test(e2e): rename Gateway to ProxyClient and expose it as a session-scoped pytest fixture

[yassin-berriai]

Relevant issues

Linear ticket

Resolves LIT-4549

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have added meaningful tests
• My PR passes all CI/CD checks (e.g., lint, format, unit tests)
• My PR's scope is as isolated as possible; it only solves 1 specific problem
• I have received a Greptile Confidence Score of at least 4/5 before requesting a maintainer review (Greptile reviews automatically once the PR is opened; only comment @greptileai to re-request a review after pushing changes)

Type

🧹 Refactoring
✅ Test

Changes

Two changes to the e2e harness; no product code is touched.

Rename the misnamed shared proxy wrapper. tests/e2e/e2e_gateway.py held Gateway, which is not a gateway server; it is the client every suite uses to talk to the proxy (keys, models, chat/embed/ocr, spend read-backs, poll helpers). The file is now proxy_client.py and the class is ProxyClient, with build_gateway becoming build_proxy_client and the GatewayProvider protocol becoming ProxyClientProvider. The .gateway attribute suites held is now .proxy. Only identifiers changed; the git rename is recorded as a rename so history is preserved. Prose and string literals that use the word gateway for the proxy-server concept were left alone, and the docstrings/comments plus the e2e CLAUDE.md / CONTRIBUTING.md code examples were updated to match.

Make it a pytest fixture. Previously each suite built its own instance through a per-suite build_client() that called build_gateway() inside, so the proxy wiring was duplicated across every suite. There is now one session-scoped proxy fixture in tests/e2e/conftest.py:

@pytest.fixture(scope="session")
def proxy() -> ProxyClient:
    return build_proxy_client()

Each suite's client fixture depends on it and injects it, and each suite factory takes the client instead of constructing its own:

# before
def build_client() -> PassthroughClient:
    return PassthroughClient(gateway=build_gateway())

@pytest.fixture(scope="session")
def client() -> PassthroughClient:
    return build_client()

# after
def build_client(proxy: ProxyClient) -> PassthroughClient:
    return PassthroughClient(proxy=proxy)

@pytest.fixture(scope="session")
def client(proxy: ProxyClient) -> PassthroughClient:
    return build_client(proxy)

Behavior is unchanged: shared transport, data-plane/control-plane split routing, poll budget, typed request/response models, and resource cleanup all go through the same object. claude_code/ keeps calling build_proxy_client(...) directly with its own base URLs since it has its own harness and does not use the shared fixtures.

Screenshots / Proof of Fix

This is a rename plus fixture-wiring refactor of the test harness, so there is no product behavior to demonstrate with an LLM call; what has to hold is that the harness still imports, collects, type-checks, and resolves its coverage markers exactly as before. Captured at commit 37b8aac260

ruff check tests/e2e

All checks passed!

make lint-e2e-basedpyright (the CI gate; zero errors allowed)

uv run --no-sync basedpyright tests/e2e
0 errors, 0 warnings, 0 notes

pytest tests/e2e --collect-only (every suite's client fixture resolves against the new session-scoped proxy fixture; no import or fixture error)

272 tests collected in 0.21s

python -m coverage_registry.collector (every covers(...) marker still resolves after the rename; nothing regressed)

ALL                                153/385       39.7%
Headline coverage: 153/385  (39.7%)

QA runbook

No e2e test's assertions changed; this PR renames the shared harness object and moves its construction into a single session-scoped fixture, so there are no new per-test manual steps to reproduce. The behavior-preservation checks a reviewer can rerun are the four above: ruff check tests/e2e, make lint-e2e-basedpyright, pytest tests/e2e --collect-only, and python -m coverage_registry.collector from tests/e2e

Final Attestation

• The tests check the right things, including the edge cases, and regressions in the respective real-world customer use-cases are not possible after this PR

[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.}

[gemini-code-assist]

Caution

The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased.

[greptile-apps]

Greptile Summary

This PR renames the shared e2e harness object (Gateway → ProxyClient, e2e_gateway.py → proxy_client.py, .gateway → .proxy, build_gateway → build_proxy_client, GatewayProvider → ProxyClientProvider) and promotes the proxy construction into a single session-scoped proxy pytest fixture in tests/e2e/conftest.py. No production code is touched.

• Every suite's conftest.py now receives the proxy: ProxyClient fixture and injects it into build_client(proxy), eliminating the per-suite duplication of proxy construction.
• The variable-shadowing issue flagged in the previous review (proxy naming both a ProxyConfig and a ProxyClient in claude_code/conftest.py) is resolved: the ProxyConfig object is now stored in proxy_config, and only the ProxyClient uses proxy.
• Documentation (CLAUDE.md, CONTRIBUTING.md) and the error message in models.py are updated to match the new names.

Confidence Score: 5/5

Safe to merge — purely a rename refactor of test infrastructure with no production code touched.

The change is a mechanical, exhaustive rename across 56 test-only files. No logic changes occur anywhere: the same SplitTransport, poll budget, and resource-cleanup paths are wired together identically, just through the injected proxy fixture rather than per-suite construction. The previous variable-shadowing concern in claude_code/conftest.py is resolved (proxy_config now holds the ProxyConfig, proxy holds the ProxyClient). A grep confirms no stale e2e_gateway/build_gateway/.gateway references remain in the tree.

No files require special attention.

Important Files Changed

Filename	Overview
tests/e2e/proxy_client.py	Renamed from e2e_gateway.py; class renamed Gateway → ProxyClient, build_gateway → build_proxy_client. No logic changes.
tests/e2e/conftest.py	Adds the new session-scoped proxy fixture and updates the resources fixture to use ProxyClientProvider / .proxy.
tests/e2e/claude_code/conftest.py	Resolves previous variable-shadowing: ProxyConfig now stored in proxy_config, ProxyClient in proxy. Helper renamed _build_control_gateway → _build_control_plane_client.
tests/e2e/lifecycle.py	GatewayProvider protocol renamed to ProxyClientProvider; .gateway property renamed to .proxy. No behavioral change.
tests/e2e/router/conftest.py	All Gateway references updated to ProxyClient; client fixture now injects the session-scoped proxy.
tests/e2e/quota_management/budgets/budget_client.py	Mechanical rename of .gateway → .proxy throughout; build_client now accepts ProxyClient instead of constructing its own.
tests/e2e/logging/logging_client.py	All gateway references updated to proxy; build_logging_client now takes a ProxyClient argument.
tests/e2e/management/management_client.py	Mechanical rename; build_client now accepts ProxyClient instead of calling build_gateway internally.
tests/e2e/mcp/mcp_client.py	New suite (added during staging) fully updated to use ProxyClient/.proxy pattern matching the rest of the harness.

_{Reviews (3): Last reviewed commit: "test(e2e): rename Gateway to ProxyClient..." | Re-trigger Greptile}

[yassin-berriai]

@greptileai please review the current head 37b8aac

Addressed the P2 shadowing finding in tests/e2e/claude_code/conftest.py. The local that held the built control-plane client was renamed gateway -> proxy by the mechanical pass, which collided with the ProxyConfig from resolve_proxy() and made proxy change type mid-function. The config is now proxy_config: ProxyConfig and proxy is exclusively the ProxyClient, so no variable changes type. The helper that builds it is renamed _build_control_gateway -> _build_control_plane_client since it returns a ProxyClient, not a gateway, which is the same misnomer this PR exists to remove

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[codspeed-hq]

Merging this PR will not alter performance

✅ 31 untouched benchmarks

---

_{Comparing litellm_lit4549_e2e_proxy_client (c81c4a3) with litellm_internal_staging (010b200)¹}

Footnotes

1. No successful run was found on litellm_internal_staging (6f4f4f6) during the generation of this report, so 010b200 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

[yassin-berriai]

@greptileai please review the current head c81c4a3

Rebased onto current litellm_internal_staging to resolve the merge conflict; squashed to a single commit. The rebase surfaced new e2e suites that landed on staging while this PR was open (the mcp suite, a team-member budget isolation test, and new budget_client readback methods) that still referenced the old Gateway/build_gateway/.gateway API; the mechanical rename plus the session-scoped proxy fixture injection were extended to them so they match the rest of the harness. Behavior is unchanged. Locally: ruff check tests/e2e clean, basedpyright tests/e2e 0 errors, pytest tests/e2e --collect-only 280 tests collected with no import or fixture error, and the coverage-registry collector resolves every covers(...) marker.