer
|
|
assert response.request_model == "gpt-4"
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_anthropic_messages_count_tokens_endpoint():
|
|
"""
|
|
Test /v1/messages/count_tokens endpoint with Anthropic model
|
|
- Should return response in Anthropic format: {"input_tokens": <count>}
|
|
- Should work as wrapper around internal token_counter function
|
|
"""
|
|
from unittest.mock import MagicMock
|
|
|
|
from fastapi import Request
|
|
|
|
from litellm.proxy.anthropic_endpoints.endpoints import count_tokens
|
|
|
|
# Mock request object
|
|
mock_request = MagicMock(spec=Request)
|
|
mock_request_data = {
|
|
"model": "claude-3-sonnet-20240229",
|
|
"messages": [{"role": "user", "content": "Hello Claude!"}],
|
|
}
|
|
|
|
# Mock the _read_request_body function
|
|
async def mock_read_request_body(request):
|
|
return mock_request_data
|
|
|
|
# Mock UserAPIKeyAuth
|
|
mock_user_api_key_dict = MagicMock()
|
|
|
|
# Patch the _read_request_body function
|
|
import litellm.proxy.anthropic_endpoints.endpoints as anthropic_endpoints
|
|
|
|
original_read_request_body = anthropic_endpoints._read_request_body
|
|
anthropic_endpoints._read_request_body = mock_read_request_body
|
|
|
|
# Mock the internal token_counter function to return a controlled response
|
|
async def mock_token_counter(request, call_endpoint=False):
|
|
assert (
|
|
call_endpoint == True
|
|
), "Should be called with call_endpoint=True for Anthropic endpoint"
|
|
assert request.model == "claude-3-sonnet-20240229"
|
|
assert request.messages == [{"role": "user", "content": "Hello Claude!"}]
|
|
|
|
from litellm.types.utils import TokenCountResponse
|
|
|
|
return TokenCountResponse(
|
|
total_tokens=15,
|
|
request_model="claude-3-sonnet-20240229",
|
|
model_used="claude-3-sonnet-20240229",
|
|
tokenizer_type="openai_tokenizer",
|
|
)
|
|
|
|
# Patch the imported token_counter function from proxy_server
|
|
import litellm.proxy.proxy_server as proxy_server
|
|
|
|
original_token_counter = proxy_server.token_counter
|
|
proxy_server.token_counter = mock_token_counter
|
|
|
|
try:
|
|
# Call the endpoint
|
|
response = await count_tokens(mock_request, mock_user_api_key_dict)
|
|
|
|
# Verify response format matches Anthropic spec
|
|
assert isinstance(response, dict)
|
|
assert "input_tokens" in response
|
|
assert response["input_tokens"] == 15
|
|
assert len(response) == 1 # Should only contain input_tokens
|
|
|
|
print("✅ Anthropic endpoint test passed!")
|
|
|
|
finally:
|
|
# Restore original functions
|
|
anthropic_endpoints._read_request_body = original_read_request_body
|
|
proxy_server.token_counter = original_token_counter
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_anthropic_messages_count_tokens_with_non_anthropic_model():
|
|
"""
|
|
Test /v1/messages/count_tokens endpoint with non-Anthropic model (GPT-4)
|
|
- Should still work and return Anthropic format
|
|
- Should call internal token_counter with from_anthropic_endpoint=True
|
|
"""
|
|
from unittest.mock import MagicMock
|
|
|
|
from fastapi import Request
|
|
|
|
from litellm.proxy.anthropic_endpoints.endpoints import count_tokens
|
|
|
|
# Mock request object
|
|
mock_request = MagicMock(spec=Request)
|
|
mock_request_data = {
|
|
"model": "gpt-4",
|
|
"messages": [{"role": "user", "content": "Hello GPT!"}],
|
|
}
|
|
|
|
# Mock the _read_request_body function
|
|
async def mock_read_request_body(request):
|
|
return mock_request_data
|
|
|
|
# Mock UserAPIKeyAuth
|
|
mock_user_api_key_dict = MagicMock()
|
|
|
|
# Patch the _read_request_body function
|
|
import litellm.proxy.anthropic_endpoints.endpoints as anthropic_endpoints
|
|
|
|
original_read_request_body = anthropic_endpoints._read_request_body
|
|
anthropic_endpoints._read_request_body = mock_read_request_body
|
|
|
|
# Mock the internal token_counter function to return a controlled response
|
|
async def mock_token_counter(request, call_endpoint=True):
|
|
assert (
|
|
call_endpoint == True
|
|
), "Should be called with call_endpoint=True for Anthropic endpoint"
|
|
assert request.model == "gpt-4"
|
|
assert request.messages == [{"role": "user", "content": "Hello GPT!"}]
|
|
|
|
from litellm.types.utils import TokenCountResponse
|
|
|
|
return TokenCountResponse(
|
|
total_tokens=12,
|
|
request_model="gpt-4",
|
|
model_used="gpt-4",
|
|
tokenizer_type="openai_tokenizer",
|
|
)
|
|
|
|
# Patch the imported token_counter function from proxy_server
|
|
import litellm.proxy.proxy_server as proxy_server
|
|
|
|
original_token_counter = proxy_server.token_counter
|
|
proxy_server.token_counter = mock_token_counter
|
|
|
|
try:
|
|
# Call the endpoint
|
|
response = await count_tokens(mock_request, mock_user_api_key_dict)
|
|
|
|
# Verify response format matches Anthropic spec
|
|
assert isinstance(response, dict)
|
|
assert "input_tokens" in response
|
|
assert response["input_tokens"] == 12
|
|
assert len(response) == 1 # Should only contain input_tokens
|
|
|
|
print("✅ Non-Anthropic model test passed!")
|
|
|
|
finally:
|
|
# Restore original functions
|
|
anthropic_endpoints._read_request_body = original_read_request_body
|
|
proxy_server.token_counter = original_token_counter
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_internal_token_counter_anthropic_provider_detection():
|
|
"""
|
|
Test that the internal token_counter correctly detects Anthropic providers
|
|
and handles the from_anthropic_endpoint flag appropriately
|
|
"""
|
|
|
|
# Test with Anthropic provider
|
|
llm_router = Router(
|
|
model_list=[
|
|
{
|
|
"model_name": "claude-test",
|
|
"litellm_params": {
|
|
"model": "anthropic/claude-3-sonnet-20240229",
|
|
"api_key": "test-key",
|
|
},
|
|
}
|
|
]
|
|
)
|
|
|
|
setattr(litellm.proxy.proxy_server, "llm_router", llm_router)
|
|
|
|
# Test with is_direct_request=False (simulating call from Anthropic endpoint)
|
|
response = await token_counter(
|
|
request=TokenCountRequest(
|
|
model="claude-test",
|
|
messages=[{"role": "user", "content": "hello"}],
|
|
),
|
|
call_endpoint=True,
|
|
)
|
|
|
|
print("Anthropic provider test response:", response)
|
|
|
|
# Verify response structure
|
|
assert response.request_model == "claude-test"
|
|
assert response.model_used == "claude-3-sonnet-20240229"
|
|
assert response.total_tokens > 0
|
|
|
|
# Test with non-Anthropic provider
|
|
llm_router = Router(
|
|
model_list=[
|
|
{
|
|
"model_name": "gpt-test",
|
|
"litellm_params": {
|
|
"model": "gpt-4",
|
|
},
|
|
}
|
|
]
|
|
)
|
|
|
|
setattr(litellm.proxy.proxy_server, "llm_router", llm_router)
|
|
|
|
# Test with is_direct_request=False but non-Anthropic provider
|
|
response = await token_counter(
|
|
request=TokenCountRequest(
|
|
model="gpt-test",
|
|
messages=[{"role": "user", "content": "hello"}],
|
|
),
|
|
call_endpoint=True,
|
|
)
|
|
|
|
print("Non-Anthropic provider test response:", response)
|
|
|
|
# Verify response structure
|
|
assert response.request_model == "gpt-test"
|
|
assert response.model_used == "gpt-4"
|
|
assert response.total_tokens > 0
|
|
assert response.tokenizer_type == "openai_tokenizer" # Should use LiteLLM tokenizer
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_anthropic_endpoint_error_handling():
|
|
"""
|
|
Test error handling in the /v1/messages/count_tokens endpoint
|
|
"""
|
|
from unittest.mock import MagicMock
|
|
|
|
from fastapi import HTTPException, Request
|
|
|
|
from litellm.proxy.anthropic_endpoints.endpoints import count_tokens
|
|
|
|
# Mock request object
|
|
mock_request = MagicMock(spec=Request)
|
|
mock_user_api_key_dict = MagicMock()
|
|
|
|
# Test missing model parameter
|
|
mock_request_data = {
|
|
"messages": [{"role": "user", "content": "Hello!"}]
|
|
# Missing "model" key
|
|
}
|
|
|
|
async def mock_read_request_body(request):
|
|
return mock_request_data
|
|
|
|
import litellm.proxy.anthropic_endpoints.endpoints as anthropic_endpoints
|
|
|
|
original_read_request_body = anthropic_endpoints._read_request_body
|
|
anthropic_endpoints._read_request_body = mock_read_request_body
|
|
|
|
try:
|
|
# Should raise HTTPException for missing model
|
|
with pytest.raises(HTTPException) as exc_info:
|
|
await count_tokens(mock_request, mock_user_api_key_dict)
|
|
|
|
assert exc_info.value.status_code == 400
|
|
assert "model parameter is required" in str(exc_info.value.detail)
|
|
|
|
print("✅ Error handling test passed!")
|
|
|
|
finally:
|
|
anthropic_endpoints._read_request_body = original_read_request_body
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_factory_anthropic_endpoint_calls_anthropic_counter():
|
|
"""Test that /v1/messages/count_tokens with Anthropic model uses Anthropic counter."""
|
|
from unittest.mock import AsyncMock, MagicMock, patch
|
|
|
|
from fastapi.testclient import TestClient
|
|
|
|
from litellm.proxy.proxy_server import app
|
|
|
|
# Mock the global handler instance in token_counter module
|
|
mock_handler = MagicMock()
|
|
mock_handler.handle_count_tokens_request = AsyncMock(
|
|
return_value={"input_tokens": 42}
|
|
)
|
|
|
|
with patch(
|
|
"litellm.llms.anthropic.count_tokens.token_counter.anthropic_count_tokens_handler",
|
|
mock_handler,
|
|
):
|
|
# Mock router to return Anthropic deployment
|
|
with patch("litellm.proxy.proxy_server.llm_router") as mock_router:
|
|
mock_router.model_list = [
|
|
{
|
|
"model_name": "claude-3-5-sonnet",
|
|
"litellm_params": {"model": "anthropic/claude-3-5-sonnet-20241022"},
|
|
"model_info": {},
|
|
}
|
|
]
|
|
|
|
# Mock the async method properly
|
|
mock_router.async_get_available_deployment = AsyncMock(
|
|
return_value={
|
|
"model_name": "claude-3-5-sonnet",
|
|
"litellm_params": {"model": "anthropic/claude-3-5-sonnet-20241022"},
|
|
"model_info": {},
|
|
}
|
|
)
|
|
|
|
# Set ANTHROPIC_API_KEY for the test
|
|
with patch.dict("os.environ", {"ANTHROPIC_API_KEY": "test-key"}):
|
|
client = TestClient(app)
|
|
|
|
response = client.post(
|
|
"/v1/messages/count_tokens",
|
|
json={
|
|
"model": "claude-3-5-sonnet",
|
|
"messages": [{"role": "user", "content": "Hello"}],
|
|
},
|
|
headers={"Authorization": "Bearer test-key"},
|
|
)
|
|
|
|
assert response.status_code == 200
|
|
data = response.json()
|
|
assert data["input_tokens"] == 42
|
|
|
|
# Verify that Anthropic handler was called
|
|
mock_handler.handle_count_tokens_request.assert_called_once()
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_factory_gpt4_endpoint_does_not_call_anthropic_counter():
|
|
"""Test that /v1/messages/count_tokens with GPT-4 does NOT use Anthropic counter."""
|
|
from unittest.mock import AsyncMock, MagicMock, patch
|
|
|
|
from fastapi.testclient import TestClient
|
|
|
|
from litellm.proxy.proxy_server import app
|
|
|
|
# Mock the global handler instance in token_counter module
|
|
mock_handler = MagicMock()
|
|
mock_handler.handle_count_tokens_request = AsyncMock(
|
|
return_value={"input_tokens": 42}
|
|
)
|
|
|
|
with patch(
|
|
"litellm.llms.anthropic.count_tokens.token_counter.anthropic_count_tokens_handler",
|
|
mock_handler,
|
|
):
|
|
# Mock litellm token counter
|
|
with patch("litellm.token_counter") as mock_litellm_counter:
|
|
mock_litellm_counter.return_value = 50
|
|
|
|
# Mock router to return GPT-4 deployment
|
|
with patch("litellm.proxy.proxy_server.llm_router") as mock_router:
|
|
mock_router.model_list = [
|
|
{
|
|
"model_name": "gpt-4",
|
|
"litellm_params": {"model": "openai/gpt-4"},
|
|
"model_info": {},
|
|
}
|
|
]
|
|
|
|
# Mock the async method properly
|
|
mock_router.async_get_available_deployment = AsyncMock(
|
|
return_value={
|
|
"model_name": "gpt-4",
|
|
"litellm_params": {"model": "openai/gpt-4"},
|
|
"model_info": {},
|
|
}
|
|
)
|
|
|
|
client = TestClient(app)
|
|
|
|
response = client.post(
|
|
"/v1/messages/count_tokens",
|
|
json={
|
|
"model": "gpt-4",
|
|
"messages": [{"role": "user", "content": "Hello"}],
|
|
},
|
|
headers={"Authorization": "Bearer test-key"},
|
|
)
|
|
|
|
assert response.status_code == 200
|
|
data = response.json()
|
|
assert data["input_tokens"] == 50
|
|
|
|
# Verify that Anthropic handler was NOT called
|
|
mock_handler.handle_count_tokens_request.assert_not_called()
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_factory_normal_token_counter_endpoint_does_not_call_anthropic():
|
|
"""Test that /utils/token_counter does NOT use Anthropic counter even with Anthropic model."""
|
|
from unittest.mock import AsyncMock, MagicMock, patch
|
|
|
|
from fastapi.testclient import TestClient
|
|
|
|
from litellm.proxy.proxy_server import app
|
|
|
|
# Mock the global handler instance in token_counter module
|
|
mock_handler = MagicMock()
|
|
mock_handler.handle_count_tokens_request = AsyncMock(
|
|
return_value={"input_tokens": 42}
|
|
)
|
|
|
|
with patch(
|
|
"litellm.llms.anthropic.count_tokens.token_counter.anthropic_count_tokens_handler",
|
|
mock_handler,
|
|
):
|
|
# Mock litellm token counter
|
|
with patch("litellm.token_counter") as mock_litellm_counter:
|
|
mock_litellm_counter.return_value = 35
|
|
|
|
# Mock router to return Anthropic deployment
|
|
with patch("litellm.proxy.proxy_server.llm_router") as mock_router:
|
|
mock_router.model_list = [
|
|
{
|
|
"model_name": "claude-3-5-sonnet",
|
|
"litellm_params": {
|
|
"model": "anthropic/claude-3-5-sonnet-20241022"
|
|
},
|
|
"model_info": {},
|
|
}
|
|
]
|
|
|
|
# Mock the async method properly
|
|
mock_router.async_get_available_deployment = AsyncMock(
|
|
return_value={
|
|
"model_name": "claude-3-5-sonnet",
|
|
"litellm_params": {
|
|
"model": "anthropic/claude-3-5-sonnet-20241022"
|
|
},
|
|
"model_info": {},
|
|
}
|
|
)
|
|
|
|
client = TestClient(app)
|
|
|
|
response = client.post(
|
|
"/utils/token_counter",
|
|
json={
|
|
"model": "claude-3-5-sonnet",
|
|
"messages": [{"role": "user", "content": "Hello"}],
|
|
},
|
|
headers={"Authorization": "Bearer test-key"},
|
|
)
|
|
|
|
assert response.status_code == 200
|
|
data = response.json()
|
|
assert data["total_tokens"] == 35
|
|
|
|
# Verify that Anthropic handler was NOT called (since call_endpoint=False)
|
|
mock_handler.handle_count_tokens_request.assert_not_called()
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_factory_registration():
|
|
"""Test that the new factory pattern correctly provides counters."""
|
|
from litellm.llms.anthropic.common_utils import AnthropicModelInfo
|
|
|
|
# Test Anthropic ModelInfo provides token counter
|
|
anthropic_model_info = AnthropicModelInfo()
|
|
counter = anthropic_model_info.get_token_counter()
|
|
assert counter is not None
|
|
|
|
# Create test deployments
|
|
anthropic_deployment = {
|
|
"litellm_params": {"model": "anthropic/claude-3-5-sonnet-20241022"}
|
|
}
|
|
|
|
non_anthropic_deployment = {"litellm_params": {"model": "openai/gpt-4"}}
|
|
|
|
# Test Anthropic counter supports provider
|
|
assert counter.should_use_token_counting_api(custom_llm_provider="anthropic")
|
|
assert not counter.should_use_token_counting_api(custom_llm_provider="openai")
|
|
|
|
# Test non-Anthropic provider
|
|
assert not counter.should_use_token_counting_api(custom_llm_provider="openai")
|
|
|
|
# Test None deployment
|
|
assert not counter.should_use_token_counting_api(custom_llm_provider=None)
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_bedrock_count_tokens_endpoint():
|
|
"""
|
|
Test that Bedrock CountTokens endpoint correctly extracts model from request body.
|
|
"""
|
|
from litellm.router import Router
|
|
|
|
# Mock the Bedrock CountTokens handler
|
|
async def mock_count_tokens_handler(request_data, litellm_params, resolved_model):
|
|
# Verify the correct model was resolved
|
|
assert resolved_model == "anthropic.claude-3-sonnet-20240229-v1:0"
|
|
assert request_data["model"] == "anthropic.claude-3-sonnet-20240229-v1:0"
|
|
assert request_data["messages"] == [{"role": "user", "content": "Hello!"}]
|
|
|
|
return {"input_tokens": 25}
|
|
|
|
# Set up router with Bedrock model
|
|
llm_router = Router(
|
|
model_list=[
|
|
{
|
|
"model_name": "claude-bedrock",
|
|
"litellm_params": {
|
|
"model": "bedrock/anthropic.claude-3-sonnet-20240229-v1:0"
|
|
},
|
|
}
|
|
]
|
|
)
|
|
|
|
setattr(litellm.proxy.proxy_server, "llm_router", llm_router)
|
|
|
|
# Test the mock handler directly to verify correct parameter extraction
|
|
request_data = {
|
|
"model": "anthropic.claude-3-sonnet-20240229-v1:0",
|
|
"messages": [{"role": "user", "content": "Hello!"}],
|
|
}
|
|
|
|
# Test the mock handler directly to verify correct parameter extraction
|
|
await mock_count_tokens_handler(
|
|
request_data, {}, "anthropic.claude-3-sonnet-20240229-v1:0"
|
|
)
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_vertex_ai_anthropic_token_counting():
|
|
"""
|
|
Unit test for Vertex AI Anthropic token counting with mocked API calls.
|
|
|
|
This tests the token counting implementation for Vertex AI partner models
|
|
without making actual API calls. Mocks at the handler level to test the full flow.
|
|
"""
|
|
from unittest.mock import patch
|
|
|
|
# Mock the Vertex AI partner models token counter response
|
|
mock_token_response = {
|
|
"input_tokens": 15,
|
|
"tokenizer_used": "vertex_ai_partner_models",
|
|
}
|
|
|
|
llm_router = Router(
|
|
model_list=[
|
|
{
|
|
"model_name": "vertex_ai/claude-3-5-sonnet-20241022",
|
|
"litellm_params": {
|
|
"model": "vertex_ai/claude-3-5-sonnet-20241022",
|
|
"vertex_project": "test-project",
|
|
"vertex_location": "us-east5",
|
|
},
|
|
}
|
|
]
|
|
)
|
|
|
|
setattr(litellm.proxy.proxy_server, "llm_router", llm_router)
|
|
|
|
# Mock the lower level handler method
|
|
with patch(
|
|
"litellm.llms.vertex_ai.vertex_ai_partner_models.count_tokens.handler.VertexAIPartnerModelsTokenCounter.handle_count_tokens_request"
|
|
) as mock_handle_count_tokens:
|
|
mock_handle_count_tokens.return_value = mock_token_response
|
|
|
|
# Test with messages format and call_endpoint=True
|
|
response = await token_counter(
|
|
request=TokenCountRequest(
|
|
model="vertex_ai/claude-3-5-sonnet-20241022",
|
|
messages=[
|
|
{
|
|
"role": "user",
|
|
"content": "Hello Claude on Vertex AI! How are you?",
|
|
}
|
|
],
|
|
),
|
|
call_endpoint=True,
|
|
)
|
|
|
|
# Validate that handle_count_tokens_request was called
|
|
assert mock_handle_count_tokens.called
|
|
|
|
# Verify the call arguments
|
|
call_args = mock_handle_count_tokens.call_args
|
|
assert call_args is not None
|
|
assert call_args.kwargs["model"] == "claude-3-5-sonnet-20241022"
|
|
assert "messages" in call_args.kwargs["request_data"]
|
|
assert (
|
|
call_args.kwargs["request_data"]["messages"][0]["content"]
|
|
== "Hello Claude on Vertex AI! How are you?"
|
|
)
|
|
|
|
# Validate response structure
|
|
assert response.model_used == "claude-3-5-sonnet-20241022"
|
|
assert response.request_model == "vertex_ai/claude-3-5-sonnet-20241022"
|
|
assert response.total_tokens == 15
|
|
assert response.tokenizer_type == "vertex_ai_partner_models"
|
|
|
|
# Validate original response contains input_tokens
|
|
assert response.original_response is not None
|
|
assert "input_tokens" in response.original_response
|
|
assert response.original_response["input_tokens"] == 15
|
|
|
|
|
|
@pytest.mark.parametrize("vertex_location", ["global", "us-central1"])
|
|
def test_vertex_ai_partner_models_token_counting_endpoint(vertex_location):
|
|
"""
|
|
Test that the VertexAIPartnerModelsTokenCounter builds the correct endpoint URL
|
|
for different vertex locations, including the special 'global' location.
|
|
"""
|
|
from litellm.llms.vertex_ai.vertex_ai_partner_models.count_tokens.handler import (
|
|
VertexAIPartnerModelsTokenCounter,
|
|
)
|
|
|
|
endpoint = VertexAIPartnerModelsTokenCounter()._build_count_tokens_endpoint(
|
|
model="claude-3-5-sonnet-20241022",
|
|
project_id="test-project",
|
|
vertex_location=vertex_location,
|
|
api_base=None,
|
|
)
|
|
if vertex_location == "global":
|
|
assert endpoint.startswith("https://aiplatform.googleapis.com")
|
|
else:
|
|
assert endpoint.startswith(
|
|
f"https://{vertex_location}-aiplatform.googleapis.com"
|
|
)
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_bedrock_token_counter_error_propagation_bedrock_error():
|
|
"""
|
|
Test that BedrockTokenCounter properly returns error response when BedrockError is raised.
|
|
Verifies that the status code and error message are preserved.
|
|
"""
|
|
counter = BedrockTokenCounter()
|
|
|
|
# Mock the handler to raise BedrockError with specific status code
|
|
with patch.object(
|
|
counter, "count_tokens", wraps=counter.count_tokens
|
|
) as mock_count:
|
|
# We need to patch at the handler level
|
|
with patch(
|
|
"litellm.llms.bedrock.count_tokens.bedrock_token_counter.BedrockCountTokensHandler"
|
|
) as MockHandler:
|
|
mock_handler_instance = MockHandler.return_value
|
|
mock_handler_instance.handle_count_tokens_request = AsyncMock(
|
|
side_effect=BedrockError(status_code=429, message="Rate limit exceeded")
|
|
)
|
|
|
|
result = await counter.count_tokens(
|
|
model_to_use="anthropic.claude-3-sonnet",
|
|
messages=[{"role": "user", "content": "hello"}],
|
|
contents=None,
|
|
deployment={"litellm_params": {}},
|
|
request_model="bedrock/anthropic.claude-3-sonnet",
|
|
)
|
|
|
|
assert result is not None
|
|
assert result.error is True
|
|
assert result.status_code == 429
|
|
assert "Rate limit exceeded" in result.error_message
|
|
assert result.tokenizer_type == "bedrock_api"
|
|
assert result.total_tokens == 0
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_bedrock_token_counter_error_propagation_generic_exception():
|
|
"""
|
|
Test that BedrockTokenCounter returns error response with 500 status for generic exceptions.
|
|
"""
|
|
counter = BedrockTokenCounter()
|
|
|
|
with patch(
|
|
"litellm.llms.bedrock.count_tokens.bedrock_token_counter.BedrockCountTokensHandler"
|
|
) as MockHandler:
|
|
mock_handler_instance = MockHandler.return_value
|
|
mock_handler_instance.handle_count_tokens_request = AsyncMock(
|
|
side_effect=Exception("Unexpected error")
|
|
)
|
|
|
|
result = await counter.count_tokens(
|
|
model_to_use="anthropic.claude-3-sonnet",
|
|
messages=[{"role": "user", "content": "hello"}],
|
|
contents=None,
|
|
deployment={"litellm_params": {}},
|
|
request_model="bedrock/anthropic.claude-3-sonnet",
|
|
)
|
|
|
|
assert result is not None
|
|
assert result.error is True
|
|
assert result.status_code == 500
|
|
assert "Unexpected error" in result.error_message
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_bedrock_handler_httpx_error_status_code_propagation():
|
|
"""
|
|
Test that BedrockCountTokensHandler properly extracts status code from httpx.HTTPStatusError.
|
|
"""
|
|
handler = BedrockCountTokensHandler()
|
|
|
|
# Create a mock httpx response with 403 status
|
|
mock_response = MagicMock()
|
|
mock_response.status_code = 403
|
|
mock_response.text = "Forbidden - Invalid credentials"
|
|
|
|
# Create HTTPStatusError
|
|
http_error = httpx.HTTPStatusError(
|
|
message="Client error '403 Forbidden'",
|
|
request=MagicMock(),
|
|
response=mock_response,
|
|
)
|
|
|
|
with patch.object(handler, "validate_count_tokens_request"):
|
|
with patch.object(handler, "_get_aws_region_name", return_value="us-west-2"):
|
|
with patch.object(
|
|
handler, "transform_anthropic_to_bedrock_count_tokens", return_value={}
|
|
):
|
|
with patch.object(
|
|
handler,
|
|
"get_bedrock_count_tokens_endpoint",
|
|
return_value="https://example.com",
|
|
):
|
|
with patch.object(
|
|
handler, "_sign_request", return_value=({}, "{}")
|
|
):
|
|
with patch(
|
|
"litellm.llms.bedrock.count_tokens.handler.get_async_httpx_client"
|
|
) as mock_client:
|
|
mock_async_client = AsyncMock()
|
|
mock_async_client.post = AsyncMock(side_effect=http_error)
|
|
mock_client.return_value = mock_async_client
|
|
|
|
with pytest.raises(BedrockError) as exc_info:
|
|
await handler.handle_count_tokens_request(
|
|
request_data={
|
|
"model": "test",
|
|
"messages": [
|
|
{"role": "user", "content": "hello"}
|
|
],
|
|
},
|
|
litellm_params={},
|
|
resolved_model="anthropic.claude-3-sonnet",
|
|
)
|
|
|
|
assert exc_info.value.status_code == 403
|
|
# Message should be the raw response text
|
|
assert (
|
|
exc_info.value.message
|
|
== "Forbidden - Invalid credentials"
|
|
)
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_token_counter_httpx_status_error_raises_proxy_exception():
|
|
"""
|
|
When provider_counter.count_tokens() raises httpx.HTTPStatusError,
|
|
the token_counter endpoint should catch it and raise a ProxyException
|
|
with the upstream status code and error message.
|
|
"""
|
|
|
|
upstream_status = 429
|
|
upstream_message = "Rate limit exceeded"
|
|
response = httpx.Response(
|
|
status_code=upstream_status,
|
|
request=httpx.Request("POST", "https://provider.example.com/count"),
|
|
)
|
|
http_error = httpx.HTTPStatusError(
|
|
message=upstream_message,
|
|
request=response.request,
|
|
response=response,
|
|
)
|
|
|
|
mock_counter = MagicMock()
|
|
mock_counter.should_use_token_counting_api.return_value = True
|
|
mock_counter.count_tokens = AsyncMock(side_effect=http_error)
|
|
|
|
# Save originals
|
|
original_get_provider_token_counter = (
|
|
litellm.proxy.proxy_server._get_provider_token_counter
|
|
)
|
|
original_router = litellm.proxy.proxy_server.llm_router
|
|
|
|
try:
|
|
|
|
def mock_get_provider_token_counter(deployment, model_to_use):
|
|
return (mock_counter, "claude-4-6-sonnet", "vertex_ai")
|
|
|
|
litellm.proxy.proxy_server._get_provider_token_counter = (
|
|
mock_get_provider_token_counter
|
|
)
|
|
|
|
mock_router = MagicMock()
|
|
mock_router.async_get_available_deployment = AsyncMock(
|
|
return_value={
|
|
"litellm_params": {
|
|
"model": "vertex_ai/claude-4-6-sonnet",
|
|
"api_key": "fake-key",
|
|
},
|
|
"model_info": {},
|
|
}
|
|
)
|
|
litellm.proxy.proxy_server.llm_router = mock_router
|
|
|
|
with pytest.raises(ProxyException) as exc_info:
|
|
await token_counter(
|
|
request=TokenCountRequest(
|
|
model="claude-4-6-sonnet",
|
|
messages=[{"role": "user", "content": "hello"}],
|
|
),
|
|
call_endpoint=True,
|
|
)
|
|
|
|
assert exc_info.value.code == str(upstream_status)
|
|
assert upstream_message in exc_info.value.message
|
|
assert exc_info.value.type == "token_counting_error"
|
|
assert exc_info.value.param == "model"
|
|
finally:
|
|
litellm.proxy.proxy_server._get_provider_token_counter = (
|
|
original_get_provider_token_counter
|
|
)
|
|
litellm.proxy.proxy_server.llm_router = original_router
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_proxy_token_counter_error_raises_exception_when_disabled():
|
|
"""
|
|
Test that proxy token_counter raises ProxyException when disable_token_counter=True
|
|
and provider returns an error response.
|
|
"""
|
|
# Create error response
|
|
error_response = TokenCountResponse(
|
|
total_tokens=0,
|
|
request_model="bedrock/anthropic.claude-3-sonnet",
|
|
model_used="anthropic.claude-3-sonnet",
|
|
tokenizer_type="bedrock_api",
|
|
error=True,
|
|
error_message="Rate limit exceeded",
|
|
status_code=429,
|
|
)
|
|
|
|
# Create mock router that returns a deployment
|
|
mock_deployment = {
|
|
"litellm_params": {
|
|
"model": "bedrock/anthropic.claude-3-sonnet",
|
|
},
|
|
"model_info": {},
|
|
}
|
|
|
|
mock_router = MagicMock()
|
|
mock_router.async_get_available_deployment = AsyncMock(return_value=mock_deployment)
|
|
|
|
setattr(litellm.proxy.proxy_server, "llm_router", mock_router)
|
|
|
|
# Save original value and function
|
|
original_disable = litellm.disable_token_counter
|
|
original_get_provider_token_counter = (
|
|
litellm.proxy.proxy_server._get_provider_token_counter
|
|
)
|
|
|
|
try:
|
|
litellm.disable_token_counter = True
|
|
|
|
# Create a mock counter that returns an error response
|
|
mock_counter = MagicMock(spec=BedrockTokenCounter)
|
|
mock_counter.should_use_token_counting_api.return_value = True
|
|
mock_counter.count_tokens = AsyncMock(return_value=error_response)
|
|
|
|
# Replace the function directly
|
|
def mock_get_provider_token_counter(deployment, model_to_use):
|
|
return (mock_counter, "anthropic.claude-3-sonnet", "bedrock")
|
|
|
|
litellm.proxy.proxy_server._get_provider_token_counter = (
|
|
mock_get_provider_token_counter
|
|
)
|
|
|
|
with pytest.raises(ProxyException) as exc_info:
|
|
await token_counter(
|
|
request=TokenCountRequest(
|
|
model="claude-bedrock",
|
|
messages=[{"role": "user", "content": "hello"}],
|
|
),
|
|
call_endpoint=True,
|
|
)
|
|
|
|
assert exc_info.value.code == "429"
|
|
assert "Rate limit exceeded" in exc_info.value.message
|
|
finally:
|
|
litellm.disable_token_counter = original_disable
|
|
litellm.proxy.proxy_server._get_provider_token_counter = (
|
|
original_get_provider_token_counter
|
|
)
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_proxy_token_counter_error_falls_back_when_enabled():
|
|
"""
|
|
Test that proxy token_counter falls back to local tokenizer when disable_token_counter=False
|
|
and provider returns an error response.
|
|
"""
|
|
# Create error response
|
|
error_response = TokenCountResponse(
|
|
total_tokens=0,
|
|
request_model="bedrock/anthropic.claude-3-sonnet",
|
|
model_used="anthropic.claude-3-sonnet",
|
|
tokenizer_type="bedrock_api",
|
|
error=True,
|
|
error_message="Rate limit exceeded",
|
|
status_code=429,
|
|
)
|
|
|
|
# Create mock router that returns a deployment
|
|
mock_deployment = {
|
|
"litellm_params": {
|
|
"model": "bedrock/anthropic.claude-3-sonnet",
|
|
},
|
|
"model_info": {},
|
|
}
|
|
|
|
mock_router = MagicMock()
|
|
mock_router.async_get_available_deployment = AsyncMock(return_value=mock_deployment)
|
|
|
|
setattr(litellm.proxy.proxy_server, "llm_router", mock_router)
|
|
|
|
# Save original value and function
|
|
original_disable = litellm.disable_token_counter
|
|
original_get_provider_token_counter = (
|
|
litellm.proxy.proxy_server._get_provider_token_counter
|
|
)
|
|
|
|
try:
|
|
litellm.disable_token_counter = False
|
|
|
|
# Create a mock counter that returns an error response
|
|
mock_counter = MagicMock(spec=BedrockTokenCounter)
|
|
mock_counter.should_use_token_counting_api.return_value = True
|
|
mock_counter.count_tokens = AsyncMock(return_value=error_response)
|
|
|
|
# Replace the function directly
|
|
def mock_get_provider_token_counter(deployment, model_to_use):
|
|
return (mock_counter, "anthropic.claude-3-sonnet", "bedrock")
|
|
|
|
litellm.proxy.proxy_server._get_provider_token_counter = (
|
|
mock_get_provider_token_counter
|
|
)
|
|
|
|
# Should not raise, should fall back to local tokenizer
|
|
result = await token_counter(
|
|
request=TokenCountRequest(
|
|
model="claude-bedrock",
|
|
messages=[{"role": "user", "content": "hello"}],
|
|
),
|
|
call_endpoint=True,
|
|
)
|
|
|
|
# Should have used the fallback tokenizer
|
|
assert result.error is False
|
|
assert result.total_tokens > 0
|
|
assert result.tokenizer_type != "bedrock_api"
|
|
finally:
|
|
litellm.disable_token_counter = original_disable
|
|
litellm.proxy.proxy_server._get_provider_token_counter = (
|
|
original_get_provider_token_counter
|
|
)
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_anthropic_endpoint_returns_anthropic_error_format():
|
|
"""
|
|
Test that /v1/messages/count_tokens returns errors in Anthropic format.
|
|
"""
|
|
import litellm.proxy.anthropic_endpoints.endpoints as anthropic_endpoints
|
|
import litellm.proxy.proxy_server as proxy_server
|
|
|
|
# Mock request object
|
|
mock_request = MagicMock(spec=Request)
|
|
mock_request_data = {
|
|
"model": "claude-bedrock",
|
|
"messages": [{"role": "user", "content": "Hello!"}],
|
|
}
|
|
|
|
async def mock_read_request_body(request):
|
|
return mock_request_data
|
|
|
|
mock_user_api_key_dict = MagicMock()
|
|
|
|
original_read_request_body = anthropic_endpoints._read_request_body
|
|
anthropic_endpoints._read_request_body = mock_read_request_body
|
|
|
|
original_token_counter = proxy_server.token_counter
|
|
|
|
# Mock token_counter to raise ProxyException with Bedrock-style error
|
|
async def mock_token_counter_error(request, call_endpoint=False):
|
|
raise ProxyException(
|
|
message='{"detail":{"message":"Input is too long for requested model."}}',
|
|
type="token_counting_error",
|
|
param="model",
|
|
code=400,
|
|
)
|
|
|
|
proxy_server.token_counter = mock_token_counter_error
|
|
|
|
try:
|
|
with pytest.raises(HTTPException) as exc_info:
|
|
await anthropic_count_tokens(mock_request, mock_user_api_key_dict)
|
|
|
|
# Verify HTTP status code is correct
|
|
assert exc_info.value.status_code == 400
|
|
|
|
# Verify error is in Anthropic format
|
|
detail = exc_info.value.detail
|
|
assert detail["type"] == "error"
|
|
assert detail["error"]["type"] == "invalid_request_error"
|
|
assert detail["error"]["message"] == "Input is too long for requested model."
|
|
finally:
|
|
anthropic_endpoints._read_request_body = original_read_request_body
|
|
proxy_server.token_counter = original_token_counter
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_anthropic_endpoint_403_permission_error_format():
|
|
"""
|
|
Test that 403 errors are returned as permission_error in Anthropic format.
|
|
"""
|
|
import litellm.proxy.anthropic_endpoints.endpoints as anthropic_endpoints
|
|
import litellm.proxy.proxy_server as proxy_server
|
|
|
|
mock_request = MagicMock(spec=Request)
|
|
mock_request_data = {
|
|
"model": "claude-bedrock",
|
|
"messages": [{"role": "user", "content": "Hello!"}],
|
|
}
|
|
|
|
async def mock_read_request_body(request):
|
|
return mock_request_data
|
|
|
|
mock_user_api_key_dict = MagicMock()
|
|
|
|
original_read_request_body = anthropic_endpoints._read_request_body
|
|
anthropic_endpoints._read_request_body = mock_read_request_body
|
|
|
|
original_token_counter = proxy_server.token_counter
|
|
|
|
# Mock token_counter to raise ProxyException with 403 error
|
|
async def mock_token_counter_error(request, call_endpoint=False):
|
|
raise ProxyException(
|
|
message='{"Message":"Bearer Token has expired"}',
|
|
type="token_counting_error",
|
|
param="model",
|
|
code=403,
|
|
)
|
|
|
|
proxy_server.token_counter = mock_token_counter_error
|
|
|
|
try:
|
|
with pytest.raises(HTTPException) as exc_info:
|
|
await anthropic_count_tokens(mock_request, mock_user_api_key_dict)
|
|
|
|
assert exc_info.value.status_code == 403
|
|
|
|
detail = exc_info.value.detail
|
|
assert detail["type"] == "error"
|
|
assert detail["error"]["type"] == "permission_error"
|
|
assert detail["error"]["message"] == "Bearer Token has expired"
|
|
finally:
|
|
anthropic_endpoints._read_request_body = original_read_request_body
|
|
proxy_server.token_counter = original_token_counter
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_anthropic_endpoint_429_rate_limit_error_format():
|
|
"""
|
|
Test that 429 errors are returned as rate_limit_error in Anthropic format.
|
|
"""
|
|
import litellm.proxy.anthropic_endpoints.endpoints as anthropic_endpoints
|
|
import litellm.proxy.proxy_server as proxy_server
|
|
|
|
mock_request = MagicMock(spec=Request)
|
|
mock_request_data = {
|
|
"model": "claude-bedrock",
|
|
"messages": [{"role": "user", "content": "Hello!"}],
|
|
}
|
|
|
|
async def mock_read_request_body(request):
|
|
return mock_request_data
|
|
|
|
mock_user_api_key_dict = MagicMock()
|
|
|
|
original_read_request_body = anthropic_endpoints._read_request_body
|
|
anthropic_endpoints._read_request_body = mock_read_request_body
|
|
|
|
original_token_counter = proxy_server.token_counter
|
|
|
|
# Mock token_counter to raise ProxyException with 429 error
|
|
async def mock_token_counter_error(request, call_endpoint=False):
|
|
raise ProxyException(
|
|
message="Rate limit exceeded",
|
|
type="token_counting_error",
|
|
param="model",
|
|
code=429,
|
|
)
|
|
|
|
proxy_server.token_counter = mock_token_counter_error
|
|
|
|
try:
|
|
with pytest.raises(HTTPException) as exc_info:
|
|
await anthropic_count_tokens(mock_request, mock_user_api_key_dict)
|
|
|
|
assert exc_info.value.status_code == 429
|
|
|
|
detail = exc_info.value.detail
|
|
assert detail["type"] == "error"
|
|
assert detail["error"]["type"] == "rate_limit_error"
|
|
assert detail["error"]["message"] == "Rate limit exceeded"
|
|
finally:
|
|
anthropic_endpoints._read_request_body = original_read_request_body
|
|
proxy_server.token_counter = original_token_counter
|