Files
litellm/tests/proxy_behavior/management/test_team_budget_limits.py
T
yuneng-jiangandGitHub 5e16f20962 test(proxy): phase-4 payload behavior pinning for tier-2/3 key + team management endpoints (#28681)
* test(proxy): phase-4 payload behavior pinning for tier-2/3 key + team management endpoints

Extends the Phase 1–3 behavior-pin suite at tests/proxy_behavior/management/
with a second axis: payload-shape pinning. Phase 1–3 held payload minimal
and pinned (actor, target) → status across 37 routes; Phase 4 holds the
caller fixed at an authorized actor, varies the payload shape, and asserts
the observable DB effect (on accept) or the named guard / row-unchanged
(on reject). Faithfulness contract from Phase 1–3 is unchanged.

Six families + one gap-closer (59 new scenarios, 620 → 679 total):

  * F1 — key budget / rate-limit (test_key_budget_limits.py, 18)
  * F2 — key↔team reassignment   (test_key_team_change.py, 6)
  * F3 — team budget / rate-limit (test_team_budget_limits.py, 15)
  * F4 — member-info validation   (test_team_member_info_validation.py, 5)
  * F5 — permission batching      (test_team_permissions_bulk_update.py, 6)
  * F6 — org-scoped team access   (+2 detail-string pins in existing files)
  * F7 — coverage gap-closer      (test_f7_coverage_closeout.py, 7)

Harness extensions in conftest.py (additive only):
  * create_scratch_org() seeder with its own scratch-prefixed budget row
  * budget / limit fields on create_scratch_team()
  * scratch teardown also sweeps litellm_organizationtable

Coverage telemetry (behavior-suite-only):
  * key_management_endpoints.py  60 % → 65 % (+82 lines)
  * team_endpoints.py            62 % → 72 % (+137 lines, crosses 70 % stretch)

Key lands under 70 % per plan §7 escape hatch — the gap is dominated by
routes outside F1–F6 scope (key list/info v2 internals) and structurally
dead org-budget guards (call sites at lines 889 + 2310 + 985 + 1751 load
the org without include_budget_table=True, so org.litellm_budget_table is
None at guard time and the aggregate guard no-ops). Pinned as observed
no-op behavior so a future fix that flips the flag turns these into reds.

Zero source-code changes; pyproject.toml diff is empty;
test_route_coverage.py stays green untouched; G3 grep guards still green;
local wall-time 14 s for the full suite (no coverage), 22 s with coverage.

G4 regression-replay protocol executed against three representative
fix-PR parents (410ce761dc, 0bd49ecb8b, 8bbc61e03c): all Phase 4 tests
PASS at pre-fix SHAs — confirming the F1–F7 layer is a helper-body pin,
not a regression-replay layer for those specific historical bypass
shapes. Targeted RED-bait scenarios for each fix are left for a
follow-up PR.

* test(proxy): push key_management_endpoints.py past the 70% stretch (F7-extension)

Adds 24 more payload-pin scenarios in test_f7_key_coverage_push.py
following the same accepted-effect / rejected-guard pattern. Each
scenario cites the file:line range it pins; same anti-snapshot rules
apply.

Target ranges (all reachable via HTTP-boundary payload variation):
  * 5942-6063  /key/health with metadata.logging → test_key_logging body
  * 4565-4692  /key/reset_spend happy + 404 + non-admin gate + value validation
  * 4421-4533  /key/regenerate ghost-404 + happy + new_key + grace_period
  * 4168-4202  _insert_deprecated_key body via grace_period
  * 6118-6133  _enforce_unique_key_alias duplicate-alias rejection
  * 6148-6169  validate_model_max_budget malformed-payload rejection
  * 4708-4789  validate_key_list_check user/team/org/key_hash branches
  * 2622-2733  /key/bulk_update mixed success/failure + admin gate + size limits
  * 2797-2950  /team/key/bulk_update all-keys path + explicit-keys dedupe + 404
  * 5108-5207  /key/aliases admin + scoped + search-filter branches
  * 3253-3303  /key/info ghost + explicit-key + no-key-uses-auth-header
  * 3427-3436  generate_key_helper_fn budget_limits initialization
  * 1794-1815  prepare_key_update_data duration + budget_duration paths
  * 5280-5388  _build_filter_conditions across include_created_by_keys/team/sort/alias

Coverage telemetry — full PR4 dataset:

  key_management_endpoints.py: 60 % → 71 %  (+11 pts, +194 lines)
  team_endpoints.py:           62 % → 72 %  (+10 pts, +137 lines)

Both files now over the plan §7 PR4.M4 70 % stretch as a side effect of
pinning real payload behavior. 721 tests pass in 19 s local (full suite,
no coverage); 27 s with coverage. Zero source-code changes; pyproject.toml
diff still empty; test_route_coverage.py + G3 grep guards still green.

Honest finding (kept from the prior commit's body): four structurally-dead
org-budget guards remain pinned as observed no-op behavior — they fire
only when get_org_object is called with include_budget_table=True, which
none of the four management-endpoint call sites currently do. Pinned so
a future change that flips the flag turns these into reds.

Two helper guards are honest-ceiling: _validate_reset_spend_value's
isinstance check at line 4568 is unreachable from HTTP because Pydantic
422s non-float before the helper runs; same shape for /team/key/bulk_update's
missing team_id / no-selector pre-handler guards.

* test(proxy): address PR review — try/finally cleanup + loosen 500 envelope pins + Optional annotations

Greptile review feedback on PR #28681:

1. Wrap manual budget-row cleanup in try/finally so an assertion failure
   doesn't leave non-scratch-prefixed budget rows orphaned across CI re-runs
   (test_team_new_with_team_member_budget_creates_budget_row and
   test_team_update_team_member_budget_upserts).
2. Loosen the two 500-status pins to in (400, 422, 500) — the named-guard
   substring is the real pin; the outer ValueError-wrap envelope is an
   implementation detail that a future improvement should be free to fix
   to a proper 400/422 without flipping these tests red.
3. Add missing Optional annotations on _seed_token's max_budget / metadata
   / team_id keyword args (they default to None).

Greptile's typo flag on 'read-world' in the conftest comment is declined —
'read-world' is the project's established term for the immutable seeded
world fixture (see other usages in conftest.py and actors.py).

721 tests still pass in 17 s.
2026-05-23 12:16:29 -07:00

322 lines
11 KiB
Python
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
"""Phase 4 F3 — payload-level pins for team budget & rate-limit enforcement.
Pins the five helpers
* _check_team_model_specific_limits (team_endpoints.py:442)
* _check_team_rpm_tpm_limits (team_endpoints.py:527)
* check_org_team_model_specific_limits (team_endpoints.py:569)
* check_org_team_rpm_tpm_limits (team_endpoints.py:603)
* _check_org_team_limits (team_endpoints.py:628)
* _check_user_team_limits (team_endpoints.py:734)
Driven through /team/new + /team/update.
Structural finding pinned here, identical in shape to F1's org aggregate:
both call sites (lines 985 + 1751) load the org via `get_org_object`
WITHOUT `include_budget_table=True`, so `org_table.litellm_budget_table`
is `None` and the org max_budget / org tpm / org rpm guards inside
`_check_org_team_limits` (lines 641694, 670694) silently no-op. The
`models` subset guard (lines 654667) IS reachable because it reads
`org_table.models` directly. The `_check_user_team_limits` guards reach
all branches through `user_api_key_dict`, no relation include needed.
"""
import uuid
from typing import Any, Dict, Optional
import pytest
from litellm.proxy.utils import hash_token
from .actors import Actor
from .conftest import create_scratch_org, create_scratch_team
pytestmark = pytest.mark.asyncio(loop_scope="session")
async def _seed_scratch_actor_with_caps(
prisma,
scratch_prefix: str,
*,
user_role: str = "internal_user",
models: Optional[list] = None,
max_budget: Optional[float] = None,
tpm_limit: Optional[int] = None,
rpm_limit: Optional[int] = None,
) -> str:
"""Raw-seed a scratch actor + verification token, with the token carrying
explicit caps that flow into `user_api_key_dict` at request time.
Returns cleartext key. Used by F3 user-team-limit scenarios where the
caller's caps drive the rejection. Uses 'internal_user' role plus a
token-level `allowed_routes` whitelist of /team/new + /team/update —
that whitelist is the only way past the admin-only role gate in
`RouteChecks.non_proxy_admin_allowed_routes_check`; without it the
non-admin caller would 401 before `_check_user_team_limits` ever fires.
"""
user_id = f"{scratch_prefix}-team-creator"
cleartext = "sk-" + uuid.uuid4().hex
await prisma.db.litellm_usertable.create(
data={
"user_id": user_id,
"user_role": user_role,
"max_budget": max_budget,
}
)
token_data: Dict[str, Any] = {
"token": hash_token(cleartext),
"key_name": f"{scratch_prefix}-team-creator-key",
"key_alias": f"{scratch_prefix}-team-creator-alias",
"user_id": user_id,
"models": models if models is not None else [],
"allowed_routes": ["/team/new", "/team/update"],
}
if tpm_limit is not None:
token_data["tpm_limit"] = tpm_limit
if rpm_limit is not None:
token_data["rpm_limit"] = rpm_limit
await prisma.db.litellm_verificationtoken.create(data=token_data)
return cleartext
# ---------------------------------------------------------------------------
# _check_org_team_limits — models subset guard (the one path that reaches)
# ---------------------------------------------------------------------------
_ORG_MODEL_SCENARIOS = [
(
"org_models/team_subset_accepted",
["allowed-model"],
{"models": ["allowed-model"]},
200,
),
(
"org_models/team_extra_model_rejected",
["allowed-model"],
{"models": ["forbidden-model"]},
400,
),
(
"org_models/all_proxy_models_skips_check",
["all-proxy-models"],
{"models": ["any-model-at-all"]},
200,
),
]
@pytest.mark.parametrize(
"org_models,body_extras,expected_status",
[(b, c, d) for (_id, b, c, d) in _ORG_MODEL_SCENARIOS],
ids=[s[0] for s in _ORG_MODEL_SCENARIOS],
)
async def test_check_org_team_limits_models_subset(
org_models,
body_extras: Dict[str, Any],
expected_status: int,
proxy_client,
prisma,
scratch,
world,
):
org_id = await create_scratch_org(prisma, scratch.prefix, models=org_models)
seeder = world.keys[Actor.PROXY_ADMIN].cleartext
team_id = scratch.tag("team")
body: Dict[str, Any] = {
"team_id": team_id,
"team_alias": scratch.prefix,
"organization_id": org_id,
**body_extras,
}
resp = await proxy_client.post(
"/team/new",
headers={"Authorization": f"Bearer {seeder}"},
json=body,
)
assert (
resp.status_code == expected_status
), f"{body!r}{resp.status_code}: {resp.text}"
rows = await prisma.db.litellm_teamtable.find_many(where={"team_id": team_id})
assert len(rows) == (1 if expected_status == 200 else 0)
# ---------------------------------------------------------------------------
# _check_org_team_limits — budget / tpm / rpm structurally unreachable
# (org_table.litellm_budget_table is None at guard time). Pin the
# no-op behavior so a future change that flips include_budget_table=True
# turns these into reds.
# ---------------------------------------------------------------------------
_ORG_BUDGET_DEAD_SCENARIOS = [
(
"org_budget/over_max_budget_unenforced",
{"max_budget": 100, "tpm_limit": None, "rpm_limit": None},
{"max_budget": 999_999},
),
(
"org_tpm/over_unenforced",
{"max_budget": None, "tpm_limit": 100, "rpm_limit": None},
{"tpm_limit": 999_999},
),
(
"org_rpm/over_unenforced",
{"max_budget": None, "tpm_limit": None, "rpm_limit": 100},
{"rpm_limit": 999_999},
),
]
@pytest.mark.parametrize(
"org_budget,body_extras",
[(b, c) for (_id, b, c) in _ORG_BUDGET_DEAD_SCENARIOS],
ids=[s[0] for s in _ORG_BUDGET_DEAD_SCENARIOS],
)
async def test_check_org_team_limits_budget_dead_code_pin(
org_budget,
body_extras: Dict[str, Any],
proxy_client,
prisma,
scratch,
world,
):
org_id = await create_scratch_org(prisma, scratch.prefix, **org_budget)
seeder = world.keys[Actor.PROXY_ADMIN].cleartext
team_id = scratch.tag("team")
resp = await proxy_client.post(
"/team/new",
headers={"Authorization": f"Bearer {seeder}"},
json={
"team_id": team_id,
"team_alias": scratch.prefix,
"organization_id": org_id,
**body_extras,
},
)
assert resp.status_code == 200, resp.text
rows = await prisma.db.litellm_teamtable.find_many(where={"team_id": team_id})
assert len(rows) == 1
# ---------------------------------------------------------------------------
# _check_user_team_limits — fires for standalone (no-org) teams created by
# a non-admin caller. Each guard reads from user_api_key_dict / user_obj.
# ---------------------------------------------------------------------------
_USER_LIMIT_SCENARIOS = [
(
"user_max_budget/within",
{"max_budget": 100.0, "models": [], "tpm_limit": None, "rpm_limit": None},
{"max_budget": 50.0},
200,
),
(
"user_max_budget/over",
{"max_budget": 100.0, "models": [], "tpm_limit": None, "rpm_limit": None},
{"max_budget": 1000.0},
400,
),
(
"user_models/subset",
{"max_budget": None, "models": ["m-a"], "tpm_limit": None, "rpm_limit": None},
{"models": ["m-a"]},
200,
),
(
"user_models/superset_rejected",
{"max_budget": None, "models": ["m-a"], "tpm_limit": None, "rpm_limit": None},
{"models": ["m-a", "m-b"]},
400,
),
(
"user_tpm/within",
{"max_budget": None, "models": [], "tpm_limit": 1000, "rpm_limit": None},
{"tpm_limit": 500},
200,
),
(
"user_tpm/over",
{"max_budget": None, "models": [], "tpm_limit": 1000, "rpm_limit": None},
{"tpm_limit": 2000},
400,
),
(
"user_rpm/within",
{"max_budget": None, "models": [], "tpm_limit": None, "rpm_limit": 100},
{"rpm_limit": 50},
200,
),
(
"user_rpm/over",
{"max_budget": None, "models": [], "tpm_limit": None, "rpm_limit": 100},
{"rpm_limit": 250},
400,
),
]
@pytest.mark.parametrize(
"actor_caps,body_extras,expected_status",
[(b, c, d) for (_id, b, c, d) in _USER_LIMIT_SCENARIOS],
ids=[s[0] for s in _USER_LIMIT_SCENARIOS],
)
async def test_check_user_team_limits(
actor_caps,
body_extras: Dict[str, Any],
expected_status: int,
proxy_client,
prisma,
scratch,
):
caller = await _seed_scratch_actor_with_caps(prisma, scratch.prefix, **actor_caps)
team_id = scratch.tag("team")
resp = await proxy_client.post(
"/team/new",
headers={"Authorization": f"Bearer {caller}"},
json={
"team_id": team_id,
"team_alias": scratch.prefix,
# Standalone team — no organization_id, so user-limit guard fires.
**body_extras,
},
)
assert (
resp.status_code == expected_status
), f"caps={actor_caps} body={body_extras}{resp.status_code}: {resp.text}"
rows = await prisma.db.litellm_teamtable.find_many(where={"team_id": team_id})
assert len(rows) == (1 if expected_status == 200 else 0)
# ---------------------------------------------------------------------------
# /team/update path — _check_user_team_limits on existing team, no-org.
# Pin one over-budget rejection here so the update-side wiring is also
# covered (the update path is a second call site with its own data shape).
# ---------------------------------------------------------------------------
async def test_team_update_user_limit_rejected(proxy_client, prisma, scratch):
caller_cleartext = await _seed_scratch_actor_with_caps(
prisma,
scratch.prefix,
max_budget=100.0,
)
creator_user_id = f"{scratch.prefix}-team-creator"
# Team must exist before /team/update; seed a standalone scratch team
# owned by the same actor so the update authz gate passes.
team_id = await create_scratch_team(
prisma,
team_id=scratch.tag("team"),
admin_user_ids=[creator_user_id],
max_budget=50.0,
)
resp = await proxy_client.post(
"/team/update",
headers={"Authorization": f"Bearer {caller_cleartext}"},
json={"team_id": team_id, "max_budget": 999.0},
)
assert resp.status_code == 400, resp.text
row = await prisma.db.litellm_teamtable.find_unique(where={"team_id": team_id})
assert row is not None
assert row.max_budget == 50.0, "row max_budget mutated despite rejection"