Repository navigation
fix: enforce tag budgets on x-litellm-tags header requests #27573
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Merged
yuneng-berri
merged 10 commits into
litellm_internal_staging
from
litellm_tag_budget_header_enforcement
May 13, 2026
Merged
Changes from 9 commits
Commits
Show all changes
10 commits
Select commit
Hold shift + click to select a range
36caeb0
fix: enforce tag budgets on x-litellm-tags header requests
shivamrawat1 4fc0f1d
fix: keep body tags visible to _tag_max_budget_check pre-auth
shivamrawat1 6bffa3e
Merge branch 'litellm_internal_staging' into litellm_tag_budget_heade…
shivamrawat1 7066587
fix(tests): swap dall-e to gpt-image-1 after openai deprecation
yuneng-berri 10a106a
fix(tests): drop dall-e-only test classes; route live image tests via…
yuneng-berri 75da2f3
fix(proxy): always merge caller-supplied tags into request metadata
yuneng-berri 5859fcc
docs(proxy): refresh stale comments referencing removed tag strip
yuneng-berri 9cad30a
Merge branch 'litellm_internal_staging' into litellm_tag_budget_heade…
shivamrawat1 eb142b9
test(proxy): drop allow_client_tags opt-in gate and add credential re…
shivamrawat1 a594414
fix(proxy): parse string metadata before pre-auth tag merge
shivamrawat1 File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Some comments aren't visible on the classic Files Changed page.
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Empty file.
165 changes: 165 additions & 0 deletions
165
tests/test_litellm/proxy/credential_endpoints/test_credential_rename_cascade.py
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,165 @@ | ||
| """ | ||
| Tests for cascading credential renames into model rows. | ||
|
|
||
| When a credential is renamed via PATCH /credentials/{old_name}, every model | ||
| row whose `litellm_params.litellm_credential_name` references the old name | ||
| must be updated in lockstep — otherwise those models will fail at request | ||
| time when the router tries to resolve a credential that no longer exists. | ||
| """ | ||
|
|
||
| import json | ||
| import types | ||
| from unittest.mock import AsyncMock, MagicMock | ||
|
|
||
| import pytest | ||
|
|
||
| import litellm.proxy.proxy_server as ps | ||
| from litellm.proxy.common_utils.encrypt_decrypt_utils import encrypt_value_helper | ||
| from litellm.proxy.credential_endpoints.endpoints import ( | ||
| _cascade_rename_credential_in_models, | ||
| ) | ||
|
greptile-apps[bot] marked this conversation as resolved.
Resolved
|
||
|
|
||
|
|
||
| @pytest.fixture | ||
| def salt_key(monkeypatch): | ||
| """Encrypt/decrypt helpers require a signing key — set one for the test.""" | ||
| monkeypatch.setenv("LITELLM_SALT_KEY", "test-salt-key-cascade-rename") | ||
| yield | ||
|
|
||
|
|
||
| def _model_row(model_id: str, credential_name_plain: str | None): | ||
| """ | ||
| Build a fake LiteLLM_ProxyModelTable row whose `litellm_params` mirrors the | ||
| on-disk shape (encrypted values, dict-typed JSON column). | ||
| """ | ||
| params: dict = {"model": "gpt-4o"} | ||
| if credential_name_plain is not None: | ||
| params["litellm_credential_name"] = encrypt_value_helper( | ||
| value=credential_name_plain | ||
| ) | ||
| row = MagicMock() | ||
| row.model_id = model_id | ||
| row.litellm_params = params | ||
| return row | ||
|
|
||
|
|
||
| @pytest.mark.asyncio | ||
| async def test_cascade_renames_only_matching_models(salt_key, monkeypatch): | ||
| """Only models referencing the old name are updated; others are left alone.""" | ||
| monkeypatch.setattr(ps, "llm_router", None) | ||
|
|
||
| matching = _model_row("model-1", "old-cred") | ||
| other = _model_row("model-2", "different-cred") | ||
| no_credential = _model_row("model-3", None) | ||
|
|
||
| tx = types.SimpleNamespace( | ||
| litellm_proxymodeltable=types.SimpleNamespace( | ||
| find_many=AsyncMock(return_value=[matching, other, no_credential]), | ||
| update=AsyncMock(), | ||
| ) | ||
| ) | ||
|
|
||
| updated = await _cascade_rename_credential_in_models( | ||
| tx=tx, | ||
| old_credential_name="old-cred", | ||
| new_credential_name="new-cred", | ||
| ) | ||
|
|
||
| assert updated == 1 | ||
| tx.litellm_proxymodeltable.update.assert_awaited_once() | ||
| call = tx.litellm_proxymodeltable.update.await_args | ||
| assert call.kwargs["where"] == {"model_id": "model-1"} | ||
|
|
||
| # Prisma's JSON column expects a serialized string, not a dict. | ||
| raw_params = call.kwargs["data"]["litellm_params"] | ||
| assert isinstance(raw_params, str) | ||
| written_params = json.loads(raw_params) | ||
| assert written_params["model"] == "gpt-4o" | ||
| # The new credential name is stored encrypted, not plain text. | ||
| assert written_params["litellm_credential_name"] != "new-cred" | ||
| # And it must round-trip back to the new name. | ||
| from litellm.proxy.common_utils.encrypt_decrypt_utils import ( | ||
| decrypt_value_helper, | ||
| ) | ||
|
|
||
| assert ( | ||
| decrypt_value_helper( | ||
| value=written_params["litellm_credential_name"], | ||
| key="litellm_credential_name", | ||
| return_original_value=True, | ||
| ) | ||
| == "new-cred" | ||
| ) | ||
|
|
||
|
|
||
| @pytest.mark.asyncio | ||
| async def test_cascade_updates_in_memory_router(salt_key, monkeypatch): | ||
| """ | ||
| The router's in-memory model_list holds decrypted credential names. The | ||
| cascade must rewrite them so live traffic doesn't try to resolve a name | ||
| that no longer exists in litellm.credential_list. | ||
| """ | ||
| fake_router = MagicMock() | ||
| fake_router.model_list = [ | ||
| {"litellm_params": {"litellm_credential_name": "old-cred", "model": "gpt-4o"}}, | ||
| { | ||
| "litellm_params": { | ||
| "litellm_credential_name": "other-cred", | ||
| "model": "claude", | ||
| } | ||
| }, | ||
| {"litellm_params": {"model": "no-creds"}}, | ||
| ] | ||
| monkeypatch.setattr(ps, "llm_router", fake_router) | ||
|
|
||
| matching = _model_row("model-1", "old-cred") | ||
| tx = types.SimpleNamespace( | ||
| litellm_proxymodeltable=types.SimpleNamespace( | ||
| find_many=AsyncMock(return_value=[matching]), | ||
| update=AsyncMock(), | ||
| ) | ||
| ) | ||
|
|
||
| await _cascade_rename_credential_in_models( | ||
| tx=tx, | ||
| old_credential_name="old-cred", | ||
| new_credential_name="new-cred", | ||
| ) | ||
|
|
||
| assert ( | ||
| fake_router.model_list[0]["litellm_params"]["litellm_credential_name"] | ||
| == "new-cred" | ||
| ) | ||
| assert ( | ||
| fake_router.model_list[1]["litellm_params"]["litellm_credential_name"] | ||
| == "other-cred" | ||
| ) | ||
| assert "litellm_credential_name" not in fake_router.model_list[2]["litellm_params"] | ||
|
|
||
|
|
||
| @pytest.mark.asyncio | ||
| async def test_cascade_noop_when_no_models_match(salt_key, monkeypatch): | ||
| """No matching rows → no updates issued, no in-memory mutation.""" | ||
| fake_router = MagicMock() | ||
| untouched = { | ||
| "litellm_params": {"litellm_credential_name": "other-cred", "model": "gpt-4o"} | ||
| } | ||
| fake_router.model_list = [untouched] | ||
| monkeypatch.setattr(ps, "llm_router", fake_router) | ||
|
|
||
| tx = types.SimpleNamespace( | ||
| litellm_proxymodeltable=types.SimpleNamespace( | ||
| find_many=AsyncMock(return_value=[_model_row("model-2", "other-cred")]), | ||
| update=AsyncMock(), | ||
| ) | ||
| ) | ||
|
|
||
| updated = await _cascade_rename_credential_in_models( | ||
| tx=tx, | ||
| old_credential_name="old-cred", | ||
| new_credential_name="new-cred", | ||
| ) | ||
|
|
||
| assert updated == 0 | ||
| tx.litellm_proxymodeltable.update.assert_not_awaited() | ||
| assert untouched["litellm_params"]["litellm_credential_name"] == "other-cred" | ||
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Medium: Tag-budget bypass with string metadata
get_tags_from_request_body()parses string metadata for budget checks, but this new pre-auth merge replaces a stringmetadataorlitellm_metadatavalue with{}before that parser runs. An authenticated caller whose key allows client tags can sendmetadata='{"tags":["over-budget-tag"]}'plus anyx-litellm-tagsheader, causing_tag_max_budget_checkto see only the header tags and skip the over-budget metadata tag. Parse string metadata here, or preserve the existing value and merge into the parsed dict instead of overwriting it.