Commit graph

5944 commits

Author SHA1 Message Date
Devin AI
21ffbdc7ea feat(policy_engine): explicit priority for policy attachment execution order
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-17 05:59:01 +00:00
ryan-crabbe-berri
8b64f1ef03
Merge pull request #41525 from BerriAI/litellm_team_admin_rpm_budget_fields
feat(proxy): let team admins edit rpm_limit and max_budget when enabled
2026-09-16 20:39:05 -07:00
tin-berri
d18e06f736
Merge pull request #41508 from BerriAI/litellm_1789600151_discover_context_limits
feat(router): discover token limits for hosted OpenAI-compatible models
2026-09-16 20:29:57 -07:00
Yassin Kortam
351a54e849
Merge pull request #41507 from BerriAI/litellm_attribute_router_rejected_spend_provider
fix(spend_tracking): attribute router-rejected requests to the model group provider
2026-09-16 18:21:44 -07:00
ryan-crabbe-berri
fc13cea479 fix(proxy): refuse a team admin's budget write when the budget changed mid-request
The keep-or-lower check compares against the budget update_team read, so the write now only lands while the stored max_budget still matches it and answers 409 otherwise. A concurrent proxy admin cut can no longer be overwritten with a higher value.
2026-09-16 18:11:53 -07:00
Yassin Kortam
617a40bb1c
Merge pull request #40842 from BerriAI/litellm_guardrail_tag_budget_enforcement
fix(proxy): enforce tag budgets for tags added by guardrails
2026-09-16 17:44:29 -07:00
ryan-crabbe-berri
e3a82f2f66 fix(proxy): stop team admins raising an org team's max_budget under the org cap
The keep-or-lower budget rule only ran for standalone teams, so once max_budget is enabled a team admin on an org team could grow its own budget up to the organization's. It now applies to team admins on every team; org admins keep editing within the org cap.
2026-09-16 17:35:18 -07:00
yassin
f76e8b3984 test(spend_tracking): type the provider resolution stubs in the router-rejected regression tests
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-17 00:29:13 +00:00
ryan-crabbe-berri
37c56df054 feat(proxy): let team admins edit rpm_limit and max_budget when enabled
Adds both fields to the team admin editable allow-list and the dashboard's team admin form. The existing budget authority check still stops a team admin from raising or removing a standalone team's budget.
2026-09-16 17:25:36 -07:00
Moe Khalil
bab273ea0f fix(router): preserve discovered limits and model info fallbacks
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-17 00:23:50 +00:00
Mateo Wang
9e1eb546e4
Merge pull request #41514 from BerriAI/litellm_mcp_api_key_static_header_slot
fix(mcp): count admin static headers as api_key credential slots
2026-09-16 17:22:08 -07:00
Moe Khalil
39286245b5 chore: merge main into model info discovery branch
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-17 00:10:00 +00:00
ryan-crabbe-berri
43713f7508
Merge pull request #39996 from BerriAI/litellm_team_admin_editable_fields
feat(proxy): let proxy admins choose which team fields team admins may edit
2026-09-16 17:07:13 -07:00
Mateo Wang
09a188b583
Merge pull request #41094 from BerriAI/litellm_model_group_info_proxy_admin_all_models
fix(proxy): show all model groups to proxy admins in /model_group/info
2026-09-16 17:06:44 -07:00
yassin
b35ca7d2c3 fix(spend_tracking): honour the global litellm_proxy override when inferring a model group provider
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-17 00:03:01 +00:00
Mateo Wang
6cdf398bea
Merge pull request #41504 from BerriAI/litellm_bedrock_agent_runtime_strip_virtual_key
fix(proxy): stop forwarding LiteLLM credential headers on Bedrock agent-runtime passthrough
2026-09-16 17:02:17 -07:00
yassin
d6b13f938d test(proxy): cover guardrail tag budget edge cases
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-16 23:52:39 +00:00
yassin
488666ccae fix(proxy): enforce tag budgets for tags added by guardrails
Auth runs the tag budget check before pre_call_hook, so a tag that a custom guardrail adds is attributed spend but never budget checked. After the pre-call hook, budget check only the newly added tags with the same exemptions auth applied (budget-free routes, zero-cost models), keep the pre-guardrail tag baseline across fallback retries, and surface an over-budget tag as the same budget_exceeded 429 auth returns

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-16 23:52:39 +00:00
yassin
5f1d87911a test(spend_tracking): cover unresolvable deployment leaving provider empty
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-16 23:50:22 +00:00
yucheng-berri
c5325b1492
Merge pull request #40596 from BerriAI/litellm_lit_7470_rate_limit_fallback_pristine_data
fix(proxy): retry rate-limit fallbacks from a pristine request snapshot
2026-09-16 16:47:35 -07:00
yucheng-berri
672f43fd54
Merge pull request #41356 from BerriAI/litellm_lit7836_call_id_endpoint_logs
fix(proxy): carry litellm_call_id through endpoint specific error logs and failure responses
2026-09-16 16:43:12 -07:00
mateo-berri
574ea15b8f fix(mcp): count admin static headers as api_key credential slots
An api_key server whose key lives in static_headers, the documented
shape for upstreams that expect a custom header name, dispatched fine
before the fail-closed check and was rejected as misconfigured after
it. The check now treats every static header the admin configured as a
credential slot for api_key mode, on both the MCP client path and the
OpenAPI tool path, with regression tests at all three layers.
2026-09-16 16:42:20 -07:00
mateo-berri
41737aeda8 test(proxy): drop the docstring from the agent-runtime passthrough regression class 2026-09-16 16:37:32 -07:00
yassin
9af014d75a fix(spend_tracking): keep inferred provider out of model reconstruction and OAuth provider lookup
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-16 23:36:54 +00:00
joshua-berri
41410e9556
Merge pull request #41364 from BerriAI/litellm_fix_mcp_auth_fail_closed_4501
fix(mcp): fail closed on missing upstream credentials
2026-09-16 23:26:38 +00:00
Moe Khalil
b7c6befb37 feat(router): discover token limits for hosted OpenAI-compatible models
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-16 23:19:54 +00:00
mateo-berri
3712d8de92 Merge remote-tracking branch 'origin/main' into litellm_model_group_info_proxy_admin_all_models
Some checks failed
LiteLLM Rust / rust-wheel (push) Has been cancelled
LiteLLM Rust / rust-lint (push) Has been cancelled
LiteLLM Rust / rust-test (push) Has been cancelled
Terraform Provider / gofmt, vet, build, test (push) Has been cancelled
Terraform Provider / Provider endpoints vs proxy OpenAPI schema (push) Has been cancelled
2026-09-16 16:18:45 -07:00
mateo-berri
2a9fa48730 fix: expand wildcard deployments for proxy admins on /model_group/info 2026-09-16 16:18:44 -07:00
ryan-crabbe-berri
81ae5caa7e fix(proxy): pass only a team admin's changed fields on to the team update 2026-09-16 16:11:52 -07:00
yassin
acd4f0eb04 fix(spend_tracking): attribute router-rejected requests to the model group provider
A request for a configured model group that the router rejects before picking a deployment (all deployments in cooldown, no healthy deployment) never gets a custom_llm_provider in its logging kwargs. The spend log payload persisted an empty provider, the daily spend tables carried it through, and the Admin UI Usage page rendered those requests under unknown even though every model in the group has a provider

get_logging_payload now takes the proxy router and, when the logged provider is missing, infers it from the model group's deployments. It only attributes when every deployment in the group resolves to the same provider; mixed groups, unknown groups and a missing router leave the value empty as before. Explicitly logged providers keep precedence

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-16 23:10:47 +00:00
ryan-crabbe-berri
be6e7ef173
Merge pull request #41345 from BerriAI/litellm_remove_duplicate_user_budget_hook
fix(proxy): remove duplicate user budget hook that 429'd zero-cost models
2026-09-16 16:06:41 -07:00
yucheng-berri
659c4ce5a1
Merge pull request #40669 from BerriAI/litellm_otel_passthrough_trace_propagation
fix(otel): propagate W3C trace context on HTTP and WebSocket passthrough
2026-09-16 16:02:46 -07:00
mateo-berri
02d9aae8c8 fix(proxy): stop forwarding LiteLLM credential headers on Bedrock agent-runtime passthrough
The agent-runtime branch of /bedrock/{endpoint} (agents, knowledgebases, flows,
retrieveAndGenerate, rerank, generateQuery, optimize-prompt) forwarded every
caller header to AWS next to the SigV4 signature, so a LiteLLM key presented in
x-api-key or x-litellm-api-key reached bedrock-agent-runtime verbatim. Build the
upstream header set explicitly: drop LiteLLM credential headers by name and any
authenticated secret by value, keep the remaining caller headers, and let the
signed headers win on collisions.
2026-09-16 16:01:52 -07:00
ryan-crabbe-berri
36eb9cdb35 test(proxy): drop the route-list membership test that the PATCH gate tests already cover 2026-09-16 15:56:27 -07:00
ryan-crabbe-berri
c37d0a2e66 fix(proxy): read general_settings without a cast and test the /team/update gate by behavior only 2026-09-16 15:51:04 -07:00
ryan
dae16264c1 test(auth): cover over-budget user on zero-cost vs paid model in common_checks
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-16 15:42:15 -07:00
ryan
287bbaa6c1 fix(proxy): remove duplicate user budget hook that 429'd zero-cost models
_PROXY_MaxBudgetLimiter re-checked spend:user:{id} against user_max_budget in
async_pre_call_hook without the zero-cost model exemption that
_user_max_budget_check applies in auth, so free models were rejected with
"Max budget limit reached." once a user was over budget. Auth already owns
this check, so the hook is deleted rather than taught the exemption again

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-16 15:42:15 -07:00
ryan-crabbe-berri
6e2ae19670 fix(proxy): enforce org budget ceilings on /team/update
update_team loaded the org without its budget row, so the org max_budget,
tpm_limit and rpm_limit checks silently passed. It now loads the budget the
same way /team/new does
2026-09-16 15:38:30 -07:00
ryan-crabbe-berri
a44a58e91a feat(proxy): let team admins edit tpm_limit when a proxy admin enables it
tpm_limit is the first field in the team-admin allow-list registry. The team
settings tab gives a team admin a form with only the enabled fields and sends
only those on save, and UI Settings labels the checkbox the same way
2026-09-16 15:38:30 -07:00
ryan-crabbe-berri
66519da9b6 fix(proxy): report the caller's team edit access on /team/info
The dashboard gated the team settings form on a role it guessed from the
is_* props, the members list and the org list. The org list is premium
gated and empty while loading, so a team admin who is also an org admin
was told team admins cannot edit, although /team/update accepts them as
an org admin

/team/info now returns caller_edit_access, resolved by the same helper
/team/update uses, and TeamInfo keys the form and the toast off that
field. The org list is only read for the organization dropdown now, and
general_settings is read through one validated accessor in both handlers
2026-09-16 15:38:30 -07:00
ryan-crabbe-berri
50f890d02d fix(proxy): keep the team-admin field allow-list importable on python 3.10
assert_never landed in typing in 3.11, so importing it from typing broke the
3.10 import smoke job and the py310 typing gate. Take it from typing_extensions
like the rest of the proxy does.

Also drop /team/update from test_neighbouring_team_routes_stay_closed. The route
is self-managed now, so the coarse gate admits the caller and update_team decides,
which the sibling test in the same file already asserts.

Claude-Session: https://claude.ai/code/session_018PUCupsaarVLJy4iDFx256
2026-09-16 15:38:06 -07:00
ryan-crabbe-berri
ac47c95db9 fix(proxy): re-read UI settings on every config reload
The persisted UI settings were read once at startup, so a proxy admin
flipping a runtime flag through PATCH /update/ui_settings only changed the
pod that served the request. Every other pod kept serving the old value
until it restarted.

add_deployment, the reload the scheduler runs every 30s, now re-reads the
row and applies the runtime flags before it takes the model reconcile lock,
so a change made through one pod reaches the rest within one reload
interval. The startup hook and the two settings endpoints share that helper
instead of each repeating the flag copy.

Claude-Session: https://claude.ai/code/session_018PUCupsaarVLJy4iDFx256
2026-09-16 15:38:06 -07:00
ryan-crabbe-berri
9b77b5c2cb feat(proxy): let proxy admins choose which team fields team admins may edit
Team admins could never reach POST /team/update: the route gate answered
401 before the handler's team-admin branch ran. This moves /team/update
into the self-managed routes, resolves proxy admin, org admin or team
admin inside the handler, and filters team admins through a new
proxy-wide UI setting, team_admin_editable_team_fields. The setting is
an allow-list of team fields. Empty means team admins cannot edit team
settings and get a 403 pointing at the proxy admin, and changing a field
outside the list fails 403 naming that field. Only values that differ
from what is stored count, since the dashboard resends the whole form.

The registry of fields the setting accepts ships empty on purpose. Each
field lands in its own follow-up PR with its value diff and dashboard
wiring. The Admin UI gains a "Team admin editable fields" section under
Settings > UI and a toast on the team page while editing is disabled.

Refs LIT-5722

Claude-Session: https://claude.ai/code/session_01A6SkwJdfZUmkzfUkrEkqX8
2026-09-16 15:38:06 -07:00
Yassin Kortam
a6e87526f8
Merge pull request #40843 from elifozdamar/litellm_fix_nonstream_parallel_slot_release
fix(proxy): release completed max-parallel slots promptly
2026-09-16 15:13:59 -07:00
mateo-berri
ea5887d4bb test: cover proxy_admin_viewer sessions in the model_group/info admin regression test 2026-09-16 15:10:48 -07:00
Yassin Kortam
f62f140f75
Merge pull request #41403 from BerriAI/litellm_key_total_spend
feat(proxy): expose lifetime total_spend on virtual keys
2026-09-16 14:49:02 -07:00
Yassin Kortam
2bfa10b6d9
Merge pull request #41311 from BerriAI/litellm_key_status_filter_deleted_lookup
feat(keys): filter /key/list by active, expired, revoked or deleted status and serve deleted keys from /key/info
2026-09-16 14:48:34 -07:00
Yassin Kortam
95abc9fb0b
Merge pull request #41330 from BerriAI/litellm_team_model_max_budget_v2
feat(team): team-level model_max_budget with key-level overrides
2026-09-16 14:48:29 -07:00
mateo-berri
da844e3dd2 Merge remote-tracking branch 'origin/main' into litellm_model_group_info_proxy_admin_all_models 2026-09-16 14:40:10 -07:00
ryan-crabbe-berri
4439631e05
Merge pull request #41379 from runjivu/fix/fallback-budget-check
fix!: re-check budget on router fallback targets
2026-09-16 14:39:56 -07:00