* fix(bedrock): route unmapped openai family model ids to converse
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(bedrock): rename the e2e openai family backend constant
global.openai.gpt-6-sol has a cost-map row now, so the constant no longer
names an unmapped model
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(logging): pass provider response headers to callbacks on every endpoint
Custom callbacks only received kwargs["response_headers"] for chat
completions. Responses, image generation and edit, speech, and
transcription calls either never recorded the provider's headers or
recorded them in one place and not the other.
Every handler now records the provider's httpx headers on the response's
hidden params as "headers" (raw) and "additional_headers" (processed,
with LiteLLM's own entries winning on a clash), and the logging object
derives model_call_details["response_headers"] from those hidden params
before cost calculation on the non-stream and both streaming success
paths, keeping a handler-set value authoritative. Binary speech responses
expose their hidden params to the standard logging payload, and the sync
OpenAI transcription request always fetches the raw response.
* test(images): point the legacy image and speech fakes at the raw response surface
Image generation now goes through the SDK's raw response so the provider headers can be read, and the speech binary response now carries hidden params. The unit fakes in the image generation, xinference, proxy provider, image edit, Vertex speech, and otel suites still pinned the old call surface and the old "no hidden params" assertion, so they read an uncalled mock or a fake response without headers.
* test(images): drop the rewritten mock comments and the generated edit PNGs
* test(images): move the llm-span test's image fake to the raw response surface
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* feat(vertex): native batch JSONL passthrough with cost tracking
Add a per-request `passthrough=true` multipart field on `POST /v1/files`
(and the same kwarg on `litellm.create_file`) that uploads a native
Vertex AI batch JSONL to the deployment's GCS bucket unchanged, so rows
using `googleSearch` and other Gemini-only features run as written and
the output, `groundingMetadata` included, comes back untouched.
Passthrough is sticky through the GCS object path
(`litellm-vertex-files/passthrough/...`), so batch create and output
retrieval inherit it without new state. Native output rows are costed
from their `usageMetadata` with the deployment's model and model_info,
in the polling and retrieve paths and for the existing global
`disable_vertex_batch_output_transformation` flag, which billed $0
before.
The proxy requires the target to resolve to vertex_ai deployments only,
refuses `passthrough` with a non-batch purpose, a non-default
`target_storage`, or pre-call guardrails, and validates native rows on
`request` instead of the OpenAI batch keys.
* refactor(vertex): keep native batch row pricing inside the Vertex adapter
Moves native Vertex batch row detection, response parsing, and per-row
pricing from litellm/batches/batch_utils.py into
litellm/llms/vertex_ai/batches/transformation.py, so batch_utils only
aggregates the rows it gets back. Adds tests/test_litellm/files to the
misc unit shard so the new test directory is claimed by a shard.
* fix(files): say what a passthrough batch upload takes when a row is not native
The missing-key 400 listed bare key names, so an OpenAI-shaped row under
passthrough=true read "Each line must be a JSON object with keys request".
The batch line shape now carries its own hint, and the passthrough one says
a passthrough upload takes native Vertex batch rows with a request key
* fix(batches): bill native Vertex embedding batch rows on the native cost path
A native Vertex output row whose response holds an embedding was validated as a
generateContent response, so the documented tokenCount-only shape counted as a failed
row. Price embedding rows from their own usage (promptTokenCount, else tokenCount) with
the helper the transformed embeddings path already used, and drop the prompt-details
helper nothing calls anymore.
* fix(batches): keep modality batch rates on native Vertex embedding rows
An embedding row that carries usageMetadata was billed from promptTokenCount alone, so
its promptTokensDetails no longer reached the audio, image, and video batch rates the
way it did before the native cost path. Run every row with usageMetadata through the
Gemini usage parser and keep the flat tokenCount fallback for embedding rows without it.
* fix(batches): price native Vertex batch rows by modelVersion under a wildcard deployment
A `vertex_ai/*` deployment hands the batch cost path `*` as the deployment model, which
no cost map resolves, so every native (passthrough or flag-on) row was billed at $0. A
wildcard deployment model now defers to the row's own `modelVersion`, the way the
transformed path already prices by the row's `model`.
Also moves the native passthrough tests under tests/test_litellm, the tree codecov
reads, and covers the raw upload chunking, the embedding output translation, the
unpriceable-row path, and the flag-on dispatch.
* fix(batches): keep explicit deployment prices for native Vertex rows without a modelVersion
Under a wildcard deployment a native batch row that carries no modelVersion (an embedding
row, or a generateContent row Vertex returned without one) was billed at $0 even when the
deployment's model_info sets explicit batch prices, because the cost calculator was never
called. The row now falls back to the wildcard name, which the cost calculator prices from
the explicit model_info, and only a row with neither a modelVersion nor a deployment model
is billed at $0 with the warning
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(prometheus): add model_group label to deployment request and rate limit metrics
litellm_deployment_total_requests, litellm_deployment_success_responses,
litellm_deployment_failure_responses, litellm_deployment_tpm_limit and
litellm_deployment_rpm_limit had no way to identify which model_group a
pooled deployment belongs to, only requested_model, litellm_model_name and
model_id, none of which name the alias a model_name resolves through when
it fans out to more than one deployment.
model_group was already resolved onto enum_values for every request in
async_log_success_event, so this is a label-list addition for the metrics
built directly from that enum_values (the two request counters). The
failure counter builds its own UserAPIKeyLabelValues locally and had a
model_group variable already in scope that it never passed through, and
the tpm/rpm limit gauges are set from a helper that took no model_group
parameter at all even though its only caller already had it on
enum_values. Both now thread the value through.
* test(prometheus): expect model_group in deployment success/total request labels
test_set_llm_deployment_success_metrics asserts the exact label set passed
to litellm_deployment_success_responses.labels() and
litellm_deployment_total_requests.labels(), which now includes model_group
since it was added to those metrics' label list.
* fix(prometheus): bound model_group on deployment failure metrics
On a pre-routing reject (no deployment selected), model_group is
caller-supplied via litellm_params.metadata and was passed through
unbounded, letting an unrecognized value mint unlimited label series on
litellm_deployment_failure_responses / litellm_deployment_total_requests.
Bound it with the same _bounded_requested_model_label used for
requested_model on this path. When a deployment is actually selected,
model_group is router-resolved and passed through as-is.
Also documents the model_group parameter on
_set_deployment_tpm_rpm_limit_metrics and the bounding behavior on
set_llm_deployment_failure_metrics.
---------
Co-authored-by: ahamedshaik16 <24526479+ahamedshaik16@users.noreply.github.com>
* test(rust): encode anthropic response serialization shape as rstest cases
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(rust): shape Anthropic Messages requests natively
The Rust Messages route only relayed the body. It now runs the request shaping
the Python handler does for the direct Anthropic provider: history sanitizers
(empty blocks, tool ids, replayed web search results, provider_specific_fields,
encrypted reasoning, advisor blocks), reasoning_effort and adaptive/legacy
thinking translation against the model's capability flags, the sampling and
speed gates under drop_params, the metadata allowlist, additional_drop_params,
reasoning auto summary, OAuth and ANTHROPIC_AUTH_TOKEN credentials,
provider_specific_header merging and anthropic-beta injection. Capability flags
and LiteLLM settings reach Rust through route_host.shaping(). A request the
route rejects before the call now maps to BadRequestError instead of
APIConnectionError
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* test(rust): port Anthropic Messages shaping tests and pin comment contracts as cases
Every Python unit test that exercises the ported shaping for the direct
Anthropic provider now has a named rstest counterpart, and every comment that
stated a behavior contract is deleted in favor of a case that pins it. Measured
with cargo-mutants over the touched files, all viable mutants are caught
Porting the tests surfaced parity gaps, fixed here to match Python: every
casing of a forwarded anthropic-beta header is merged, replayed web search
results are rewritten from their own block (an empty result keeps its slot and
a server_tool_use with a non-string query stays), an empty output_config.effort
falls back to medium, speed and reasoning effort errors quote values the way
Python does, additional_drop_params apply after metadata validation and the
auto summary and never touch model or messages, and a non-string
metadata.user_id is rejected before the call
* fix(rust): resolve Messages credentials through the secret source and scope headers by resolved provider
The native Messages route read ANTHROPIC_API_KEY, ANTHROPIC_AUTH_TOKEN and the
base URL straight from the process environment, so a key or base held in a
configured secret manager was never found. Each provider config now declares
its secret names and the route resolves them through the same SecretSource the
OCR route uses, with the Python bridge passing in litellm's configured manager
provider_specific_header entries were scoped by the explicit
custom_llm_provider only, falling back to anthropic, so an azure_ai/ model
lost its azure_ai scoped headers. Scoping now happens in the route after the
provider is resolved from the model, as Python's handler does
The Azure config now adds the same anthropic-beta feature headers Python's
Azure route adds, and the metadata allowlist, reasoning auto summary and
history sanitizers move from the core route into the llms crate, mirroring
their home in Python's messages handler
* test(rust): escape the dot in the metadata.user_id match pattern
* refactor(rust-bridge): project Messages capabilities without mutable dicts
The capability flags and effort tiers were built as dict comprehensions,
which the type-discipline gate counts as mutable construction, and the
asdict call carried a mutable-ok suppression that suppressed nothing. The
flags are now passed one by one and the effort tiers are a frozen dataclass,
which asdict projects to the same map the native side reads
---------
Co-authored-by: Yujong Lee <yujong@berri.ai>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Non-streaming responses served by the Rust core already carry
x-litellm-rust: true through _hidden_params.additional_headers, which the
SDK exposes and the gateway renders as a response header. Native streams
did not, because the lifecycle Stream and SyncStream objects had nowhere
to hold hidden params and the marker writer skips objects without them.
Give both stream classes the same _hidden_params bag every other litellm
response has, so the existing marker attaches without wrapping the stream
or changing its identity.
Co-authored-by: Yujong Lee <yujong@berri.ai>
* refactor(types): replace Any with proven types in 13 files
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(types): drop unused executor import from utils type-checking block
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(realtime): keep reserved-key filtering on azure realtime health params
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(realtime): pin reserved-key filtering in azure realtime health auth params
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(realtime): exercise the real azure header builder in the reserved-key test
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): fail parked DB lookups at a deadline and flip readiness while they stall
Under a load burst with a slow authentication database every request parked inside
the pod with no deadline while /health/readiness kept answering 200 (its own ping
gets a fresh connection), so the load balancer kept sending traffic until the pod hit
its memory limit, and the parked requests completed against the provider minutes
after every client had hung up
Every pre-request read (key, team, user, end user, budget, membership, organization,
object permission, jwt mapping, project, proxy budget, spend counter reseed) now runs
under one deadline, PROXY_DB_LOOKUP_DEADLINE_SECONDS (default 10 s). A lookup that
hits it fails the request with the existing 503 "authentication database is
temporarily unreachable" answer, honours allow_requests_on_db_unavailable, and never
triggers the transport reconnect (the transport is fine, the query is slow), which is
what turned the repro's stall into "too many clients". Writes stay unbounded
A deadline hit marks the pod stalled for PROXY_DB_LOOKUP_STALL_WINDOW_SECONDS
(default 30 s, 0 disables), during which /health/readiness answers 503 with
"db": "stalled" behind the same fail-open gate, so the pod leaves rotation before it
fills its memory. The existing litellm_in_flight_requests gauge already exposes the
parked set on /metrics
The deadline is enforced on the wall clock: bounded_db_lookup waits on the lookup
task with asyncio.wait and raises DBLookupDeadlineExceeded when the deadline passes
even if the lookup absorbs its cancellation, where asyncio.wait_for on 3.12+ would
sit on the cancelled task for as long as it takes
The failure spend-log row no longer re-runs the key and team lookups when the
failure itself is a database connection or deadline error, so a request that hit
the deadline is answered after one deadline instead of two
* fix(proxy): bound the spend counter gate wait and narrow the stalled lookup shortcut
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): keep the global spend lookup on the prisma client handle
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
Co-authored-by: yassin <yassin@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): list key and team model aliases in GET /v1/models
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): keep alias listing helpers within the type discipline budget
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(proxy): cover alias rows on GET /v1/models and /v1/models/{id}
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): apply team then key aliases like chat completions and keep the alias as the retrieved id
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): apply key aliases twice like chat completions and skip only malformed alias entries
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): apply the global model_alias_map between the key alias passes like chat completions
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): list only the caller's own aliases and never rewrite a listed model id on retrieval
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* style(proxy): ruff format model_info alias lookup
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): hide undiscoverable names from model retrieval so an alias named like one resolves to its target
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): keep undiscoverable models retrievable by id while excluding them from the alias guard
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(proxy): pass an immutable name sequence into the model_info alias guard
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(proxy): type the model list alias test helpers
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(proxy): annotate the new alias listing test fixtures and helpers
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(s3_v2): bound concurrent S3 uploads per flush and add opt-in JSONL batch files
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(s3_v2): keep failed uploads queued, parse env-backed flags, add integration coverage
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(s3_v2): type test helpers and honor constructor bound when config value is null
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(s3): annotate required casts for the type-discipline gate
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(s3_v2): keep tenant prefixes, stable retries and cold storage safety in batch file mode
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(s3_v2): cover root-level batch file keys for codecov patch target
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(s3_v2): audit matrix across chat, messages and responses surfaces with sink faults
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(s3_v2): read sink objects under the lock in the audit cells
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(s3_v2): rebind the retry queue instead of slicing in place
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(s3_v2): ignore stray non-POST requests in the surface upstream
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(s3_v2): keep fake upload state on the fake client instead of nonlocal counters
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: yucheng <yucheng@berri.ai>
* test(integration): reproduce passthrough upstream error body missing from logs and spend row
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(passthrough): log upstream 4xx/5xx error bodies and carry them into the failure hook
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(error_normalization): let the passthrough prefix win over upstream body text
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(passthrough): honor message redaction for upstream error bodies
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(passthrough): bound the upstream error body read and sanitize it before logging
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(passthrough): use the Sequence import directly in the allowed-routes cast
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(passthrough): rechunk the upstream error stream so the preview read stays bounded
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(integration): audit matrix for passthrough upstream error visibility
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(passthrough): drop the restating docstring on the upstream failure logger
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(integration): drop the retired covers markers from the passthrough error tests
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(passthrough): keep the upstream status when the error body peek fails
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(passthrough): cover the relay aclose in the mid-read failure test
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(passthrough): relay decoded partial body on mid-read failure
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: yucheng <yucheng@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): do not requeue a daily spend batch whose commit already left for postgres
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): settle an interrupted daily spend commit from the shutdown flush instead of blocking the cancelled tick
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(integration): burst two workers and SIGTERM during daily spend COMMIT, expect exactly once
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: yucheng <yucheng@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): keep deployment labels on cache-hit post_call rejections
A post-call failure on a response served from the litellm cache set no first_api_call_start_time, so the failure hook flagged it as rejected before routing and dropped the model_id and provider labels
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): read the cache hit from caching_details in the failure hook
model_call_details[cache_hit] is stamped inside the enqueued success handler, so a post-call failure can observe it too early; logging_obj.caching_details is set synchronously before the cached response returns
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(integration): cover cache-hit guardrail reject deployment labels across endpoints, modes and chaos
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(integration): bound the worker-kill reject count by in-flight losses
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(proxy): assert provider and model labels on the cache-hit regression test
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: yucheng <yucheng@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(logging): bound the exceeded budget regex so a crafted error message cannot stall the proxy
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(integration): audit normalized_error clustering on long messages with a real two worker proxy
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(integration): poll for the budget denial and correlate upstream 503 bursts by request marker
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(logging): replace the bounded exceeded budget regex with a linear scan that keeps the original semantics
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(integration): compare upstream error wording against the decoded message
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(integration): tolerate a reaped worker while listing proxy children
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: yucheng <yucheng@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(presidio): stream non-Anthropic raw SSE through the post_call hook unbuffered
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(presidio): keep the pytest.raises block to a single await
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(presidio): move raw SSE format check into a helper to keep hook complexity flat
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(presidio): fold the raw SSE format check into the existing bytes branch to stay within the complexity budget
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(presidio): decide raw SSE stream shape on a complete first frame, not a transport fragment
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(integration): cover presidio post_call streaming for native gemini passthrough and anthropic messages
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(presidio): cap first SSE frame coalescing at 64 KiB so an unterminated first event cannot buffer unbounded
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(presidio): name raw SSE passthrough in the skipped output masking warning
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: yucheng <yucheng@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(model_prices): registry audit 2026-09-23, in-region Bedrock Claude and OpenRouter price fixes
Absorbs #42698
Co-authored-by: coldStoneSoul <shekharpachlore@gmail.com>
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(model_prices): keep registry formatting unchanged
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test: use eu.amazon.nova-pro for regional pricing probe after in-region parity
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test: lock in-region parity for bare Bedrock Claude ids
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test: compare every pricing field for bare Bedrock Claude parity
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(model-catalog): add above_32k cost fields to ModelInfo round-trip
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* Revert "fix(model-catalog): add above_32k cost fields to ModelInfo round-trip"
This reverts commit c71d5a3de3.
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: coldStoneSoul <shekharpachlore@gmail.com>
* fix(proxy): document responses API request and response schemas in openapi
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): namespace colliding openapi defs instead of overwriting existing components
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* chore(proxy): regenerate lazy openapi snapshot and dashboard schema types
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): require model and input in responses schema, document event stream, fix def collision refs
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): mark responses request fields readonly required
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): reuse existing OpenAPI components when a $defs entry has the same shape
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: kerry <kerry@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(ollama): read the JSON thinking field on non-streaming completions
Ollama's /api/generate returns reasoning in a top-level `thinking` field, but
the completion transport only looked for inline <think> tags. reasoning_content
was therefore always null, and a model that spent its whole turn reasoning
returned an empty assistant message with tokens billed.
Port the precedence the ollama_chat transport already uses: the field wins and
inline tags stay the fallback. Applied to both non-streaming paths, including
the JSON-mode text fallback. The two fields are read through a small validated
model rather than off the untyped JSON, so absent and explicitly null
`response` stay distinct exactly as before.
* fix(ollama): keep the thinking field on JSON-mode completions
The first pass read `thinking` for plain replies and for JSON-mode text that
failed to parse, but the three JSON-mode branches that succeed still dropped
it: an empty `response`, a valid JSON object, and a function-call shaped one.
A model that spent its whole turn reasoning under `format: json` therefore
still came back blank with the tokens billed.
Carry the field on all three, type the new test helper's parameters, and cover
the null and malformed `response` fallbacks.
---------
Co-authored-by: Pawan-Shahane <shahanepawan511@gmail.com>
* fix(models): sync openrouter prices from the models API
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(models): allow above_32k_tokens cost fields in price map schema test
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: kerry <kerry@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): gate disable_global_guardrails on keys and teams to proxy admins
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test: cover metadata smuggle with explicit false and UI toggle gating
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(proxy): satisfy PT017 in resend-stored guardrail flag test
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(proxy): keep regenerate_key_fn under the C901 ceiling via a guardrail opt-out helper
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* chore(ui): regenerate schema.d.ts for guardrail opt-out docstrings
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): gate disable_global_guardrails on caller-sent metadata, not server defaults
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(integration): audit cells for disable_global_guardrails admin gate
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(integration): restore contracts.json formatting
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(integration): share guardrail opt-out helpers
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(ui): hide the team disable_global_guardrails switch from non proxy admins
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(integration): drop covers markers and bound the slow sink check to the sink delay
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: yucheng <yucheng@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(proxy): honor model_info.discoverable on the model listing endpoints
A model_list entry marked model_info: {discoverable: false} is left out of GET /v1/models (OpenAI and Anthropic shapes, scope=expand and wildcard routes included), the list path of GET /v1/model/info and GET /model_group/info for every caller without the admin view, while direct requests naming the model keep routing to it. The field defaults to None so an absent flag reads as discoverable and nothing is persisted or echoed for configs that never set it.
* fix(proxy): hide flagged team models under their public name and cover the scope=expand filter
The discoverability lookup now resolves a listed name with the caller's team context, so a team-scoped deployment marked discoverable: false drops out for that team's keys under its public name instead of failing open. The scope=expand branch is now exercised by a team admin caller, and the OCI secrets test builds a real UserAPIKeyAuth instead of a spec mock that has no pydantic fields.
* perf(proxy): resolve only candidate names in the discoverable filter
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(mcp): reject duplicate MCP server names and aliases
MCP server_name and alias were unchecked at write time, so two servers
could share one tool prefix and tool routing resolved to an arbitrary
winner. Writes now run inside an advisory-locked transaction that
rejects a collision on either column case-insensitively with a 400
naming the colliding identifier, covering create, edit, connector
import and restricted-admin submission. Server reload logs one warning
per identifier already shared in the database.
Co-Authored-By: bot_apk <apk@cognition.ai>
* fix(ui): block duplicate MCP server names and aliases before submit
The create and edit forms now check the normalized name/alias against
the loaded server list (case-insensitive, spaces to underscores, own
row excluded on edit) and show a field error instead of submitting.
Structured proxy error bodies are unwrapped so a 400 no longer renders
as 'Error: [object Object]'.
Co-Authored-By: bot_apk <apk@cognition.ai>
* fix(mcp): check identifier conflicts when an alias is cleared
Clearing an alias drops the tool prefix to the stored server_name, so
that name must go through the conflict check too; an explicit alias:null
is now treated as an identifier write. Also narrows the new db tests to
behavioral assertions instead of pinning prisma where shapes.
Co-Authored-By: bot_apk <apk@cognition.ai>
* fix(mcp): treat an empty alias as a clear in conflict checks
An empty-string alias was written unchecked even though the prefix falls
back to server_name; the update path now treats any falsy alias like a
clear. The edit form likewise compares a cleared alias as empty instead
of re-checking the alias being removed.
Co-Authored-By: bot_apk <apk@cognition.ai>
* test(mcp): cover clearing an alias to an empty string
Co-Authored-By: bot_apk <apk@cognition.ai>
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: bot_apk <apk@cognition.ai>
* fix(otel): root post-response service spans in their own trace linked to the request
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(otel): trim service span context docstring
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: yassin <yassin@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(redis): authenticate sync clusters with IAM credential providers
Signed-off-by: Silu Panda <31051721+SiluPanda@users.noreply.github.com>
* test(redis): exercise IAM cluster authentication over TCP
Run Azure and GCP regressions against a real local cluster with only
cloud token issuance stubbed. Build a checksum-verified Redis server
in the compatibility workflow and report its coverage.
Signed-off-by: Silu Panda <31051721+SiluPanda@users.noreply.github.com>
* test(redis): separate unit and cluster integration coverage
Keep the mapped test tree mock-only. Run the live cluster cases
from the existing local caching integration file, selected by
explicit node IDs in the Redis compatibility workflow.
Signed-off-by: Silu Panda <31051721+SiluPanda@users.noreply.github.com>
* test(ci): isolate workflow coverage audit fixtures
Replace the stale unrun caching-file assumption with isolated workflow
fixtures for file and node-ID selectors. Keep the unnamed-file negative
check and clarify which live caching cases remain outside CI.
Signed-off-by: Silu Panda <31051721+SiluPanda@users.noreply.github.com>
---------
Signed-off-by: Silu Panda <31051721+SiluPanda@users.noreply.github.com>
Co-authored-by: yassin <yassin@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(bedrock): serve the OpenAI models on bedrock-runtime's native Responses API
AWS serves the OpenAI models on bedrock-runtime through an OpenAI-compatible
surface at /openai/v1/responses, alongside Converse. LiteLLM had no Responses
config for the bedrock provider, so /v1/responses fell back to the Chat
Completions bridge and was translated into Converse. A realistic Codex session
does not survive that translation: its function_call / function_call_output
history becomes Converse toolUse / toolResult blocks with no toolConfig, and
Converse rejects the request outright.
Add a Responses config for that surface, opted into per model from the price-map
supported_endpoints so models without the signal keep the bridge exactly as
before. Auth is Bearer when a Bedrock API key is present, SigV4 otherwise.
Both Bedrock endpoints reject the Codex history item types agent_message,
context_compaction and local_shell_call, so the normalization bedrock_mantle
carried privately moves into a shared module and both providers use it. They are
history items, so they only bite from the second turn onward -- a first-turn
smoke test passes and hides the problem. Verified against bedrock-runtime with
global.openai.gpt-5.6-sol: additional_tools is accepted there (unlike on
bedrock-mantle) while those three types are rejected, so the two endpoints do
not share one validator and each provider opts in explicitly.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(bedrock): build the Responses endpoint from the region's partition suffix
get_complete_url hardcoded amazonaws.com in an f-string, so every non-commercial
partition got the wrong host: cn-north-1 resolved to amazonaws.com instead of
amazonaws.com.cn, and GovCloud/ISO regions were wrong the same way. Defer to
BaseAWSLLM._select_default_endpoint_url, which this config already inherits and
which resolves the suffix per partition.
test_no_fstring_hardcodes_the_commercial_dns_suffix scans the whole tree, so it
caught this even though it is not one of this PR's test files. Register the
config in ENDPOINT_BUILDERS so the cn/GovCloud endpoint sweep covers this
surface from now on rather than only the f-string guard.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* feat(bedrock): opt the gpt-6 family into the native Responses API
* fix(bedrock): drop the Responses tool types bedrock-runtime rejects
Codex sends a web_search tool on every turn. api.openai.com runs that tool
itself, and the Converse bridge dropped it silently, but bedrock-runtime's
native Responses endpoint rejects the whole request with 400 "web search is
not supported for this request". Filter the request's tools down to the
types bedrock-runtime's own validation error names, logging what was dropped,
through a helper shared with the Mantle route, which already did the same.
* fix(bedrock): emulate file_search and collapse custom Responses paths
* fix(bedrock): keep background and remote image inputs working on the native Responses route
* fix(bedrock): inline remote images inside tool outputs on the native Responses route
* fix(bedrock): inline remote computer screenshots on the native Responses route
---------
Co-authored-by: Leonardo Freitas dos Santos <leonardo.freitas.s@outlook.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* feat(rust-bridge): declare native chat completions and responses bindings
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(dispatch): cover chat completions and messages dispatch
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(rust-bridge): keep secret manager stub formatting unchanged
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(rust-bridge): match stub parameter names and exports to the native surface
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(rust-bridge): name declining entrypoint parameters and export embeddings in the stub
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(rust-bridge): cover embeddings bindings in the route matrix
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(rust-bridge): keep secret manager stub formatting unchanged
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: Yujong Lee <yujong@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(mcp): return 401 challenge for REST token-exchange tool calls without a subject token
Co-Authored-By: bot_apk <apk@cognition.ai>
* fix(mcp): keep tool_server_mismatch when server_id disagrees with the tool prefix
Co-Authored-By: bot_apk <apk@cognition.ai>
* test(mcp): type the token-exchange challenge test helpers
Co-Authored-By: bot_apk <apk@cognition.ai>
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: bot_apk <apk@cognition.ai>
The tool_result user message built by the /v1/messages MCP loop used tuple
content, which the Messages to Chat Completions adapter silently dropped, so
non-Anthropic models re-requested the tool until the iteration cap or the
provider rejected the follow-up. Emit list content so the existing tool_result
branch translates it into a role tool message keyed by tool_call_id.
Resolves LIT-8474
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: bot_apk <apk@cognition.ai>
* feat(guardrails): speak the Straiker v3 platform API (/api/v3/detect)
The Straiker guardrail posted a webhook envelope to /api/v1/detect/webhook.
The v3 platform exposes /api/v3/detect instead, and its integration keys
(sk_agt_…) are rejected by the v1 route with an empty 401, so a tenant on
the v3 platform could not run this guardrail at all. Measured on a
customer gateway on 2026-09-17 after they rotated to a v3 key.
v3 parses the gateway's own traffic server-side, the same contract as
Straiker's unified Kong plugin. So on v3 the guardrail relays: the
request phase posts the provider body LiteLLM received (Anthropic
Messages or OpenAI chat), the response phase posts
{straiker_phase, sse, model, request}, the answer beside the request it
answers, and Straiker derives prompt, answer, agent and archetype. Both
phases also carry the flat prompt / app_response pair: a gateway-mode
integration key scores only the flat pair and an api-mode key only the
relayed body, each ignoring the other, so one payload serves whichever
key the console issued and it is one turn either way (measured on tenant
123, both key modes, 2026-09-18).
- api_version: "v1" | "v3", unset follows the key prefix, so a v3 key
needs no extra configuration. Explicit override still wins.
- The relayed body is an allowlist of provider fields. The hook sees the
client body merged with proxy state: `deployment` carries the resolved
provider credential and `proxy_server_request` the client's own
Authorization header. Neither travels. Identity survives as the
metadata subset Straiker's LiteLLM adapter reads.
- Identity never sends a proxy placeholder. `default_user_id` and the
master-key alias were being forwarded as a user and became the
session's identity on the platform.
- Headers: x-tool: litellm (ingress), x-straiker-phase, x-straiker-user,
and x-claude-code-session-id forwarded when the client sent it.
- Verdict: hookSpecificOutput.permissionDecision on the gateway envelope,
`action` on the flat one; block on block/deny, and on a non-empty
blocked_by as a backstop. A detect-mode control reads NONE.
- An error status from Straiker is now a webhook failure. LiteLLM's HTTP
client raises on any non-2xx and the retry loop caught only connection
errors, so a 401 or 503 from Straiker escaped the guardrail as an
exception and was relayed raw to the client, bypassing fail_open /
fail_closed. Retryable statuses retry; the rest are final.
- v1 is unchanged: same envelope, same X-Straiker-Webhook-Format header.
Tests: 15 new, fixtures from the request dict a hook sees on 1.98.0 and
the verdict envelopes the v3 platform returned on 2026-09-18. Each fix
was mutation-checked (handling removed, the test fails). Live: the same
eight-case battery (chat, /v1/messages, streaming, tool call; benign,
injection, PII) passes on a gateway-mode and an api-mode key, blocks at
pre_call with the tenant's block message, and lands under the declared
agent with the end user attributed.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* feat(guardrails): name the agent per application on v3 (x-s6r-agent)
One integration key can front several applications. Straiker enumerates them
as separate agents when the turn names one, which is what the unified Kong
plugin sends as x-s6r-agent. Without it every application on a gateway
collapses onto a single agent.
- Forwards a client-supplied x-s6r-agent.
- New `agent_ref` config names one agent for a route when the client sends
nothing. The client wins, matching Kong's precedence.
- Neither set: no header, and the platform derives the agent from the traffic.
Verified live on tenant 123 against an integration whose connector is
`gateway`: three distinct values minted three observed agents, and a turn
with no hint derived one from the traffic shape. An integration whose
connector is `custom-agent` declares its agent, so every turn attributes to
that one agent and the hint is ignored (agent_ref_source: attested).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* docs(guardrails): which v3 shape is scored depends on the connector, not the key mode
The earlier comment said a gateway-mode key scores only the flat pair. Re-measured
on tenant 123 across all three integration types with one injection prompt:
custom-agent connector (Add Agent) raw body ignored flat prompt scored
gateway connector raw body scored flat prompt scored
api mode raw body scored flat prompt ignored
Behaviour unchanged: the payload already carries both shapes, which is why it works
on every type. Comment only.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(guardrails): send exactly what the unified Kong plugin sends on v3
The v3 platform parses the gateway's traffic itself and derives agent,
archetype and identity from it. The earlier commits added to the relayed
body (a flat prompt / app_response pair, source, user_name) and to the
headers (x-tool, x-straiker-phase, x-straiker-user). None of that is in
the Kong v0.12 contract, and traffic through this guardrail was not
classifying by shape the way the same traffic through Kong does. Match
Kong byte for byte and leave classification to the platform.
Request phase: the provider body, plus session_id and
original.processed.Meta.user. Response phase: {straiker_phase, sse,
model, request} plus the same two. No flat fields, no phase or user
headers, no x-tool.
Session id follows Kong's precedence: the client's x-claude-code-session-id,
then the session LiteLLM resolved, then an md5 of system prompt + first
message so a conversation that states no session still groups across its
replays.
Routing hints complete the Kong set: x-s6r-agent (client header, else
`agent_ref`), and new `client` (x-s6r-client) and `format_hint`
(x-s6r-format) config, both optional.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* test(guardrails): sort imports in the v3 session test
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* fix(straiker): send a streamed Messages answer back in the Messages shape on v3
On a streamed /v1/messages call the proxy rebuilds the answer as a chat
completion before the post-call hook runs, and that is what the plugin put in
the response envelope's sse field. Straiker's coding-agent reader parses a
Messages answer, so a Claude Code turn relayed this way came back
coding_agent/claude with no session and zero events scored: the model's tool
calls were never screened on the response phase. Captured live on 2026-09-18
against tenant 123, a real Claude Code Bash tool call through the proxy.
The proxy's own Anthropic adapter turns the rebuilt answer back into a Messages
response when the call arrived on the anthropic_messages route, which is what a
transport relay forwards. Chat completions calls keep the chat completion shape
and a buffered Messages answer is relayed untouched.
The regression test's fixture is the chat completion the proxy actually built
for that captured turn. After the fix the same turn scores on the response
phase (session resolved, one event, the Bash tool_use block present).
* style(straiker): ruff format the v3 guardrail and its tests
* refactor(straiker): one attempt per call in the webhook retry loop
The HTTPStatusError branch added for v3 duplicated the non-200 branch and put
_post_webhook over the strict complexity ceiling. One attempt is now its own
method that returns the verdict or a failure marked retryable, and the loop only
decides whether to try again. Behaviour is unchanged: retryable statuses and
transport errors retry, everything else is final.
* fix(straiker): name Claude Code's client and agent on v3 so its session lands under one coding agent
Straiker types a gateway turn as a coding agent from the "You are Claude Code"
preamble, which only the main agent turns carry. Claude Code's title and
topic-detection sidecars have their own system prompts, so they resolved by
shape as autonomous, and because they share the session id with the main turns
the whole session was filed under Autonomous rather than under a coding agent.
Kong does not hit this because its plugin config names the client and agent on
every call.
The User-Agent (claude-cli/...) is on every call including the sidecars, so the
plugin now reads it and sends x-s6r-client: claude plus, when the route names no
agent, x-s6r-agent: "Claude (LiteLLM)". A client-supplied x-s6r-agent or the
agent_ref config still wins. Verified live on tenant 123: a real Claude Code
session now lands as one coding_agent labelled "Claude (LiteLLM)" with its turns
scored, where before it split across Autonomous.
Identity: the key's own user (email then id) now outranks the end user the
request named. LiteLLM resolves Claude Code's hashed metadata.user_id as the end
user when nothing better is set, so a per-user key was being shadowed by a
session token. The key is the authenticated principal, the way a Kong consumer
is, so it wins; the request end user is the fallback.
* refactor(straiker): build the v3 request, envelope and headers as frozen mappings
The v3 builders seeded dicts and grew them, which the type-discipline gate
counts as mutable accumulators. Each is now one expression over a tuple of
pairs, frozen with MappingProxyType, and the JSON encoder unwraps a frozen
mapping through a default. The session seed and the verdict parser no longer
rebind locals. The wire is unchanged: 36 live calls through the proxy on this
commit carry the same fields, shapes, headers and identities as before, with
no mappingproxy text in any body.
* fix(straiker): satisfy basedpyright on the v3 builders
The frozen-mapping refactor left a shadowed headers local, a Mapping handed to
an HTTP client that takes a dict, an unguarded optional response, a turn id
typed object, and a redundant isinstance on already-typed texts. No behaviour
change: 4 live calls (chat, Messages, Bedrock, injection) return 200 with the
expected verdicts on this commit.
* fix(straiker): type the v3 config fields at the initializer and keep the verbose log as JSON
The four v3 routing fields (api_version, agent_ref, client, format_hint)
travelled through the untyped kwargs passthrough, which basedpyright counts
against the budget. They are now validated through a small Pydantic model at
the initializer and passed by name.
The verbose log serialized the frozen payload with default=str, which printed
a Python repr instead of JSON once the builders returned MappingProxyType.
Every serializer now unwraps a frozen mapping first. A test asserts the logged
payload parses as JSON and carries the identity; mutating the log site back to
default=str fails it.
* fix(straiker): address review findings on the v3 relay
Text completions relay their prompt: `prompt`, `suffix`, `echo` and `best_of`
join the provider allowlist, so /v1/completions traffic is screened.
The route's `agent_ref` now outranks the caller's `x-s6r-agent` header. The
header is caller-supplied, and letting it beat a pinned route would let any key
file its traffic under another application's agent and controls. On a route
that names nothing the header still names the application, which is how
several applications enumerate behind one key.
Credentials inside `tools` and `mcp_servers` (an OpenAI `mcp` tool's `headers`,
Anthropic's `authorization_token`) are replaced with `[redacted]` before the
body leaves the proxy, on both phases and in the verbose log. Detection reads
tool names, descriptions and schemas, never these.
A 200 whose body is valid JSON but not an object now reports an invalid
schema and follows the failure policy instead of raising out of the hook.
Comments that restated a constant are gone. Tests cover each change and the
failure paths (unreadable error body, client exceptions, missing response,
unmodellable request, session seeds from Anthropic block shapes); every fix
fails its test when reverted.
* fix(straiker): scrub tool credentials one level deep, without recursion
* fix(straiker): scrub only the fields that carry a credential, never a schema
The credential set is now the three fields that actually hold one on a tools
or mcp_servers entry (headers, authorization, authorization_token), read one
level deep. A function tool whose parameter schema defines a token, headers or
api_key property is relayed exactly as sent; a test pins that, and fails
against the recursive version.
* test(straiker): use example.com identities; drop a comment that restated its branch
* fix(straiker): present a legacy completion as the chat exchange it is
Straiker scores chat on both phases of a gateway turn but has no reader for a
text_completion answer: the request phase of a /v1/completions call was
scored and the response phase was refused with 501, whether or not the call
named an agent. A completion is one user turn and one assistant turn, so both
phases now present that exchange: the prompt becomes the single user message
and the TextCompletionResponse becomes a chat completion. Measured through the
proxy on this commit, both phases return 200 and score, and the derived
session is shared between them.
The derived session seed accepts the tuple the conversion produces; the test
pins the session on both phases and fails against the list-only check. The
unreachable "parsed is None" branch is folded into the failure branch, and a
malformed tools value is shown to relay as sent.
* fix(straiker): screen a completions prompt as the text the model receives
LiteLLM's /v1/completions accepts a string, a list of strings, a list of
token ids or a list of token-id lists, and decodes token ids with the
text-davinci-003 tokenizer before calling the model. The relay now renders
the prompt the same way, one user message per prompt, so a pre-tokenized
prompt is screened as the text it stands for rather than as digit strings.
A prompt in a shape this cannot render (empty, mixed, or with no tokenizer
available) is relayed untouched instead of being replaced with something
else. Tests cover all four accepted shapes and six unrenderable ones.
* fix(straiker): seed the derived session on the preamble and the first user turn
An OpenAI chat body carries its system prompt as messages[0], and the derived
session seeded on the Anthropic `system` field plus messages[0] with no role
check. For that shape the seed was the system prompt twice and the first user
turn never counted, so every unnamed conversation behind one system prompt
collapsed into one Straiker session. The seed now takes the preamble from
wherever the API puts it (`system`, `instructions`, or a leading system or
developer message) and the first message with role `user`, else a Responses
`input` string, else `prompt`. Two conversations sharing a system prompt are
two sessions again; a replayed conversation stays one.
* fix(straiker): seed the derived session on every text block of the first turn
A user turn that opens with an image or a document block and carries its
text later seeded the session on an empty string, so two different
conversations under the same preamble shared one Straiker session. Read
every text block of the turn instead of only the first block. A plain
string or a single text block seeds exactly as before.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* test(straiker): cover the tokenizer fallback, a textless first turn and Responses instructions
Three branches of the v3 relay had no test: a token-id prompt relayed as
sent when the tokenizer cannot be fetched, a first user turn with no text
seeding the session on the preamble alone, and a Responses API body
seeding on its instructions and first input turn. Each test fails when
its branch is mutated.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* fix(straiker): seed the derived session on the principal as well as the conversation
Straiker skips turns it has already scored for a session. The derived
session hashed the system prompt and the first user turn alone, so two
users who opened a conversation with the same words shared one session,
and the second user's copy of an attack came back as a replay: unscored
and allowed. Measured live on 2026-09-20: the first user's SSN turn was
blocked (`social_security_number`, scored=2), the second user's identical
turn was allowed (`controls: []`, replayed=2).
The principal now joins the seed. Explicit session ids, the Claude Code
header and LiteLLM's own session are unchanged.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* fix(straiker): derive the session id with sha256 and drop comments that restated constants
The derived session now hashes the principal, and CodeQL flags MD5 over an
identity as a weak hash on sensitive data. SHA-256 truncated to the same
32 hex characters keeps the id shape. Comments that only labelled the
allowlist groups or restated a constant are removed; the two that explain
a non-obvious choice stay.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* fix(straiker): keep a blocked conversation blocked when it is replayed
Straiker de-duplicates turns it has already scored per session and
answers a replay `allow`, whatever the first verdict was. A client that
resends a blocked request, or grows the conversation past the blocked
turn, was let through: measured on 2026-09-20, `block` then `allow,
events_replayed=2` for the same session and body, and Claude Code's
automatic retry after the 400 turned a blocked poisoned-file read into
a pass.
The guardrail now remembers, per session, a fingerprint of every
conversation it blocked (a bounded, day-long in-memory cache) and blocks
a request that repeats or extends one without asking again. A different
session with the same words is a new conversation and is scored afresh.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* fix(straiker): scope the block memory by session or principal, never by content alone
A request with no derivable session keyed the replay memory on the
conversation fingerprint alone, so one caller's block could answer
another caller's identical request. The memory is now scoped by the
session, else by the principal, and a request with neither is not
remembered at all.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* fix(straiker): remember only a block that names a control, never one that comes from state
The replay memory kept every block, including one the platform returns
because a kill switch is engaged (`action: block` with `blocked_by: []`).
An administrator lifting the kill switch then left the conversation
refused by the remembered copy: measured on 2026-09-21, traffic stayed
blocked after `POST /inventory/agents/{id}/restore` returned `engaged:
false`.
The same words are the same attack tomorrow, so a control-named block is
still worth remembering; state is not ours to cache. The parsed verdict
now carries `blocked_by` so the two can be told apart.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Phimmasone Phonpaseuth <PhimmStraiker@users.noreply.github.com>
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
* fix(prices): add baseten/zai-org/GLM-5.3-Fast pricing with cost tracking e2e
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(e2e): assert message instead of comment on breakdown row
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(baseten): drop the live e2e cost tracking test
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: kerry <kerry@berri.ai>
* ci: add dashboard and core smoke checks across supported Python versions
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* ci: tighten merge smoke harness and keep mapped test diffs additive
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* ci: terminate proxy on readiness timeout and use contextlib.suppress in teardown
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: yuneng <yuneng@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(params): carry stream_chunk_size through litellm_params instead of provider params
* test(integration): fence stream_chunk_size out of every provider request body
* test(bedrock): type parametrized stream chunk test params
* test(integration): drop the contracts manifest resurrected by the main merge
* test(bedrock): type the stream_chunk_size test helpers
* test(params): finish AGENTS.md typing pass on stream_chunk_size tests
* test(integration): drop the covers marker from the stream_chunk_size wire test
---------
Co-authored-by: shrey kharbanda <shreshth@berri.ai>
* test(autoroute): wait for a valid fuzzy selection index and cancel the prompt on driver failure
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(autoroute): read the fuzzy selection through the public InquirerPy property
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(proxy): inject the HIBP client into change_password so the breached-password test never touches the network
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(mcp): only use the 200ms read timeout in the silent mode of the transport completion test
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(proxy): record HIBP requests so the ordering test asserts no lookup happened
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* ci(codeql): filter the weak-sensitive-data-hashing false positive on the HIBP k-anonymity lookup
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(cost-map): add Azure Foundry pricing for gpt-6-sol and gpt-6-luna
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(cost-map): give azure/eu gpt-6-sol and gpt-6-luna full model metadata
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: kerry <kerry@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
