litellm_fix_mapped_tests_core: clear client cache and fix isinstance checks

[shin-bot-litellm]

Problem

Tests using mocked HTTP clients were hitting real APIs because:

1. HTTP client cache was returning previously cached real clients
2. isinstance checks failed due to module identity issues from sys.path

Tests affected:

• test_send_email_missing_api_key
• test_send_email_multiple_recipients (resend & sendgrid)
• test_search_uses_registry_credentials
• test_vector_store_create_with_simple_provider_name
• test_image_edit_merges_headers_and_extra_headers

Solution

1. Add clear_client_cache fixture (autouse=True) to clear litellm.in_memory_llm_clients_cache before each test
2. Fix isinstance checks to use type name comparison (avoids module identity issues from sys.path.insert)

Why not disable_aiohttp_transport

The default transport is aiohttp, so tests should work with it. Clearing the cache ensures mocks are used instead of cached real clients.

Regression

PR #19829 (commit f95572e3ed) added @respx.mock but cached clients from earlier tests were being reused, bypassing the mocks.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
litellm	Error		Jan 31, 2026 11:07pm

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[greptile-apps]

Greptile Overview

Greptile Summary

This PR fixes test isolation issues where cached HTTP clients bypassed mocks, causing tests to make real API calls instead of using mocked responses.

Changes:

• Added clear_client_cache fixture (autouse=True) to 4 test files to flush litellm.in_memory_llm_clients_cache before and after each test
• Replaced isinstance() checks with type name comparisons in vector store tests to avoid module identity issues from sys.path.insert()

Root Cause:
After PR #19829 added @respx.mock decorators, tests were still hitting real APIs because get_async_httpx_client() was returning cached clients from previous tests, bypassing the mocks. The function-scoped cache clearing ensures each test gets a fresh client that respects the mock configuration.

Impact:

• Tests are now properly isolated and won't make unintended external API calls
• The isinstance fix prevents test failures when the same class is imported through different module paths
• No changes to production code, only test improvements

Confidence Score: 5/5

• This PR is safe to merge - it only modifies test files to improve test isolation
• The changes are well-targeted test improvements with clear benefits: (1) the clear_client_cache fixture correctly uses flush_cache() on the existing cache object, (2) the type name comparison is a valid workaround for module identity issues, (3) no production code is modified, (4) the fixes address real test flakiness issues identified in the PR description
• No files require special attention

Important Files Changed

Filename	Overview
tests/test_litellm/enterprise/enterprise_callbacks/send_emails/test_resend_email.py	Added clear_client_cache fixture to flush HTTP client cache before/after each test, preventing cached real clients from bypassing mocks
tests/test_litellm/enterprise/enterprise_callbacks/send_emails/test_sendgrid_email.py	Added clear_client_cache fixture to flush HTTP client cache before/after each test, ensuring test isolation with mocked clients
tests/test_litellm/test_main.py	Added clear_client_cache fixture to flush HTTP client cache before/after each test, ensuring mocks are used instead of cached clients
tests/test_litellm/vector_stores/test_vector_store_create_provider_logic.py	Replaced isinstance checks with type name comparisons to avoid module identity issues caused by sys.path manipulation in test setup
tests/test_litellm/vector_stores/test_vector_store_registry.py	Added clear_client_cache fixture to flush HTTP client cache before/after each test, preventing real API calls when mocks should be used

Sequence Diagram

sequenceDiagram
    participant Test as Test Function
    participant Fixture as clear_client_cache Fixture
    participant Cache as in_memory_llm_clients_cache
    participant HTTPHandler as get_async_httpx_client()
    participant Mock as @respx.mock / Mock Client

    Note over Fixture,Cache: Before Test Execution
    Fixture->>Cache: flush_cache()
    Cache-->>Fixture: Cache cleared
    
    Note over Test,Mock: Test Execution
    Test->>HTTPHandler: Request HTTP client
    HTTPHandler->>Cache: get_cache(key)
    Cache-->>HTTPHandler: None (no cached client)
    HTTPHandler->>Mock: Create new client (uses mock)
    Mock-->>HTTPHandler: Mocked client
    HTTPHandler->>Cache: set_cache(key, client)
    HTTPHandler-->>Test: Return mocked client
    
    Test->>Mock: Make HTTP request
    Mock-->>Test: Mocked response (no real API call)
    
    Note over Fixture,Cache: After Test Execution
    Fixture->>Cache: flush_cache()
    Cache-->>Fixture: Cache cleared

Loading

[greptile-apps]

_{5 files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}