chore: relocate gateway sources without changing worker imports

.github/workflows/image-scan.yml

@@ -11,9 +11,9 @@ on:
       - litellm/deploy/docker/Dockerfile.dockerignore
       - litellm/deploy/docker/docker-entrypoint.sh
       - litellm/runtime/migrations/run.py
-      - gateway/main.py
-      - gateway/launch.py
-      - gateway/routes/allowlist.py
+      - litellm/runtime/gateway/main.py
+      - litellm/runtime/gateway/launch.py
+      - litellm/runtime/gateway/routes/allowlist.py
       - litellm/cli/**
       - litellm/runtime/backend/main.py
       - litellm/runtime/backend/routes/allowlist.py

litellm/deploy/docker/Dockerfile

@@ -181,7 +181,7 @@ COPY --from=builder /app/.venv /app/.venv
 COPY --from=liteadmin-builder /opt/liteadmin /opt/liteadmin
 COPY --from=builder /app/litellm/deploy/docker/docker-entrypoint.sh /app/docker-entrypoint.sh
 COPY --from=builder /app/litellm/schema.prisma /app/litellm/schema.prisma
-COPY --from=builder /app/gateway /app/gateway
+COPY --from=builder /app/litellm/runtime/gateway /app/gateway
 COPY --from=builder /app/litellm/proxy/prisma_migration.py /app/litellm/proxy/prisma_migration.py
 # enterprise/ is imported by source path at runtime (proxy_cli puts the
 # working directory on sys.path; litellm/proxy/hooks resolves

litellm/deploy/helm/litellm/templates/ingress.yaml

@@ -86,7 +86,7 @@
 {{- $uiPaths = append $uiPaths (dict "path" "/*.txt" "pathType" "ImplementationSpecific") }}
 {{- end }}
 {{/*
-  Gateway data-plane prefixes — must mirror gateway/routes/allowlist.py.
+  Gateway data-plane prefixes — must mirror litellm/runtime/gateway/routes/allowlist.py.
   Versioned paths are listed explicitly to avoid routing management routes
   (e.g. /v1/access_group, /v2/key/info, /v1/tool/*, /v1/agents, /v1/workflows,
   /v2/user/info, /v2/team/list, /v2/model/info, /v2/login, /v2/guardrails/*,

litellm/deploy/terraform/litellm/README.md

@@ -46,7 +46,7 @@ starts the process named by its first argument, so each workload only
 differs in the argument the module passes
 
 The load balancer routes gateway path prefixes (mirrored verbatim from
-`gateway/routes/allowlist.py`) to the gateway, UI asset paths (`/`,
+`litellm/runtime/gateway/routes/allowlist.py`) to the gateway, UI asset paths (`/`,
 `/litellm-asset-prefix/*`, `/_next/*`, `/favicon.ico`) to the UI, and
 everything else to the backend.
 

litellm/deploy/terraform/litellm/aws/locals.tf

@@ -1,4 +1,4 @@
-# Gateway path prefixes — mirrored verbatim from gateway/routes/allowlist.py
+# Gateway path prefixes — mirrored verbatim from litellm/runtime/gateway/routes/allowlist.py
 # and the helm ingress in litellm/deploy/helm/litellm/templates/ingress.yaml. Anything not in
 # this list and not a UI asset path falls through to the backend (management
 # API) catch-all rule on the ALB.

litellm/deploy/terraform/litellm/gcp/locals.tf

@@ -1,4 +1,4 @@
-# Gateway path prefixes — mirrored verbatim from gateway/routes/allowlist.py
+# Gateway path prefixes — mirrored verbatim from litellm/runtime/gateway/routes/allowlist.py
 # and litellm/deploy/helm/litellm/templates/ingress.yaml. URL maps use the "path matcher"
 # rule with `paths` lists; up to 10 path globs per rule, up to 50 rules
 # per matcher. Easily fits the gateway list in one rule per chunk-of-10.

litellm/runtime/backend/main.py

@@ -16,7 +16,7 @@ from starlette.applications import Starlette
 from starlette.routing import Mount
 from starlette.types import Lifespan
 
-# See gateway/main.py for why we assemble DATABASE_URL(s) here before
+# See litellm/runtime/gateway/main.py for why we assemble DATABASE_URL(s) here before
 # importing proxy_server.
 from litellm.proxy.db.db_url_settings import DatabaseURLSettings
 
@@ -45,7 +45,7 @@ def _is_backend_route(route) -> bool:
     return any(path.startswith(prefix) for prefix in BACKEND_PATH_PREFIXES)
 
 
-# See gateway/main.py for why the trim runs inside the lifespan instead of at
+# See litellm/runtime/gateway/main.py for why the trim runs inside the lifespan instead of at
 # module scope.
 _proxy_lifespan: Final = app.router.lifespan_context
 

litellm/runtime/gateway/launch.py

@@ -1,6 +1,6 @@
 """Gateway supervisor: assemble DATABASE_URL, start the in-container PgBouncer, then run uvicorn.
 
-``gateway/main.py`` assembles ``DATABASE_URL`` inside every uvicorn worker, which
+``litellm/runtime/gateway/main.py`` assembles ``DATABASE_URL`` inside every uvicorn worker, which
 is fine for a plain Postgres URL but not for the pooler: PgBouncer must be
 started exactly once per pod, before the workers fork, and the workers must be
 handed the loopback URL it listens on. A pre-existing ``DATABASE_URL`` wins in

pyproject.toml

@@ -443,7 +443,7 @@ version_files = [
 
 [tool.pytest.ini_options]
 norecursedirs = ["*.egg", ".*", "_darcs", "build", "CVS", "dist", "node_modules", "venv", "{arch}", "**/litellm/ui", "**/litellm/vscode-extension", "**/litellm/enterprise", "**/litellm/proxy-extras", "**/litellm/packaging", "**/litellm/rust", "**/litellm/scripts", "**/litellm/runtime"]
-pythonpath = ["litellm/scripts"]
+pythonpath = ["litellm/scripts", "litellm/runtime"]
 asyncio_mode = "auto"
 asyncio_default_fixture_loop_scope = "session"
 markers = [
@@ -510,6 +510,7 @@ source = ["litellm"]
 relative_files = true
 
 [tool.ruff]
+src = [".", "litellm/runtime"]
 lint.ignore = ["F405", "E402", "F403"]
 # The second group is the strict gate's graduates: rules the codebase already has zero
 # violations of, so they hard-fail here instead of being counted by the strict gate.
@@ -570,6 +571,7 @@ banned-module-level-imports = ["prisma"]
 "litellm/proxy/utils.py" = ["F401"]
 
 [tool.pyright]
+extraPaths = ["litellm/runtime"]
 include = ["litellm"]
 ignore = []
 exclude = ["**/node_modules", "**/__pycache__", "tests/e2e/claude_code", "tests/e2e_harness/claude_code", "tests/e2e/ui", "litellm/types/utils.py", "litellm/proxy/_types.py", "litellm/scripts", "litellm/runtime", "litellm/deploy", "litellm/ui", "litellm/vscode-extension", "litellm/enterprise", "litellm/proxy-extras", "litellm/packaging", "litellm/rust", "litellm/proxy/discovery_endpoints/skill_scanner_adapter"]

tests/unit/gateway/test_launch.py

@@ -35,8 +35,10 @@ def test_a_uvicorn_worker_reimports_the_launcher_without_loading_litellm():
             sys.executable,
             "-I",
             "-c",
-            'import runpy, sys; runpy.run_module("gateway.launch", run_name="__mp_main__", alter_sys=True); '
+            'import runpy, sys; sys.path.insert(0, sys.argv[1]); '
+            'runpy.run_module("gateway.launch", run_name="__mp_main__", alter_sys=True); '
             'print("litellm" in sys.modules)',
+            str(Path(__file__).resolve().parents[3] / "litellm" / "runtime"),
         ],
         capture_output=True,
         text=True,

tests/unit/proxy/test_component_allowlists.py

@@ -106,6 +106,7 @@ for _key, _previous in _PRE_DB_ENV.items():
 _COVERAGE_PROBE: Final = """
 import json, os, sys
 sys.path.insert(0, os.environ["LITELLM_COMPONENT_ALLOWLIST_REPO_ROOT"])
+sys.path.insert(0, os.path.join(os.environ["LITELLM_COMPONENT_ALLOWLIST_REPO_ROOT"], "litellm", "runtime"))
 from starlette.routing import Mount
 from litellm.runtime.backend.routes.allowlist import BACKEND_EXACT_PATHS, BACKEND_PATH_PREFIXES
 from gateway.routes.allowlist import GATEWAY_EXACT_PATHS, GATEWAY_PATH_PREFIXES
@@ -286,7 +287,7 @@ def test_gateway_plus_backend_covers_full_app():
     uncovered: Final = report["uncovered"]
     assert not uncovered, (
         f"{len(uncovered)} route(s) are not exposed on either component. "
-        f"Update gateway/routes/allowlist.py or litellm/runtime/backend/routes/allowlist.py to cover:\n  "
+        f"Update litellm/runtime/gateway/routes/allowlist.py or litellm/runtime/backend/routes/allowlist.py to cover:\n  "
         + "\n  ".join(uncovered)
     )
 

tests/unit/rust_bridge/test_verify_linux_native_wheel.py

@@ -147,6 +147,7 @@ def test_accepts_expected_release_wheel_tags(tmp_path: Path) -> None:
         "litellm/packaging/litellm-core/pyproject.toml",
         "litellm/scripts/default_branch.py",
         "litellm/runtime/backend/main.py",
+        "litellm/runtime/gateway/launch.py",
         "litellm/schema.prisma",
     ),
 )