SpendLogsMetadata gains a typed azure_spillover key so a request Azure
served off pay-as-you-go capacity is visible in spend tracking, stamped
from the provider response headers or the processed llm_provider- headers
on the standard logging payload. The header parsing moves into a shared
azure_spillover() helper that is_spilled_over_ptu_request() now wraps.
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
The spec pinned Base UI's collision behaviour, not our code: it only passes
while the template popup happens to fit under the trigger at 1280x900, and
#41315's taller Add Auto Router form broke that premise for the second time
in three weeks. #41527 tried to scroll the trigger into the upper half, but
the dialog content is shorter than its max height, so nothing scrolls and CI
still fails 3/3 with the trigger at y=487
The guarantee #38554 introduced is that the popup never covers the trigger,
and the sibling spec keeps asserting that at a viewport with no room below
The chat-shaped output handler now takes the input translation as its
request scoping, so the logged request is scoped exactly once and with
the pre-call semantics of the surface it arrived on. This drops the
unscoped chat_shaped_request_conversation detour from af312dc8, which
made the Anthropic response scan remove in-sequence system turns under
skip_system while the request scan kept them
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Azure PTU deployments carry zeroed per-token pricing because the reservation
is billed flat by the hour. When Azure spills a request onto pay-as-you-go
capacity it returns x-ms-is-spilled-over: true, and that traffic was still
priced at zero. The response cost calculator now detects the spillover header
on the result's hidden params or the logged provider response headers and
skips the zeroed custom pricing only for genuine PTU deployments while the
feature flag is on. Azure sync streaming now also records response headers on
the logging object, matching the async paths.
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
A stream_options.include_usage usage chunk (empty delta plus usage) was folded into the final
transform round and rebuilt without its usage, so token counts and cost vanished from clients.
Metadata-only chunks are now replayed after the final text flush.
A terminal tool-call chunk arriving while earlier text was still held back carried
finish_reason=tool_calls ahead of that text. The finish_reason is now deferred to the final
text chunk whenever the choice has held text.
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
test_router_fallbacks_with_cooldowns_and_dynamic_credentials expected a
caller-supplied credential to register its own deployment and cool it down.
#41289 stopped registering it, so cooldown logic skips that id and the
assertion can never hold. The test now asserts what the router guarantees
today: a 429 to a forwarded credential cools down none of the shared
deployments, the next credential is still served, and a 429 owned by a shared
deployment still cools it down. The final live OpenAI call becomes a mock
The auto-router template spec assumed the Add Auto Router form left room
below the Template select at 1280x900. #41315 added classifier fields above
it, so the options opened upward. The spec now scrolls the trigger to the top
of the dialog and asserts it sits in the upper half before checking placement
/key/info now serves a deleted key from the archive with status deleted
instead of answering 404, so the delete test's convergence predicate never
settled and the read timed out against a 200 it kept discarding.
The predicate now waits for status deleted through the same
_key_info_everywhere helper the rest of the file uses, and KeyInfo carries
the status field. The chat-rejection assertion after it is unchanged, so
the test still proves the key stops serving.
The bedrock-realtime extra pinned aws-sdk-bedrock-runtime 0.7.x, whose Config and BedrockRuntimeClient surface is gone in 0.11. The handler now resolves AsyncBedrockRuntimeConfig, builds AsyncBedrockRuntimeClient with the awscrt duplex transport, closes the client when the session ends, and tells an absent SDK apart from an installed but unsupported version. Moves the pin to >=0.10.0,<0.12.0 with the awscrt extra
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
The Bedrock InvokeModel transformations decided whether to send the
tool-search-tool-2025-10-19 beta from hardcoded model name lists (a pattern
list on the messages path, an "opus-4" substring on the chat path), so Opus 4.8,
Opus 5 and Sonnet 5 never got the beta on the messages path, Opus 5 and Sonnet 5
never got it on the chat path, Opus 4.1 got it without support, and
/v1/model/info reported supports_tool_search as unset for all three.
Both paths now read the model map through one shared helper: the Bedrock
entries for Opus 4.8, Opus 5 and Sonnet 5 carry supports_tool_search
explicitly, and a claude-tool-search fallback rule flags Claude 4.5 and newer
for unmapped ids, inference-profile ARNs and mapped entries with no opinion,
so the next Claude gets the beta with no code change. An explicit false on a
resolved entry still wins.
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.
Forward streaming_transform_mode from guardrail litellm_params into PromptSecurityGuardrail so incremental_diff is reachable from config; the default stays block_only. In incremental_diff the guardrail now returns stream_holdback_chars alongside the rewritten texts so that a value split across streamed chunks (or across an abbreviation period) is never partially released before the vendor rewrite arrives. Each response text gets its own protect call so modified_text maps back to the right choice when n > 1, and custom_guardrail no longer logs a clean response as mask just because the guardrail attached holdback metadata
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
