Extract the Soniox SRT/VTT cue grouping and rendering into a shared
litellm_core_utils/audio_utils/subtitle_utils module, have Gemini
transcription request word timestamps whenever response_format is srt or
vtt, and let the http handler rewrite the response text into the
synthesized subtitle document (dropping the internally requested words
array) for any provider config that opts in via
supports_subtitle_synthesis
* feat(health): opt-in model-group allowlist for background health checks and health-check routing
* fix(health): merge shared health states per writer scope instead of replacing
* refactor(health): drop restating comment and parameterize test scope annotations
* chore: remove stray generated prisma migration file
* fix(health): merge health states against the Redis snapshot, not the pod-local copy
* fix(health): fall back to the pod-local snapshot when the Redis read returns nothing
Gemini Live sends no usageMetadata and no turnComplete for
gemini-3.5-transcribe-live sessions, so realtime spend logged as 0.0.
Attach estimated usage to the input_audio_transcription.completed event
using Google's published billing estimate (25 audio tokens/sec of input,
175 text tokens/min of output) derived from the streamed pcm16 audio
duration, gated to audio_transcription-mode models so conversational
Live models keep billing through usageMetadata. Also capture that usage
in the provider_config backend path so realtime cost calculation sees it.
* fix(ui_sso): resolve highest privilege Entra app role, not first in claim
A user assigned more than one Entra app role — commonly by belonging to
several assigned groups — arrives at the Microsoft SSO callback with every
role in the id_token `roles` claim. LiteLLM stores a single role per user,
and get_microsoft_callback_response collapsed the list by taking the first
value that resolved to a LitellmUserRoles and breaking.
Entra does not guarantee the ordering of the `roles` claim, so which role
won was effectively arbitrary: a user in one group mapped to internal_user
and another mapped to proxy_admin_viewer could be silently demoted to
internal_user, and proxy_admin could lose to either.
The generic/Okta path already resolves this correctly via
determine_role_from_groups, which walks a documented privilege hierarchy.
Hoist that hierarchy into LITELLM_USER_ROLE_HIERARCHY and reuse it, so
app-role logins and group-mapping logins agree.
Extract the selection into MicrosoftSSOHandler.get_user_role_from_app_roles
so it is directly testable — the existing tests re-implemented the loop
inline, which is why the ordering bug was not caught.
Behaviour is unchanged for single-role claims, unrecognised values, and
empty claims. Roles the hierarchy does not rank (org_admin, team, customer)
are resolved deterministically rather than by claim order.
* refactor(ui_sso): trim role selection prose and use immutable annotations
Addresses review feedback on the app role selection helper.
Drop the explanatory comments and the Args/Returns docstring boilerplate that
restated the control flow, keeping only the part a reader cannot infer from the
code: that Entra does not guarantee claim ordering, and how unranked roles
resolve.
Type the parameter as Sequence[str] rather than list[str] and build the resolved
set as a frozenset, so the helper stops adding an LIT001 mutable-collection
annotation. Make LITELLM_USER_ROLE_HIERARCHY a tuple for the same reason.
No behaviour change: the ordering regression tests still fail against the
previous first-match-wins logic and pass here.
Adds a Gemini audio transcription config that maps /v1/audio/transcriptions
onto the Interactions API (speaker attribution and word timestamps land on
the OpenAI verbose_json shape), registers both models with published pricing,
routes text-only Live sessions to TEXT responseModalities so
gemini-3.5-transcribe-live sessions survive, and makes the token-priced
transcription cost path provider-aware instead of hardcoding OpenAI.
Adds pricing (0.15/0.50 per 1M tokens, 0.03 cached read), the 1M context window, and capability flags (tools, parallel tools, tool choice, response schema, reasoning, vision) for Together AI's zai-org/GLM-5.3-Flash, mirrored into the backup cost map, with exact-value regression tests.
OpenAI's chat completions API rejects tool_reference content parts in
role tool messages, so a mixed text plus reference tool result carried
through the Anthropic adapter turned a previously working request into
a 400 on chat-routed OpenAI and Azure deployments. Strip the reference
parts there, keeping a reference-only result as an empty-text tool
message so the preceding tool_call stays answered, mirroring the
Responses bridge skip.
* feat(newrelic): per-team cost and usage metrics via team callbacks
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(newrelic): retry transient 429/408 metric posts instead of dropping
* fix(newrelic): drop only records queued when the drain began, not mid-drain arrivals
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
`test_gemini_chat_returns_content_and_logs_cost` asks gemini-2.5-flash to
"reply with the single word pong" under `max_tokens=32`, and has been seen
returning no content at all:
completion_tokens=29, reasoning_tokens=29, content=None
gemini-2.5-flash defaults to dynamic thinking, and `max_tokens` maps to
`maxOutputTokens`, which on the 2.5 family counts thinking tokens as well as
visible output. So the model is free to spend the entire budget on thoughts and
emit nothing, which is exactly what the usage above shows.
Raising the limit alone does not fix this. Dynamic thinking on 2.5 Flash is
documented up to 24576 tokens, so no budget small enough to be reasonable for a
one-word smoke test is safe. The fix is to take thinking out of the picture:
`reasoning_effort="none"` maps to `thinkingConfig.thinkingBudget=0` for the 2.5
family, so the whole limit is available to visible output. Verified against this
checkout:
get_optional_params(model="gemini-2.5-flash", custom_llm_provider="gemini",
max_tokens=32)
-> {'max_output_tokens': 32} # no thinkingConfig at all
get_optional_params(model="gemini-2.5-flash", custom_llm_provider="gemini",
max_tokens=64, reasoning_effort="none")
-> {'max_output_tokens': 64,
'thinkingConfig': {'thinkingBudget': 0, 'includeThoughts': False}}
This mirrors what the OpenAI tool tests in this same file already do with
gpt-5.6 for the same failure mode. `max_tokens` goes to 64 for headroom; with
thinking disabled that is ample for a one-word answer.
Neither `covers` claim changes: the call still exercises the gemini chat
translation path and still produces a costed SpendLogs row.
`_cacheable_system_block` embedded the per-run marker in all 300 paragraphs, so
the block's token count moved with the marker's own tokenization. Measured over
40 random markers the size ranged 3611-5408 tokens (median 4509): 15% of runs
landed under the 4096-token minimum cacheable prefix of Haiku 4.5, despite the
docstring claiming the prompt was comfortably above it.
When the system block is under the minimum, no cache entry is written at the
system breakpoint. The entry at the second breakpoint still gets written,
because system + first user turn clears the minimum -- which is why the failures
report a large cache_creation with cache_read stuck at 0
(`cache_creation_input_tokens=5610 cache_read_input_tokens=0`, and 5610 is the
whole prefix, not the user turn's share). `_prime_prompt_cache` rotates the user
turn on every attempt, so that second entry never prefix-matches the next
attempt either. Every attempt re-creates the full prefix, cache_read never rises
above 0, and the loop burns its 60s deadline:
prompt cache never became readable in full within 60.0s
That is the single most frequent flake in the e2e suite, 9 of 38 runs, and it
hits all three provider classes identically because they share this helper.
Move the marker out of the repeated paragraph so it appears once, and size the
block at 1500 paragraphs. The prefix is now 8056-8060 tokens across markers --
spread 4 tokens instead of 1797, and 1.97x the minimum in the worst case. The
same marker-per-repetition pattern in `_first_turn_user_text` is fixed the same
way. Both copies of the helpers stay byte-identical.
The test claiming mgmt.key.generate.happy_path signed in and then only read
/key/list, so nothing proved the session key an admin's sign-in mints is
actually accepted on /key/generate. It now does what an admin filling in
Create New Key does: POST /key/generate under the session key, read the new
key back from /key/info, see it in the dashboard's own /key/list, and drive
real traffic through it to confirm its model scope is enforced.
Adds ManagementClient.generate_key with the same caller_key seam update_key
and key_list already use, so the suite can call the route as the master key
or as a virtual key. Also wraps the over-long models import.
Refusing the dashboard session key on /key/generate turns only this test red;
the master-key generate, the key edit, and regenerate stay green.
Google withdrew gemini-live-2.5-flash-preview-native-audio-09-2025 from the
Vertex Live API. Every session dies at setup:
received 1007 (invalid frame payload data)
gemini-live-2.5-flash-preview-native-audio-09-2025 is not supported in the live api.
The client sees session.created (the proxy synthesizes it on connect) and then
nothing, so both vertex_ai realtime tests time out waiting for session.updated.
Confirmed by probing the Vertex Live endpoint directly with the e2e stack's own
credentials:
gemini-live-2.5-flash-preview-native-audio-09-2025 -> 1007, not supported
gemini-live-2.5-flash-native-audio -> setupComplete
so this swaps to the non-preview sibling, which is the same native-audio class
and is what the cost map already carries for vertex_ai.
Not a litellm regression. The suspicion fell on #38395 because it removed the
native-audio speechConfig strip, but the setup payload this suite sends is
byte-identical either side of that change: the strip only fires when a client
sends a voice, and the e2e SessionConfig has no voice field. Google's rejection
names the model, not a field.
The gemini (Google AI Studio) provider keeps the -09-2025 id, which still works
there; only the Vertex endpoint dropped it.
* feat(complexity_router): heuristic-first classifier chaining
Adds classifier_type 'heuristic_first', which scores locally on every request and
only calls the LLM classifier for traffic the scorer could not place at or below
heuristic_first_max_tier. A request short-circuits when the scorer landed at or
below the threshold and produced at least one signal; everything else escalates.
The signal requirement is load-bearing. A prompt where no dimension fires scores
exactly 0.0, which is under simple_medium, so the score-to-tier mapping calls it
SIMPLE by default rather than by evidence, and that is about half of general
traffic. Gating on the tier alone would route it to the cheapest model without
ever consulting the classifier.
Introduces uses_llm_classifier as the single owner of 'does this router call the
classifier model', replacing the classifier_type == 'llm' comparisons in the
config validator, the prompt prebuild, the health dependency graph, the
routing-test authorizer, and six dashboard sites.
* fix(complexity_router): reuse the heuristic verdict on classifier failure, load the threshold on edit
Three review findings, one push.
The heuristic-first fallback re-scored the prompt after a classifier failure,
which the README already documented as a reuse. The outcome computed before
escalation is now handed to the failure path, so the scorer runs once per request.
The edit modal never hydrated heuristic_first_max_tier, while save rebuilds every
managed key from form state, so opening a heuristic-first router and saving it
dropped a field the proxy requires. The dropdown's display fallback hid it. Both
are fixed, and the hydration is extracted into a pure function so a test can pin
the invariant: every managed key present in a stored config survives an untouched
open-and-save. That test also covers every field added later.
Classifier radio labels lost their em dashes, per the repo writing convention.
