main's test_one_text_per_row_over_a_system_prompt_is_rejected_by_name assumed the
top-level system prompt stays out of the scanned texts. This branch scans it, so one
text per structured row now lines up and the rewrite is applied; the test asserts that,
and a multi-block system prompt keeps the length-guard rejection covered.
The one-hour cache write fallback now takes the applied cache write rate, so an off-peak write price carries into it instead of the input rate
The cost estimate test for a cost-map model without cache prices now expects writes at the input rate, which is what the proxy bills
The recording logger in the deferred guardrail test types its callback parameters
A chat row whose content carried an empty text part counted two slots in the
chat completions handler while Prompt Security read one text out of the
modified row, so the structured rewrite was dropped and the request got the
named rejection. One shared helper now lists a row's slot texts and both the
slot count and the modified-row reader use it.
With a post-call guardrail the proxy defers async success logging, and every nested wrapper on a
/v1/messages call bridged to the Responses API overwrote the stored closure, so the spend log was
built from the outermost Anthropic-shaped reply under Responses semantics and recorded the prompt
tokens without the cache hit. The first wrapper to exit now keeps the slot, which is the innermost
provider response, the same one the non-deferred path logs.
The flat cost path also billed cache-creation tokens at 0 when the model had no
cache_creation_input_token_cost. It now falls back to the input rate, and the 1h rate to the
creation rate, matching the tiered path and the custom pricing helper.
The three guardrail translation handlers imported the exception from the proxy policy engine through a function-local import, which CodeQL flagged as a cyclic import. The exception and its helper now live next to the handlers in the shared guardrail translation utils, and the tests import it from there.
The Prompt Security modify-mode helper is also restructured into early-return TypedDict displays so the LIT002 budget stays at its limit
Usage tracking, compliance and the dashboard now treat both not_run (older spend logs) and skipped as unevaluated through a shared UNEVALUATED_GUARDRAIL_STATUSES set, so old records stop counting as passed. The skipped record is no longer written when the request carried images, since images without text were never dispatched to guardrails before this change and that gap is not a message-scoping skip
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
The per-guardrail status a scoped-out evaluation records is now skipped, matching the
skip_*_in_guardrail settings that cause it. Request-level rollup still maps it to not_run
so the StandardLoggingPayload status contract is unchanged
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Bedrock only runs a contextualGroundingPolicy when the ApplyGuardrail payload
carries grounding_source and query qualifiers. Callers sending ordinary system
and user messages never got those, so a configured grounding threshold was
silently skipped on /v1/chat/completions and /guardrails/apply_guardrail.
When no explicit grounding_source or query tags are present, system and
developer text is sent as grounding_source and the latest user message as
query. The apply_guardrail response branch now forwards the request messages,
which it previously dropped.
Resolves LIT-4224
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
The sandbox read user_api_key_* off the top level of request_data, where the proxy never writes them, and only forwarded request_data["metadata"], which is empty on /v1/messages, /v1/responses, batches and files because those routes keep proxy state in litellm_metadata. Merge both buckets (litellm_metadata wins) and resolve ids from the merged dict
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Prompt Security reports a queued sanitization job as status "created" before
it moves to "in progress" and "done". The poller treated anything other than
those two known strings as an error and returned HTTP 500 on the first poll, so
every image or file request through the guardrail failed while the vendor job was
still queued.
Only "done" is terminal now. Every other status is logged and polled again until
max_poll_attempts or the outer file_sanitization_timeout, after which the existing
fail-open or fail-closed (408) policy applies.
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
A not_run entry from the base guardrail only carries guardrail_name, while the
content filter's evaluated entry carries guardrail_id. Keyed apart, one request
listed twice in the monitor for a logging_only guardrail (Not run and Passed).
Resolve the id from a same-name sibling in the payload so the severity pick applies
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
A guardrail that rewrites text per chat message now returns the rewritten
rows as structured_messages instead of only texts, so the Responses and
chat handlers write the rewrite back through the structured path. The
generic guardrail API response accepts an optional structured_messages
list, Prompt Security modify builds one from modified_messages, and rows a
server echoes back exactly as shown are restored to the original row
objects because the request model drops undeclared keys. Texts-only
per-message answers keep the named rejection on both endpoints.
