* fix(a2a): bill completion-bridge agents at their own prices on /a2a send and stream
message/send on an agent served through the completion bridge (custom_llm_provider
agents such as LangGraph or databricks_agent) logged a $0, 0-token row, and
message/stream logged no row at all, so cost_per_query and per-token prices never
reached spend.
The bridge send path now records counted usage, the agent's pricing keys, and the
agent id on its logging object, the same way URL agents do. The bridge stream is
wrapped in the A2A streaming logger, which now reads the bridge's artifact event
for the reply text and gets the agent name from the proxy. Streamed calls also
count prompt tokens again, since the message text reader now reads pydantic
messages as dicts.
* fix(a2a): count chunked and Task replies once and skip agent prices on error replies
Bridge streams now follow the A2A append flag per artifact, so chunked
pydantic-ai, watsonx, and AgentCore replies count the whole reply once.
Send replies that are a Task count their artifact text. A JSON-RPC error
reply is logged at cost 0 instead of the agent's price.
* fix(a2a): keep one reply text per stream and bill every artifact and status reply
* fix(a2a): ignore a null error field on streams and run the stream per-token billing test
A stream event whose error field is null no longer zeroes the call's cost. The integration test for per-token stream billing now runs instead of skipping, and the URL agent SDK case skips when the optional a2a-sdk is not installed.
* fix(a2a): bill streams the client drops and skip the bridge's working placeholder
A stream the client closes or cancels mid-reply now logs its spend row at the agent's prices on both the bridge and URL paths, and a stream dropped before the agent sends anything still logs nothing. The bridge's "Processing request..." status no longer counts as reply text, so an empty bridge reply bills no completion tokens, as send does. Adds proxy-level tests that a bridge stream bills under the agent's card name.
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
Strategies, pre-call checks, cooldown and fallback helpers, and the auto, complexity and quality
sub-routers now take a RoutingContext instead of the Router, and read the router only through
registry(), settings(), health() and the context's forwarding methods. Router-free helpers move
from router_utils to litellm_core_utils, outside callers of router state switch to entry points,
and the surface allowlist shrinks to the endpoint methods plus the entry points
* refactor(guardrails): record which policy pipeline steps and deployment-level hooks ran
* fix(guardrails): name only the request pipeline steps that ran in x-litellm-applied-guardrails
* fix(guardrails): drop the response pipeline steps that never ran from x-litellm-applied-guardrails
* test(guardrails): integration cells for request pipeline steps in x-litellm-applied-guardrails
* fix(guardrails): keep the agent cells on the shared proxy and shorten a long suppression line
* fix(guardrails): keep request steps when a response pipeline blocks before a later request pipeline
* refactor(guardrails): record which policy pipeline steps and deployment-level hooks ran
* fix(guardrails): name only the request pipeline steps that ran in x-litellm-applied-guardrails (#49263)
* fix(guardrails): name only the request pipeline steps that ran in x-litellm-applied-guardrails
* test(guardrails): integration cells for request pipeline steps in x-litellm-applied-guardrails
* fix(guardrails): keep the agent cells on the shared proxy and shorten a long suppression line
* refactor(proxy): inject database repositories through per-repository dependencies
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(proxy): model Prisma client cleanup
* test(rust): include NativeDatabase in bridge API surface
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(proxy): give database shutdown one lifespan owner
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(proxy): log database cleanup errors and bound shutdown waits
Database.aclose now logs a failure in either cleanup step instead of raising
out of the lifespan exit, which matches how main handled stop errors and keeps
setup_prisma_client from raising out of cleanup after a failed connect
Shutdown waits for an in-flight reconnect for a fixed five seconds instead of
the reconnect timeout plus grace, and joins engine wait threads with a fixed
limit instead of indefinitely. The threads are daemon threads, so a thread
still alive after the limit cannot keep the process from exiting
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(batches): cost an unmanaged gs:// batch through the deployment that created it
The create now stores the router deployment id as model_id inside the
managed object row's file_object JSON and CheckBatchCost reads it before
falling back to the first deployment matching the base model in the
gs:// path. Rows without the field keep the old lookup
* fix(batches): tolerate a non-string or malformed file_object when reading the stored deployment
* fix(batches): read the stored creating deployment from the writer, not a lagging read replica
* fix(batches): resolve the writer database once so the type gate count does not grow
* fix(batches): decode a dict-shaped file_object when reading the stored deployment, accumulate wire samples, annotate test locals
---------
Co-authored-by: devin-ai-integration[bot] <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(sso): create the team_id_jwt_field team on UI SSO login when team_id_upsert is set
* fix(sso): keep the upserted team_id_jwt_field team out of result.team_ids so restricted_sso_group and the access-token group fallback are unchanged
* fix(sso): only upsert the team_id_jwt_field team on UI SSO login when enable_jwt_auth is on
The API JWT path only reaches team_id_upsert inside the enable_jwt_auth branch, so a deployment
with a leftover litellm_jwtauth block and JWT auth off never created teams. Gate the UI SSO upsert
the same way
* fix(sso): read enable_jwt_auth through the proxy_server module so the basedpyright gate stays at its ceiling
Seven integration tests fail on main for reasons that are not product bugs. Two are order or race bugs in the tests, one reads rows other tests wrote, one passed on an error body, and three still expect behavior that #48644, #39864 and #39540 changed on purpose
* fix(proxy): raise live budget counters for failure spend rows
A stream that breaks mid-flight writes its recovered cost to the spend log, but the failure hook
had already released the reservation and never touched the live counters, so an over-budget key
kept passing until the counter was next repaired from the database
Every charged failure row with a cost now raises the same counters the success path raises. The
error waits at most a second for them, so a hung Redis cannot hold it back
Fixes#35568
* test(proxy): cover failure spend counters across budgets, endpoints and outages
Integration cells for team, user, end-user, tag and window budgets,
concurrent streams, billed retries, native /v1/messages, Bedrock guardrail
cost, a Redis stall on two workers and disable_spend_updates
* test(proxy): check stalled failure counters and use the real spend writer
- The Redis stall cell now reads the burst key's live counter and requires at least the full failure spend
- Unit tests drop the charging writer double, bare dicts, list mutation and wall-clock timestamps
- Partial-stream test models are frozen
* test(proxy): move the stalled counter bound reason into the assert message
---------
Co-authored-by: yucheng-berri <293926326+yucheng-berri@users.noreply.github.com>
* fix(proxy): apply key and team router_settings to /v1/images/generations
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* style(proxy): format router settings helper
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): resolve key model_group_alias once on image generations
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): persist sent model_info fields on POST /model/update
POST /model/update rebuilt litellm_params from the request but dropped the
request's model_info, so changing base_model (or mode, tier, etc.) on an
existing DB model was silently ignored. Clients that only use this endpoint,
like the Terraform provider, could never fix a deployment's base_model
without deleting and recreating it.
Persist the model_info fields the caller actually sent, on top of the stored
ones. ModelInfo defaults never overwrite stored values, server-owned, team,
PTU and pricing fields stay on their dedicated write paths, and values echoed
from a /model/info response are not pinned.
* fix(proxy): keep echoed and blank model_info values out of POST /model/update writes
Blank strings from typed clients no longer overwrite stored values. Values echoed from
/model/info (the stored base model's cost map entry and discovered limits) are not
pinned, and an echoed catalog value resets a differing stored override like PATCH does.
direct_access and access_via_team_ids stay display-only. The tests stop patching the
auth check class and give the team lookup a real row instead.
* fix(proxy): lock the model row and skip unchanged model_info on POST /model/update
* fix(proxy): treat only keyed /model/update bodies as echoes and write the stored model_info as saved
A typed model_info value that matched the cost map, or the previous base model's entry,
was dropped and could erase a stored override. Echo detection now follows PATCH, only a
body carrying the /model/info key counts, compared against the base model the row has
after the write. The locked write patches the stored JSON instead of dumping the hydrated
Deployment, so litellm_params pricing is no longer copied into model_info. Tests run the
real Router with discovery over a mock upstream
* docs(terraform): note that model_info updates now take effect and overwrite values set outside Terraform
* test: expect POST /model/update to save the cache point flag, and wrap long test lines
test_x06 pinned the old contract, where POST /model/update answered 200 and ignored model_info.
This PR makes POST save model_info the way PATCH does, so x06 now asserts what its PATCH sibling
x01 asserts, the stored flag flips to true and cache points come back on.
* fix(proxy): answer 400 when POST /model/update gets an access_groups that is not a list of strings
* fix(proxy): read discovered model info through the router registry entry point
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* test: bedrock claude prompt caching on /v1/messages reports and charges cache tokens
* ci: re-run after ci-coverage and assert-shard-coverage were cancelled by the GitHub Actions incident
---------
Co-authored-by: devin-ai-integration[bot] <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): evict a deleted key from every worker's auth cache
/key/delete dropped the key object only from the in-memory cache of the worker
that served the call. Auth serves key objects cache-first, so every other
worker, and the Redis copy when enable_redis_auth_cache is on, kept
authenticating the deleted key until user_api_key_cache_ttl expired. /key/update
already evicts and broadcasts through evict_and_broadcast, delete now does the
same for the tokens it deletes
* fix(proxy): evict only the deleted rows and broadcast them in one batch
Scope the cross-worker eviction on /key/delete to the rows the caller was allowed to delete and send it through delete_cache_key_objects, so a caller-supplied name can no longer drop another team's cached entries on every worker
* refactor(proxy): drop the comment above the delete eviction call
* fix(proxy): run the usage cache eviction beside the auth eviction so a hung Redis costs one wait
* test(proxy): wait on the broadcast instead of a sleep and type the peer eviction tests
* test(proxy): wait for the peer to apply the broadcast instead of polling its cache
* test(proxy): cover hashed keys and aliases when deleting a key warmed on a peer
* fix(key_management): evict every requested key on delete, including keys whose row is already gone
A batch delete that names a live key and a key whose row was removed outside the API evicted only the live key, so the gone key kept authenticating from every worker's cache. Every requested key hash is evicted now, while raw non-key cache keys are still left alone
* test(key_management): split the gone-row eviction assertion into one comprehension per worker
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(spend): keep a NaN or infinite cost from wedging daily spend writes
A non-finite provider-reported cost now counts as not reported, so the token price bills the request. Any other non-finite cost is logged and recorded as 0, and spend log JSON writes NaN and Infinity as null so Postgres accepts the row.
* fix(spend): price non-finite provider costs on every route and drain parked NaN batches
Perplexity, OpenRouter pass-through, and Responses streaming now treat a NaN or infinite reported cost as unreported, so the token price bills the request. A daily spend batch parked in the Redis buffer by an older version with a NaN spend now drains as 0 instead of failing every flush. The cost header never carries NaN or Infinity, and a set holding a non-finite float no longer breaks spend log JSON
* fix(spend): record non-finite savings as 0 and warn when a parked NaN batch drains
An overflowing price on a compressed request made compression savings infinite, which still wedged the daily spend batch. Every savings column now records a non-finite value as 0 with a warning. The Redis drain warns when it reads a NaN an older version parked, since the spend summed into it is lost from daily spend. The writer test now runs the real payload builder and spend log insert
* fix(spend): keep non-finite costs out of headers, responses, and admin JSON
Writing NaN as null is now opt-in for spend log columns, the Redis buffer, and
streamed frames, so admin writes and logging integrations keep their output.
The cost calculator records an overflowing total as 0 with a warning, cost
headers leave out non-finite values, a non-streaming NaN usage.cost is dropped
instead of failing the response, and a parked NaN spend log row reads back as 0.
* fix(spend): keep caller-set call ids from splitting the non-finite cost warning
The overflow warning strips newlines from litellm_call_id, and neither
non-finite cost warning prints the raw value, which said nothing beyond
"non-finite"
* fix(spend): drop non-finite Responses and OpenRouter costs and the breakdown of an unpriceable request
* fix(types): drop a usage cost breakdown that holds a non-finite number
* fix(streaming): keep a dropped non-finite final usage cost from falling back to an earlier chunk's cost
* fix(spend): keep the dropped-cost flag out of reach of provider usage keys
A provider usage key named dropped_non_finite_cost shadowed the Usage method on the instance, so the stream builder raised and dropped the request's spend. The builder now reads the flag through the class, and Usage sets the flag after copying provider keys, so a _dropped_non_finite_cost key cannot set it either.
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(router): check and count deployment itpm and otpm on every Responses WebSocket turn
The success hook only reconciled an ITPM/OTPM reservation found in top-level
metadata or the standard logging object, but on the proxy the reservation lives
in litellm_params.metadata, so no reservation was ever settled there. The hook
now always calls the reconcile, which already returns early without a reservation
and clears what it settled, so a second success event is a no-op.
A Responses WebSocket turn after the first now has its deployment's current-minute
ITPM and OTPM counters checked and is refused with a 429 frame once a counter is at
its limit. An admitted turn carries a zero reservation on the current minute's
keys, so its actual tokens land on those counters on both the native and the
managed relay.
* fix(responses): settle the socket open's itpm/otpm reservation on the first turn only
After the first turn takes its logging copy, both websocket relays drop the
open's ITPM/OTPM reservation keys from the connection metadata, the same way
the budget reservation is detached. Later turns, unsolicited backend events,
turns after limits are removed mid-socket, and sockets with no turn admission
no longer settle or refund the open's reservation a second time.
* refactor(responses): read the itpm/otpm reservation keys from core metadata helpers instead of router_utils
* fix(responses): keep a socket open's hold for its first real turn and check tokens before request limits
A turn that will settle the socket open's itpm/otpm reservation no longer counts
that reservation against itself or swaps it for a zero one, so Codex's warmup
frame on a managed provider, or a refused first frame on the native socket, no
longer blocks the first real prompt or strands the hold for the minute.
The token check now runs before the rpm/rpd checks, so a turn refused for
tokens never uses up a request slot. Token-check tests pin the clock, and the
managed relay settle test drives a real streamed call through a custom provider.
* test(router): pin the router clock in the socket turn admission tests
Both tests seeded this minute and the next to dodge a minute boundary.
They now use the file's pinned-minute fixture and seed only that minute,
so neither depends on when the test runs.
* fix(router): check a responses socket's open without reserving and read shared token counters first
A Responses WebSocket open now runs the same itpm/otpm check a turn runs and
reserves nothing. It used to reserve an estimate that falls back to the
model's max output, so a socket could not open under any otpm below that.
The open still records a zero reservation for its own minute, so the first
frame, which the relay forwards without a turn check, counts its actual
tokens like every later turn.
This drops the plumbing that only carried the open's hold into the first
turn. The turn admission callable takes no arguments again, and the native
relay detaches only the budget reservation, as before this change.
The itpm/otpm turn check reads Redis first and falls back to this
instance's in-memory count only without Redis or when the circuit breaker
is open. These counters go down when a call settles, so an in-memory copy
at the limit could refuse turns after the shared count dropped. The tpm
check keeps its in-memory shortcut, since its counter only grows within a
minute.
Responses usage that reports input_tokens_details.cached_tokens now leaves
cached input out of itpm, as chat usage already did.
* fix(router): leave a socket turn's cached input out of itpm
The native Responses WebSocket relay logs a turn with its list of events,
so the deployment rate limit hook finds no usage on the response and falls
back to the standard logging payload. That fallback read cached input from
metadata.cache_read_input_tokens, a key the payload's cleaned metadata never
carries, so every cached token of a socket turn was counted against itpm.
The fallback now reads the cached count from the payload's usage_object,
the same usage the logging path already built for cost, so a native socket
turn counts only its uncached input. The managed relay and HTTP
/v1/responses already counted correctly through the response's own usage
* fix(router): mark only responses socket opens as check-only and count a turn on the minute it finishes
The check-only branch for a socket open keyed on a websocket value in the request kwargs, which any HTTP body can carry and which the realtime route also sets. The router now passes an opens_responses_socket flag into set_io_token_rate_limit_request_kwargs only for its _aresponses_websocket call, and the flag lives in the same context value as the request kwargs, so no request body can set it and realtime opens reserve and get refused as before
A zero reservation holds nothing to refund, so a socket turn now counts its real tokens on the minute the call finishes, read from the clock at settle time. A non-zero reservation still settles on the keys it reserved on so its refund hits the same counter
When the shared Redis read returns nothing, the check now falls back to this instance's in-memory count instead of admitting the turn with no count at all
* fix(router): keep io token reservation keys out of the metadata sent to providers
The itpm/otpm pre-call check wrote its reservation keys into both metadata and litellm_metadata. On the routes the router serves through its generic helper (Responses, Anthropic Messages, and the rest), metadata is the client's own field and goes to the provider as is. So _litellm_itpm_cache_key and _litellm_otpm_cache_key, which carry the deployment id and the model name, reached the provider, came back in the echoed response metadata, and counted toward the provider's 16-key cap. The managed Responses socket relay merged each turn's reservation into the frame's metadata the same way, and a native socket opened with metadata sent the keys on every turn through its request defaults
The router now passes the limiter the metadata field it uses for its own fields on that call (metadata for chat, litellm_metadata for the generic routes), and the reservation goes only there. metadata_channel is a required keyword, so a new caller has to name the field it writes. The managed relay puts each turn's reservation only on litellm_metadata
With the reservation only on litellm_metadata, a retried call on the generic routes refunded its first attempt twice and pushed the deployment's counters below zero. The failure hook read the reservation from the logging copy in litellm_params.metadata and cleared only that copy, so the router's litellm_metadata still held it and the retry refunded it again. The hooks now also read and clear litellm_params.litellm_metadata, which is the router's own dict
The proxy does not strip client keys that start with _litellm_ on the way in, so dropping that prefix at the provider boundary would also drop keys clients send today
The io token tests now run Redis in process with fakeredis instead of a TCP fake server
* test(responses): count socket turns through real logging callbacks instead of a Logging subclass
The native relay test now hands the rate limit check to a real Logging object as its success and failure callback, so the turn's tokens go through the same dispatch path production uses.
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(mcp): bill both rounds of a streamed chat completions MCP auto-execute
A streamed /v1/chat/completions request with an MCP tool set to require_approval "never" runs two model calls on one logging object, and the final-stream dedupe kept only the first round's spend row. The final usage chunk also showed only the second round.
The MCP stream iterator now holds the first round's assembled response, prices it, and logs one success row that carries the usage and cost of both rounds, with the fixed margin charged once. The final usage chunk shows the same summed usage. A single-round stream still logs its one row when the answer finishes.
When the follow-up round fails, the first round is logged from a copy of the logging object taken before the follow-up opened, and the failure row carries only the follow-up's partial usage with no second fixed margin. With proxy post-call guardrails armed, only (assembled, cache_hit) tuples are parked for the proxy, and the deferred log is armed only while the iterator reads, so the proxy's orphaned-callback cleanup never sees it after the Router's first-chunk read
* fix(mcp): send the usage chunk of a streamed chat MCP request that ends in one round
When a streamed chat completions MCP request ends in its first round (the model answers without a tool call, or no tool results come back), the iterator drained the inner stream to fire its spend log and dropped the usage chunk that stream produced, so a client that asked for include_usage got no usage chunk at all.
The drained usage chunk is now held and sent to the client after the final content chunk and before [DONE], the same place a plain stream sends it, so the last usage chunk matches the billed row. A two-round request still drops the first round's usage chunk and sends the merged one from the follow-up, and a stream without include_usage still sends none
* fix(mcp): hold round one's log for the guardrail, bill the chunk cost once, and keep round one's id
When the follow-up call of a streamed chat MCP auto-execute fails, round one's
success log now goes through the proxy's deferred guardrail release instead of
straight to the handlers, so a post-call guardrail runs on it and the row
carries it. Without a proxy release it logs directly as before
The last usage chunk now takes its cost from the combined row's billed total,
so a fixed cost margin is counted once in the chunk, as it is on the row. The
follow-up's usage chunk is held until the follow-up ends to make that possible
The combined row now keeps round one's id, the stream's first chunk id, which
matches how a single-round stream is logged
* fix(mcp): bill every chat MCP stream round once on failure, hang-up, and sync drop
When the follow-up call fails, round 1's success row is now logged on a copy of the
logging object that does not share the request's guardrail results, so a post-call
guardrail that reports on the failed request counts once instead of on both rows
A client hang-up during the follow-up now bills round 1 plus the follow-up's partial
answer on one success row with round 1's id and the fixed margin counted once, the
same way a plain stream's hang-up bills its partial answer
The last usage chunk now carries round 1's usage and cost when only round 1 is billed,
and a synchronous reader that drops the stream mid-answer bills round 1 on the loop it
read with instead of discarding it
The deferred release's logged_on argument is typed and every release closure accepts
it, and the parked dispatch constants and protocols both MCP billing paths use now live
in one module. The unit tests wait on a logging signal instead of a fixed delay and
check the billed row through the proxy stream hook instead of private callback state
* test(mcp): drive the unassembled follow-up through an upstream audio delta that is not base64
The test replaced litellm.stream_chunk_builder with a stub that raises. Round 2 now streams an audio delta whose data is not valid base64, so the real builder raises, the stream parks only best-effort usage, and the client's last usage chunk must carry round 1's tokens and cost
* fix(mcp): count the fixed margin once in the combined stream row's usage cost
* fix(mcp): bill round 2's partial answer when a client hangs up on a routed stream
On the proxy the Router wraps the MCP stream, and when a client hangs up during
round 2 the router's own cleanup closes that stream before the proxy's disconnect
cleanup runs. The close logged round 1 alone, so round 2's partial answer was
never billed. The MCP stream now bills round 1 plus round 2's partial answer as
soon as the hang-up's cancellation reaches its read, before anything closes it
* fix(mcp): count the follow-up's own messages in round 2's partial prompt
The follow-up stream shares the request's logger, so it took the client's
original messages as its prompt. When round 2 ended early, from a hang-up or a
broken upstream, its partial usage counted only those messages and left out the
assistant tool call and the tool results the follow-up actually sent. The
follow-up stream now counts the messages the follow-up call sent
* fix(mcp): bill round 2's partial when a routed stream's client hangs up between reads
On a proxy the stream the disconnect cleanup holds is the router's fallback wrapper, so the
cleanup never found the MCP stream's in-flight billing and released round 1 alone when the client
hung up while the proxy sat at a yield. The round billing now registers its in-flight biller on
the request's logging object, next to the parked release, and the cleanup looks it up there, so
any wrapper chain bills round 1 plus round 2's partial once
* fix(mcp): send the MCP stream usage chunk under the billed row id and count the partial off the loop
A streamed chat completion that runs a LiteLLM-hosted MCP tool bills both model calls on one row under round 1's
chatcmpl id, but the usage chunk it sent the client was round 2's held chunk and kept round 2's id, so a caller could
not join the usage chunk to its row. The usage chunk now carries round 1's id in every path that sends it, with or
without a deferred post-call guardrail, while its usage and cost and every content chunk's id stay as they were
When the client hangs up during round 2, assembling round 2's partial answer counts the follow-up messages, tool
result included, with the local tokenizer. That ran on the event loop and stalled every other request on the worker
for about a second on a large tool result. The assembly now runs on a worker thread through asyncify from both the
reader's cancellation and the proxy's disconnect cleanup, shielded from cancellation, and round 1 is checked again
after the await so a release that happened meanwhile is never billed twice
* test(mcp): wait on the loop with a hang guard instead of a 300 ms deadline
* fix(mcp): give a proxy-rejected follow-up's failure row round 1's own cost breakdown
* refactor(streaming): fold the partial response's empty guards into one check
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(proxy): name shared-handler errors by the status they answer
handle_exception_on_proxy pinned type=internal_server_error on every
exception it wrapped, so a management route rejecting a caller's own
argument answered {"type":"internal_server_error","code":"400"} and an
SDK branching on the type read its own bad request as a gateway outage.
Derive the type from the status the response actually carries, keeping
a type the exception already names.
* test(proxy): pin a 500 for exceptions whose status is None
* fix(proxy): keep internal_server_error for timeouts and 5xx, redact hosts in shared handler errors
Provider 5xx and timeouts keep internal_server_error so retry loops still retry them,
the shared handler redacts internal hosts from raw exception messages, and /key/list
routes its HTTPException through the shared handler instead of its own copy of the
old label.
* fix(proxy): label shared-handler errors like the chat route, pick the status before the type, redact only gateway failures
* fix(proxy): pass shared-handler error messages through as on main
Redacting timeout and 5xx messages changed what every shared-handler route
returns, which relabeling the error type does not need, so messages go back
to the verbatim text main sends.
* fix(proxy): redact only the status-less failure main never sent
* fix(proxy): answer a generic 500 for a status-less failure and expect auth_error on the /user/new 403
A failure with no status now answers "Internal server error" instead of its own text, since that text can
name a guardrail's private host and pattern redaction misses cluster DNS names and IPv6 addresses. The
/user/new integration test now expects the auth_error type main's route catch-alls give a 403 the proxy raises
* fix(proxy): answer only a 4xx or 5xx carried status and fall back to the route default otherwise
A numeric-string status outside 400 to 599 made /v1/messages, images, rerank and the OpenAI pass-through answer HTTP 200 with an error body or drop the connection, and non-ASCII digits or a 5000-digit string slipped through isdecimal. error_status_code now accepts an int or a three-digit ASCII string in 400 to 599, the range the chat route already enforces
* refactor(proxy): parse a three-digit string status without recursing
* fix(bedrock): route unmapped converse-only family model ids to converse
#42713 stopped Bedrock-native OpenAI ids that are not in the cost map yet from
falling to the invoke route. The same gap exists for the other families that
have no invoke transformation at all: minimax, nvidia, writer, zai, google and
xai. An id from one of them that is newer than the cost map hits
get_bedrock_route -> "invoke" -> get_bedrock_invoke_provider() -> None ->
BedrockError, when Converse would have served it
Add a second family regex for those six and default them to converse, the same
way the openai family is handled. _OPENAI_FAMILY_MODEL_RE is left alone because
the prompt-caching rule is openai-specific
Every shipped id of these families already resolves to a converse entry via its
base model (checked against the cost map), so this only changes behaviour for
ids that are not in the map yet
Tests: six unmapped ids route to converse, and an unmapped meta.* id still
routes to invoke since meta has an invoke handler
* fix(bedrock): route a model id with no Invoke provider to Converse
A bedrock/<model> id missing from the cost map fell to the Invoke route,
where an id that names no Invoke provider (a new vendor's model, a bare
provisioned-model, imported-model, or custom-model-deployment ARN) fails
locally with "Unknown provider=None" before any request is sent. Such ids
now go to Converse, which replaces the per-family regex with the general
rule. Ids that name an Invoke provider and explicit route prefixes are
unchanged.
* test(bedrock): expect Converse for ids with no Invoke provider in the nova ARN and not-found tests
* fix(bedrock): accept tools on Converse for a model with no cost-map row or Invoke provider, and keep a 404 for an invalid id
* fix(bedrock): answer 404 for an invalid model id instead of AWS's 400
AWS answers a mistyped Bedrock model id with a 400 ValidationException. The mapping already raised NotFoundError, but passing AWS's 400 response through made the openai base class report 400, so the proxy answered 400 and the router skipped its cooldown
* fix(bedrock): drop lookaround regexes from tool schemas for a model the cost map does not list
* fix(bedrock): apply the unknown-model defaults only to bare ids LiteLLM knows nothing about
A bare Bedrock id with no cost-map row and no Invoke vendor gets Converse's open params, its reasoning_effort as sent, lookaround-free tool schemas, and a 404 for AWS's invalid-identifier reply, which keeps AWS's response headers. An explicit route like converse/, an ARN, or an id LiteLLM recognizes keeps main's params, regexes and error status, so drop_params still drops tools there
* fix(bedrock): judge lookaround support by model id on every route, ignore price-only rows
The regex check now strips route and region prefixes from the model and its
base_model before asking whether LiteLLM knows the model, so
bedrock/converse/<id> gets the same answer in a direct transform call as on the
completion path. A cost-map row counts only when it names a provider, so a
price-only register_model row leaves an unlisted id unknown, as on the proxy.
AmazonInvokeConfig now inherits the Invoke provider lookup instead of keeping
an identical copy.
* fix(bedrock): treat a region path like the bare id when the cost map does not list it
A deployment like bedrock/us-east-1/<unlisted id> now gets the same tools,
reasoning_effort and typo 404 as bedrock/<unlisted id>, so Claude Code works
on it and a mistyped one still answers 404.
* fix(bedrock): keep today's 404 for ids that reach Converse only by default, and keep an empty id local
* fix(bedrock): answer 404 for an invalid id only when Converse rejected it, and keep a region path with no id local
* fix(bedrock): keep a path-shaped id local, and map /v1/completions errors with the model's provider
An id with a slash that is not an ARN, like bedrock/../../converse or an unknown region path, keeps Invoke and its local 404 instead of reaching AWS Converse.
atext_completion mapped every error as OpenAI, so a Bedrock typo on /v1/completions answered 400 while chat answered 404. It now resolves the provider from the request's model, as the other async entry points do.
* fix(text completion): keep /v1/completions error mapping as it is on main
Mapping async /v1/completions errors with each provider's rules is wider than this Bedrock routing change. It moves classes, statuses, cooldown and fallbacks for most providers, and some lose context-window fallbacks. So this restores main's mapping here and leaves that change to its own PR. A Bedrock typo on /v1/completions answers AWS's 400, while chat keeps the 404.
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(sso): bind CLI SSO and Claude Code gateway logins to the first team on the user record
* fix(proxy): keep SCIM and JWT team sync in record order, and look up SSO groups once
SCIM PATCH and JWT team sync saved a user's teams from a set, so the first
team on the record, the one a Claude Code gateway sign-in binds, could change
on any sync. Both now keep the record's order and append new teams in the
order the IdP sent them.
SSO group sync now drops claim groups that are not LiteLLM teams with one
lookup before adding the missing teams one at a time, so a claim full of
non-team groups no longer costs one failed add per group on every sign-in.
* test(sso): expect the gateway team dropdown in the user record's order
* fix(sso): read team existence from the writer, keep new-user and SCIM replace team order
* fix(scim): keep the teams a user already has in record order on PATCH and PUT
* fix(scim): keep record team order when a SCIM POST matches an existing email
* fix(scim): join new teams in request order so a failed roster write keeps it
* fix(scim): let a groups replace win over later group ops in a PATCH again
Applying group adds and removes after a replace made a PATCH that deactivates a user
and adds a team that does not exist answer 404 and leave the user active, where it
answered 200 before. Keep the record's team order, but ignore group ops after the last
replace as the base did.
* test(sso): type the CLI team detail row helper
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
Co-authored-by: rj-cairn <349020452+rj-cairn@users.noreply.github.com>
* feat(usage): show subscription-covered providers and account fees separately
* fix(usage): write only the named fee columns and sync the root schema
The fee update now writes just the columns the request names, so two admins changing
different fields at once both land, and the dashboard edit form sends only what changed.
The root schema.prisma carries the new table like the two copies, and a billing period
anchored in the calendar's last month ends on December 31, 9999 instead of failing.
* fix(usage): register the subscription table span and keep provider logic under llms
* fix(subscription_accounts): register the fee routes as management routes and reject an empty account id
* fix(usage): run the new tests in CI and close subscription card review gaps
Adds the subscription test file to its unit shard, sets the migration lock timeout, rejects infinite fees, keeps the card from rewriting login files, lists each deployment once without aliases, shows the provider badge on entity views, and asks before removing a fee
* fix(usage): resolve named-credential ChatGPT accounts and keep admin viewers read-only on the fee card
* fix(usage): count a subscription fee only for accounts behind a current deployment
Usage listed a fee row for an account no deployment uses while another ChatGPT
deployment existed and dropped it when none did. It now lists only accounts behind a
current deployment in both cases. Adds a real proxy and Postgres integration test for
the fee API and for ChatGPT Responses rows on a deployment's own login, and names the
card's multi-component test with the integration suffix.
* fix(usage): keep a stored subscription fee listed after its account's deployments are removed
The usage route lists every account behind a subscription deployment plus every account carrying a stored fee, and reads the fees on every proxy, so a fee stays visible, counted, and editable for past ranges after its last deployment goes. The integration test now polls every day the calls spanned, so a UTC midnight crossing no longer fails it.
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(proxy): apply published model rows on every worker before the debounced resync
Config sync messages now carry the ids of model rows a create, update, or upsert wrote. Every worker that receives one loads those rows by id right away, including while its debounced full resync is still waiting, so a group added one deployment at a time is complete on every worker as soon as the write returns. Deletes, bulk writes, and messages from older pods carry no ids and keep the full resync path
* fix(proxy): load published model rows from the primary under the reconcile lock
Read the by-id rows through the writer so a lagging read replica cannot put an old row back, and read them inside MODEL_RECONCILE_LOCK so a concurrent delete or reload stays the last word. Look the ids up in bounded chunks, reload DB credentials only when a row names one this worker lacks, and leave auto-router rows to the debounced full resync. Add an integration test that boots two proxies and needs the peer to fall back to a just-added backup within 4 s.
* test(proxy): cover a published model row that names no credential
* fix(proxy): keep the unknown model name read-through on the replica and outside the lock
Only the by-id load that follows a published write reads the primary inside
the reconcile lock. A request for a model name a worker does not know reads
through the replica routing before taking the lock again, as on main, so
unknown names put no load on the primary and do not wait behind a reload.
* fix(proxy): keep access group edits from dropping staying deployments and batch published config ids
* fix(proxy): read access group rows from the primary so quick edits are not lost to a lagging replica
* fix(proxy): install ordinary rows queued with an auto router and check access group edits before writing
A batch of published model ids that held one auto router skipped every row in it, so ordinary
rows written at the same moment waited for the full reload. The by-id load now leaves only the
auto router rows to that reload.
An access group update now reads every deployment from the primary, answers 404 for an unknown
group and 400 for an unknown model id or name before it writes anything, then removes the group
from the deployments that leave before it adds it to the ones that join, so a failed write leaves
the group narrower than asked, never wider.
* fix(proxy): write an access group update in one transaction after a fresh read
The update now checks model names first, reads every deployment from the
primary, and commits every join and leave in one batch right after that
read, so a failed write changes nothing and a field another admin set
during validation is kept. Other workers hear about the rows only after
the commit.
* fix(proxy): lock the model rows while an access group update reads and rewrites them
* test(proxy): run the model resync tests on the real installer and a real router
The by-id resync tests now install rows through ProxyConfig's real installer and credential loader on a real Router, faking only the database, and the config sync test row model is frozen.
* fix(proxy): lock only the rows an access group update touches and answer 404 before 400
An access group update locked every model row, so a lock on any unrelated
row stalled it for 30 s and then failed it with a 500, and unrelated model
edits on other proxies waited behind it. It now takes a per-group advisory
lock, locks only the rows that carry the group or that it selects, and
reads them again once locked. The group existence check runs before the
model name check again, so a missing group answers 404 as on main.
* fix(proxy): never wait on a locked row in an access group update
The update now takes its row locks with NOWAIT. When another edit holds one
of the group's rows, it rolls back, holding nothing, and tries again for about
10 s before answering 409. Holding rows while waiting deadlocked against team
deletes and let a concurrent model edit undo the update's change. The
per-group lock also moves to the two-key advisory space, apart from team ids.
* fix(proxy): retry access group row locks for 30 s with SKIP LOCKED and lock only the rows model_ids select
* fix(proxy): take the access group lock with pg_try_advisory_xact_lock and retry while another edit holds it
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(router): keep a deployment's own model_id over the request body's
A request body model_id overrode the deployment's pinned model_id, so a client could move SageMaker chat and embedding calls to another inference component. The router now drops the request's model_id when the deployment sets one.
* fix(router): let deployments opt back in to a client model_id and stop client component headers on SageMaker chat
Drop the request's model_id after the router defaults are applied, so a default_litellm_params model_id cannot win
over the deployment either, and apply the same precedence on the sync text_completion path. A deployment that lists
model_id in configurable_clientside_auth_params keeps the client's value. SageMaker chat drops any case variant of a
client's X-Amzn-SageMaker-Inference-Component header when the deployment sets a component
* fix(router): keep a SageMaker deployment's component when it pins it through extra_headers
* fix(router): read the SageMaker component header as a pin only on SageMaker deployments
* fix(sagemaker): send the inference component only as a header, never as a body model_id
SageMaker picks the component from X-Amzn-SageMaker-Inference-Component, so the chat
container body and legacy parameters no longer carry model_id. Nova endpoints accept a
fixed body field list, so a pinned Nova deployment could reject the extra field.
* fix(health): probe header-pinned SageMaker deployments at their component
Health checks bypass the router, so a legacy sagemaker/ deployment pinned only
through its inference component header sent no component on /health while routed
requests now send it. Also update the two SageMaker integration tests that still
asserted the old precedence and the dropped body field.
* fix(health): apply the router's model_id pin rule to health probes
Health probes now share the router's pin function, so a deployment that lets
clients set model_id gets no header pin and its own model_id wins without
reading the header. Header keys that are not strings are skipped when
reading the inference component header instead of failing the call.
* fix(sagemaker): keep header keys that aren't strings and report a string opt-in in health checks
The chat transformation now keeps extra_headers keys that aren't strings, so such a deployment fails the way it did before instead of answering 500, and the model_id pin checks configurable_clientside_auth_params with isinstance, so a string there reaches the health probe's own error capture again.
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(realtime): send Gemini Live setup on the client's first message so session.update modalities apply
Gemini and Vertex AI Live accept exactly one setup per session, and the proxy sent its
default audio setup at connect, so a client's session.update (GA output_modalities or
beta modalities) was dropped and every answer came back as audio. The setup now goes
out on the client's first message: a session.update becomes that setup, anything else
gets the model default. Bedrock Nova Sonic keeps its connect-time setup through a
provider hook. Text setups label deltas response.output_text.delta, and downgrading
TEXT to AUDIO on an audio-only model logs a warning
* fix(realtime): keep one Gemini Live response modality and skip the "inf" token limit in the setup
* fix(realtime): merge every session.update sent before the first content frame into the Gemini Live setup
Gemini and Vertex Live accept one setup per connection, so the bridge used to
turn only the first session.update into that setup and drop the rest without
telling the client. Now every session.update that arrives before the first
content frame is held, deep-merged (later values win per field), acknowledged
with a session.updated built from the setup it would produce, and the merged
session becomes the one setup sent with the first conversation item, audio
frame, or tool result. The session events also surface the systemInstruction
text as instructions, which the old string check never matched because the
setup carries it as a Content object.
* fix(realtime): cap the session.update merge depth so the recursive detector accepts it
_merge_session_fields recurses once per nested mapping level, so CI's
recursive_detector flagged it. Bound the descent at eight levels (the OpenAI
session object nests four deep at most), replace anything deeper wholesale
with the later session.update's value, and list the helper with the other
depth-capped functions the detector allows.
* fix(gemini-live): accept a string output modality and drop values Live cannot produce
A bare string output_modalities is read as a one-element list and modality values Gemini Live cannot produce are dropped with a warning instead of reaching the backend, so a malformed session.update no longer costs the session. Transcription guardrails now fold their VAD disable into the first held session.update only, so the merged setup carries it exactly once.
* fix(gemini realtime): answer an error for session updates Gemini Live cannot apply
An output modality the model cannot produce, or a session.update sent after the
first content, now gets an invalid_request_error event instead of a silent swap
or drop, unless drop_params is set. The synthetic session.created reports the
model's default modality, setupComplete no longer adds an extra session.updated,
and merging nested session updates is linear in their size
* fix(gemini realtime): send the setup before guardrail blocks and idle deadlines, mark push-to-talk turns
A pre_call guardrail blocking the first typed message now goes out after the
setup, so Gemini and Vertex no longer close the session with 1007.
Push-to-talk sessions (turn_detection null or create_response false) wrap each
audio turn in empty activityStart and activityEnd messages, which Gemini needs
when automatic activity detection is off.
A session with no content after 60 seconds sends its setup, since Vertex
refuses a setup 300 seconds after connect. Held session.update frames past
10 MB answer session_update_too_large, and deferred mode now errors on a
second session.update unless drop_params is set.
* fix(realtime): ack unchanged Gemini Live session.update resends and honor null turn_detection
An update that leaves the applied setup unchanged now gets session.updated
instead of session_update_not_applied, so clients that resend their session
with the first response keep working. Deferred setup mode merges held updates
like the default mode, turn_detection null turns Gemini's activity detection
off, drop_params drops a voice outside the documented Live voices, and the
lazy setup deadline moves to 240 seconds
* fix(realtime): keep Vertex activity detection off for null turn_detection and set up deferred sessions before text content
* fix(realtime): refuse unscreenable Gemini audio, send held setup after buffered audio, reject unknown Vertex voices
A realtime_input_transcription guardrail cannot screen Gemini Live audio before the answer starts, so audio input under one is refused with an audio_input_not_screenable error instead of reaching the model. In deferred setup mode a session.update that follows buffered audio is sent as setup at once. Only conversation content triggers the setup, so control events no longer lock in a default one. Vertex rejects an undocumented voice on the session.update, or drops it with drop_params, while Google AI Studio forwards any name.
* fix(realtime): hold first content so a following session.update joins the Gemini setup
Convert flat realtime tool schemas through Gemini's schema builder, and collapse a text and audio pair to what the model can produce on text-only models too
* fix(realtime): check AI Studio Live voices like Vertex
* fix(realtime): forward any voice to AI Studio Live models that accept it
* fix(realtime): map OpenAI TTS voices before the Live voice check, keep outage spend rows, share the playground session config
The Live voice gate now judges the Gemini voice an OpenAI TTS name maps to, so fable and onyx reach Vertex as Umbriel and Orus again. An upstream close before the held setup was sent is logged as a session, not a refusal. The playground sends one server_vad session config for typed and voice turns, and the e2e, integration, and dashboard tests follow lazy setup.
* fix(realtime): take the lazy setup clock as a dependency so its tests run on a stepped clock
The deadline and grace tests waited on real 50 ms timers, and the merge cost
check compared wall-clock runs. The session now reads an injected clock,
time.monotonic by default, the two tests step it, and the merge cost check
moves to the CodSpeed benchmarks. The Nova Sonic style handler test passes
the backend through wrap_backend, which main added after the test was written
* fix(proxy): restore spend logging for /cursor/chat/completions (#30126)
The /cursor/chat/completions route processes requests through the Responses API pipeline, but _get_metadata_variable_name did not treat it as a litellm_metadata route. Proxy auth metadata therefore landed in the user-facing metadata param, so the spend tracking callback saw user_api_key=None and silently skipped the SpendLogs write. Add the route to LITELLM_METADATA_ROUTES and cover it with regression tests.
* test(proxy): pin provider metadata and key billing on /cursor/chat/completions
A Responses-format body on the Cursor route now sends OpenAI only the metadata the client sent and writes a spend row that bills the key. The chat-body branch, which this fix also moves onto litellm_metadata, still bills the key with its team and request tags. The scripted chat reply gets a unique id per test, since a shared id let only the first spend row per database land
* fix(proxy): pick the Cursor route's metadata field from the body shape
A Responses-style body on /cursor/chat/completions keeps proxy metadata in
litellm_metadata, the same as /v1/responses, while a Chat Completions body
keeps metadata, the same as /v1/chat/completions. Adding the whole route to
the internal-metadata list let a client's chat-body tags replace the key's
tags in spend logs and skip tag budgets.
* fix(proxy): keep get_metadata_variable_name callable with only the request
* fix(proxy): check client tags against tag budgets and read agent limits from litellm_metadata on Responses bodies
A Responses-style body on /cursor/chat/completions now keeps its proxy fields in litellm_metadata, so the pre-call tag budget check and the v3 agent rate limiter stopped seeing the client's metadata.tags and the x-litellm-agent-id value. The pre-auth tag check now merges the caller's metadata.tags into the selected metadata field, sharing one helper with request setup, and the limiter reads agent_id from either field
* fix(proxy): check client tags sent in a JSON-string metadata against tag budgets on Responses bodies
* fix(proxy): answer 422, not 500, when a budget rejects a Responses body whose metadata is a JSON string
* fix(proxy): keep key guardrails, metadata type errors and failure-row client details on Responses-style bodies
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(websearch-interception): stream native server-tool blocks to clients
Streaming native clients (Anthropic SDK, Claude Desktop, Claude Code)
never received the server_tool_use and web_search_tool_result blocks the
interception layer builds for them, so citation panels stayed empty and
result counts read zero. Two gaps in the same flow:
FakeAnthropicMessagesStreamIterator._create_content_block_chunks only
knew text, thinking, redacted_thinking and tool_use; any other block
type emitted a bare content_block_stop with no start and no payload,
silently dropping the block from the fake stream. server_tool_use now
streams as content_block_start plus input_json_delta and any other
server-generated block rides whole in content_block_start, matching
Anthropic's native streaming of these types.
try_short_circuit_search only recognized a native client by finding a
web_search_* tool in the tools list, but the pre-request and deployment
hooks convert that tool to the LiteLLM standard shape before the
short-circuit runs, recording the conversion in
WEBSEARCH_EMIT_NATIVE_BLOCKS_KEY. The flag now counts as native too, so
the synthetic response carries the native blocks on the converted path
instead of only when hook ordering happens to leave the tool untouched.
Fixes#35333
* fix(anthropic): accept server_tool_use and web_search_tool_result blocks when costing a /v1/messages response
The short-circuit now returns native web search blocks, and the cost path validated the served body against a content union that only allowed text, tool_use and thinking blocks, so non-streaming requests answered 500
* fix(anthropic): keep every server_tool_use field when re-streaming a /v1/messages reply
* fix(anthropic): keep tool_use fields, redacted thinking data and text citations when re-streaming a /v1/messages reply
* test(websearch-interception): build the cost test's logging object with a datetime start time
* fix(websearch-interception): send search snippets once in the direct search answer
* fix(guardrails): mask web search queries, titles and urls on /v1/messages and deliver stream rewrites past server_tool_use blocks
Output guardrails now scan the native web search blocks the short-circuit emits, ahead of the reply text so guardrails that read the last text still scan the reply. A guardrail reply of a different length falls back to the reply-only write-back. Ended-stream tool call rewrites land on server_tool_use blocks too, the ended-stream scan key covers the result strings, and legacy-hook pipeline steps pass through the result strings their response view lacks. The wire test reads the streamed server_tool_use input from its input_json_delta.
* fix(guardrails): deliver a legacy post-call hook's tool-call rewrites on streams instead of discarding the stream rewrite
A legacy hook that masked a tool call's arguments on a buffered stream made the whole
rewrite undeliverable, text included, so the original stream went out. With web search
blocks in the stream that now hit any hook masking the search query. The adapter hands
same-count tool-call rewrites to the translation's write-back, as a unified guardrail's
already are, and still discards a rewrite that adds or drops a tool call.
* fix(guardrails): show a legacy post-call hook the web search result blocks of a /v1/messages stream
The legacy hook's view of a buffered Anthropic stream is rebuilt from the assembled
ModelResponse, which keeps web_search_tool_result blocks only in provider-specific
fields, so a hook masking result titles on a stream never saw them. The hook's view now
carries those blocks, its rewrite of them goes back through the ended-stream write-back,
and the adapter no longer needs to pass texts outside the hook's view through untouched.
* test(guardrails): type the legacy hook stream test helpers
* fix(guardrails): scan a streamed web search query as text like the non-streaming reply
* test(guardrails): type the masking guardrail test helpers
* fix(guardrails): keep the block id when scanning a streamed web search query
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(azure_ai): keep OpenAI prompt cache breakpoints on the Foundry chat completions path
* fix(azure_ai): keep only OpenAI content part types when preserving prompt cache breakpoints
* fix(azure_ai): share the breakpoint check with the cache-mark step and keep only text parts
A deployment's model_info opt-in is registered under its provider-prefixed cost-map key without a litellm_provider, so that prefix now names the provider the flag speaks for.
* refactor(azure_ai): read text parts through a walrus and mark new test values Final
* fix(azure_ai): read a router-registered breakpoint opt-in only for azure_ai and keep a message mark off dropped file parts
* fix(azure_ai): read the breakpoint opt-in from each deployment's own model_info and base_model
A router deployment without its own supports_prompt_cache_breakpoint flag now
reads the cost map as loaded, so a sibling's opt-in on the same model string no
longer turns its requests into 400s. The shared hook no longer reads router
registered keys for azure_ai.
* fix(azure_ai): send breakpoints in Foundry's shape under drop_params and fall back to the model name
* test(integration): expect the injected breakpoint on the Foundry GPT-6 chat wire
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(router): cap the default OTPM reservation for requests without max_tokens
* fix(router): reconcile ITPM and OTPM reservations after every successful call
* fix(router): read the default OTPM reservation from the served model and make its fraction configurable
The default reservation now looks up max output from litellm_params.model, as main does, so a pricing base_model never changes it. NO_MAX_TOKENS_TPM_RESERVATION_FRACTION sets the fraction for both the router check and the key and team TPM floor. A max_tokens that is not a number or string counts as no cap instead of raising inside the check.
* fix(router): refund a failed attempt's io token reservation in make_call
The proxy passes one logging object to every retry, and it runs failure
callbacks only for the first attempt, so the last attempt's ITPM and
OTPM reservation stayed held until the minute ended. Each attempt that
raises now gives its reservation back before the retry or fallback
* fix(router): count Anthropic cache writes once in ITPM reconcile
Native /v1/messages usage reports input_tokens without cache reads and writes, so the reconcile now maps it through AnthropicConfig.calculate_usage before subtracting cached tokens.
* fix(router): use a deployment max_tokens, charge late usage to its finish minute, and read the reservation fraction as a fraction
* fix(router): reserve with the deployment cap only through the param the request's API sends
* fix(router): never let a deployment's output cap raise the reservation above the default
* test(router): move the reservation fraction env var test to the integration suite
The unit test started a child interpreter, which the unit guide forbids. The integration test boots a real proxy with the env var set and checks admission against the held reservation.
* test(router): retry the reservation fraction scenario when the minute rolls over
The OTPM counter is keyed by minute, so a rollover between the held stream and the capped requests gave the capped request a fresh counter. The control request now leaves room for usage the first attempt left in the new minute.
* test(router): give the rollover retry its own deployment
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(cost): bill a provider reported usage.cost on every path unless the deployment is priced
* test(integration): a reported cost is attributed to output in the gateway parity cases
* fix(cost): skip a chat usage.cost that already holds litellm's stamped bill and Responses API objects
* fix(cost): record litellm's bill on the Responses terminal event so a later pricing pass does not read it as the upstream's charge
* fix(cost): mark usage.cost litellm wrote as its own bill and keep a breakdown for a reported cost
* fix(cost): keep the token breakdown when a header reported cost repeats an already priced bill
* fix(cost): read the stored breakdown total through the logging object's typed field
* refactor(cost): name the providers billed from their own usage fields for what they are
* fix(cost): keep the upstream's usage.cost beside litellm's bill and stamp a zero operator price on the stream frame
---------
Co-authored-by: devin-ai-integration[bot] <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): stop answering with the literal string "None" on the proxy-server, auth, and health routes
The error objects these routes return defaulted `type` and `param` to the
four-character string "None", so a client's error handler matched no known
OpenAI type and fell into its generic branch.
Route them through the shared openai_error_payload helpers instead, so `type`
comes from the carried type or the HTTP status and `param` serializes as JSON
null.
* fix(proxy): stop answering with the literal string "None" on the spend and management routes
The error objects these routes return defaulted `type` and `param` to the
four-character string "None", so a client's error handler matched no known
OpenAI type and fell into its generic branch.
Route them through the shared openai_error_payload helpers instead, so `type`
comes from the carried type or the HTTP status and `param` serializes as JSON
null.
* test(proxy): pin the mid-stream failure frame and the DB outage body to a JSON null param
* fix(proxy): type upstream timeouts as timeout_error and update integration asserts for the null param
A 408 with no carried type now answers timeout_error instead of invalid_request_error, so a client branching on type keeps retrying it. Three integration suites still compared auth rejections against param="None".
* fix(proxy): keep a 408 on main's invalid_request_error type and pin the transcription file cap's 400
* fix(proxy): type a route failure by the status it answers with after the tag-routing and no-deployment remap
* fix(proxy): keep auth_error on proxy-raised 401/403 route failures and ban the string None as an error type or param
* fix(proxy): type a hook's 401 or 403 on completions, moderations and transcription by its status, as chat does
Only a status the proxy remapped into 401 or 403, such as a tag-routing denial, keeps auth_error.
* test(proxy): build the transcription file-cap test on a real router so it holds after the router refactor
* feat(dashscope): add image edits and text-to-speech
* fix(dashscope): bill reported speech characters and input images, follow the chat region for edits, keep the call timeout on the audio download
* fix(dashscope): bill deployment image prices as main does and route edits to the host speech uses
A deployment that sets its own input_cost_per_image, output_cost_per_image or input_cost_per_pixel bills through the default image calculator again, so an override is never read as a per-input-image price and pixel pricing on qwen-image-3.0-pro no longer raises.
Edits now resolve the caller's api_base or DASHSCOPE_API_BASE through the same API-root helper speech uses, on any host, with DASHSCOPE_API_BASE_IMAGE still winning. An image response whose usage is null or not an object keeps its images instead of raising.
* fix(images): bill a deployment's own prices on image edits through the proxy
The proxy builds the logging object before the router picks a deployment, so image_edit never handed the deployment's price fields to its cost calculation and billed the cost-map price instead. Pass the custom pricing params with the pre-call update, as the Responses API does.
* fix(images): bill per-pixel image prices at a DashScope W*H size
DashScope's mapped size reads 1328*1328, which the cost size reader rejected, so a deployment's input_cost_per_pixel billed every DashScope generation or edit at the 1024x1024 default.
* fix(dashscope): bill image prices by their meaning, report the rendered size, prefer the deployment edit base, fetch audio over https
* fix(dashscope): read the rendered image size from output_width and output_height
* test(images): bill a fal edit at its deployment's own image price
* fix(dashscope): keep the qwen-image-3.0 input-image fee out of the shared cost map row
Released proxies read the cost map from main and bill input_cost_per_image as the
flat price of each DashScope image, so the row's $0.003 input fee billed every
qwen-image-3.0 image there at a tenth of its $0.03 price. A deployment can still
set input_cost_per_image to bill DashScope's input images
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(proxy): bill streamed /v1/completions without stream_options like chat
* test(proxy): mark the text completion stream usage test variable Final
* fix(completions): drop Azure's empty prompt-filter frame when the proxy asked for stream usage
* test(spend): drop the module docstring from the text completion stream usage test
* fix(proxy): keep the internal stream usage strip flag out of the logged request body
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(openapi): build native extension for base counts
* test(openapi): exercise native import in base worktree
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Without drop_params, every copy of a tool from additional_tools input items
is forwarded as sent, so a provider that rejects duplicate tools returns its
own 400. Codex keeps the merge because the proxy turns drop_params on for it.
* test: add a manually dispatched integration trial on larger runners
Runs the providers, database and management integration groups on 16-core GitHub-hosted runners with Postgres and Redis as service containers, to measure wall time and how far each group scales with more workers
run_integration.sh now keeps a preset INTEGRATION_WORKERS, takes the service container names from the environment, and seeds test order from the GitHub run when there is no CircleCI one
* test: run the integration trial on push to this branch
* test: give integration trial workers a local redis-server
* test: run the providers trial on a 64-core runner with 32 workers
The 16-core run reached 53% of providers in 45 minutes with 16 workers, which points at CPU. Postgres gets max_connections=1000 and a 1 GB shm because one server now backs 32 worker proxies, where each CircleCI machine backs 4
* test: run every integration group in the trial on sized runners
* test: add a scan that refuses to upload integration results holding the license
* test: drop the member-name pass the archive byte scan already covers
* test: use a secret whose base64 and base64url encodings differ
* ci: run integration groups on run-ci pull requests and scheduled main runs
* ci: scope the integration license to a protected environment
* fix(otel): stop OTel v2 vendor presets from printing every span to stdout
With LITELLM_OTEL_V2 on and a vendor callback such as langfuse_otel or arize
holding operator credentials, the proxy printed every span as JSON to stdout,
prompts included, on the request path. Nobody set OTEL_EXPORTER, but the
console placeholder that OpenTelemetryV2Config folds in for an empty exporter
list stayed next to the vendor's own exporter.
The provider now skips that placeholder when the config has any other
exporter. A console exporter the operator names with OTEL_EXPORTER=console is
kept, and so is the placeholder when it is the only exporter.
* test(otel): type the console-placeholder tests and drop inherited OTEL_* env from the stdout check
* fix(otel): print the console default only for the otel callback
The proxy's background V2 logger (published when no callback built one, which is
what key-only and team-only Langfuse destinations run under) also kept the
unconfigured console exporter, so it printed every span with prompts to stdout.
The console default now prints only for a listed otel callback or an exporter the
operator named.
* fix(otel): keep the otel callback's console output on service-name routes
A key or team otel_service_name moves the call onto a provider the tenant
tracer cache builds, which never got the console choice, so the plain otel
callback stopped printing those calls. The no-operator-keys integration rig
also drops inherited Langfuse keys.
* feat(guardrails): scan sampled stream windows in the background (LIT-9338)
streaming_scan_in_background runs each sampled post_call scan as a task so
the stream keeps flowing; a block ends the stream before the next chunk.
* fix(guardrails): hold the stream end for the background verdict and cancel the scan on upstream errors (LIT-9338)
* fix(guardrails): wait for the cancelled background scan on upstream errors (LIT-9338)
* test(guardrails): cover /v1/responses, a failing scan peer and a mixed burst for background scans (LIT-9338)
* test(guardrails): build the burst calls without a nested comprehension (LIT-9338)
* fix(otel)!: keep a credential-less langfuse_otel on OTel v2 so key destinations get the Langfuse mapper
With LITELLM_OTEL_V2 on, langfuse_otel in callbacks and no operator
LANGFUSE_PUBLIC_KEY / LANGFUSE_SECRET_KEY, the preset raised at startup and
the callback fell back to the legacy logger. Key and team destinations were
then served by a generic v2 logger without the Langfuse mapper, so their
traces lost tags, user and cost, and the legacy logger added an orphaned
raw_gen_ai_request span.
The preset now always degrades to the credential-gated exporter when the
operator has no keys, the same way arize_preset does, so the startup logger
stays on v2, exports nowhere for traffic without a destination, and carries
the Langfuse mapper to every key and team destination.
* test(otel): type the langfuse tenant-only test helpers
* fix(otel): give key/team Langfuse destinations the Langfuse fields when langfuse_otel is not a callback
With langfuse_otel only in a key's or team's logging, the startup logger is the
generic one and its spans reached the key's Langfuse project without user,
session or tags. The emitter now adds the Langfuse mapper for a request with a
Langfuse destination, and every non-Langfuse exporter strips the Langfuse
request metadata, since any logger can now write it.
* fix(otel): stop a key's Langfuse logger from re-sending spans to the operator's collector
With langfuse_otel only in key or team logging, the proxy publishes the generic
V2 logger at startup and it already exports to the operator's OTLP collector.
The Langfuse logger a key's request builds kept the same base exporter, so the
collector got every model span twice. A logger built after the publish now keeps
only its preset's own exporters, as it already did beside a configured V2 logger,
and a later call that would rebuild one exporting nowhere reuses it.
* test(otel): assert the collector gets one copy instead of reading the key logger's exporters
* test(otel): count the collector's spans with one comprehension clause
* fix(otel): keep the legacy langfuse_otel logger and stand it down for requests a Langfuse destination carries
Reverts the preset degrade and the published-provider exporter trim, so every
request without a key or team Langfuse destination keeps main's legacy logger.
For a request with one, the V2 fan-out already delivers the whole trace with
the Langfuse fields, so the legacy logger emits nothing for it.
* test(otel): pin the legacy langfuse_otel paths a tenant's project still gets without a destination
* fix(otel): shed a destination's Langfuse request metadata before its other Langfuse fields
At its attribute limit a span sheds the Langfuse fields only a destination asked
for first, in the mapper's order, so the generation marker went before the
request metadata blob. Put the request metadata first, as a Langfuse-configured
logger already does.
* fix(otel): keep the legacy langfuse_otel span for a destination's call rejected before any provider
The OTel v2 fan-out records no call for a request the proxy rejects before a
provider (rate limit, pre-call hook), so standing the legacy logger down there
lost the user, session and tags main sent the tenant. Stand it down only for a
call the fan-out records, using the same check the v2 logger uses.
The tenant and collector integration tests now hold their exact counts for a
few seconds, so a late duplicate copy fails them.
* fix(otel): give a key's Langfuse destination the langfuse_environment the legacy logger wrote
The destination's resource carries langfuse.environment from the key's or team's langfuse_environment, else LANGFUSE_TRACING_ENVIRONMENT, so a key's project files its traces under the environment it chose instead of default. Only that destination's export carries it.
* fix(otel): keep the Langfuse fields written for a key's Langfuse destination off every other exporter
* fix(otel): give a key's Langfuse destination the prompt and response the legacy logger sent, only where it ran
* fix(otel): keep a key's Langfuse destination on the global capture setting when it sets none, with or without global Langfuse keys
* fix(guardrails)!: make lakera_v2 on_flagged block reject PII-only detections, add mask
on_flagged block now returns 400 when Lakera flags only PII, instead of
masking it and forwarding the request. on_flagged mask keeps the old
behavior: mask PII-only detections and block anything else.
* test(guardrails): cover lakera_v2 on_flagged on every endpoint, hook and a Lakera outage
* test(guardrails): type the lakera_v2 on_flagged unit test helpers
* feat(guardrails): opt-in response_masked_data on PANW Prisma AIRS block
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(guardrails): integration cells for PANW include_masked_response_on_block
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(guardrails): reject non-boolean include_masked_response_on_block
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* docs(guardrails): update PANW tool-call comment for include_masked_response_on_block
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(guardrails): build PANW kwargs without mutation and clean up DB-created test guardrails
Restore main's single-expression kwargs in the PANW initializer. The two DB-backed integration tests now create their guardrail inside the scenario and delete it on exit.
* fix(guardrails): type raw PANW boolean setting before validation
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(vertex_ai): use HTTP/2 httpx client for search_api vector store
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(httpx): exercise configured request path
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(httpx): verify HTTP/2 request negotiation
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(httpx): preserve HTTP/2 for sync proxy mounts
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* build(deps): add h2 to proxy extra so vertex search HTTP/2 is on by default
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(http_handler): isolate ssl context for http2 clients
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(http): add an HTTP/2 opt-out and harden the Vertex AI Search HTTP/2 client
litellm.http2 is now unset by default, so http2: false or LITELLM_HTTP2=False
turns the provider's HTTP/2 default off. A caller's own SSLContext keeps the
search on HTTP/1.1, the HTTP/2 async client honors DISABLE_AIOHTTP_TRUST_ENV,
forced IPv4 transports get the CA bundle and client cert, and vector store
search retries once when the server drops the connection with a GOAWAY.
* fix(http): retry vector store search only on HTTP/2 GOAWAY and keep aiohttp pool and proxy behavior
Scope the one retry to a connection Google closes with an HTTP/2 GOAWAY and
log it. When HTTP/2 replaces the aiohttp transport, keep aiohttp's idle
keepalive and connection cap, and only then skip env proxies under
DISABLE_AIOHTTP_TRUST_ENV, so the global LITELLM_HTTP2 toggle keeps HTTPS_PROXY
* test(http): cover the vector store search GOAWAY retry in unit tests and drop duplicate SSL cache comments
* fix(http): open another HTTP/2 connection once 100 streams are in use
httpcore keeps every request to an origin on one HTTP/2 connection and caps it at 100 concurrent streams, so request 101 and later waited with no timeout. The handlers now spread requests over extra connections, opening one only when every open connection has 100 streams in flight. The force_ipv4 wire tests now cover ssl_verify=False and the client certificate.
* fix(vertex_ai): reuse one Vertex search OAuth token across searches
Every vector store search built a new search config, so the token cache it carries was thrown away and each search fetched a new Google OAuth token with a blocking call on the event loop. Under a burst of searches that blocking fetch held back every HTTP/2 response. The provider config is now shared, so searches with the same credentials reuse one token until it expires.
* fix(http): keep the HTTP/2 connection spread within the connection cap
* fix(http): keep sync vertex search on HTTP/1.1, since httpcore's sync HTTP/2 is not thread-safe
httpcore 1.0.9 hands two threads the same HTTP/2 stream id, so threaded sync
searches failed with protocol errors. Only the async client, which the proxy
uses, now defaults to HTTP/2, and the sync connection spread is removed.
* fix(http): split the connection cap across HTTP/2 lanes
Each lane had its own pool with the full connection cap, so an origin answering
HTTP/1.1, or several origins on one client, could exceed max_connections once
more than 100 requests were in flight. Lanes now open only while lanes times
100 streams stay under the cap, and each lane pool gets at most 100 of it.
* test(http2): scope wire-test peer logs to each fixture instance
* fix(http2): keep TE, force_ipv4 trust, evicted-client and opt-out behavior on the HTTP/2 path
- drop TE values other than trailers on https requests through the HTTP/2 spread, since h2 refuses them
- force_ipv4 transports keep httpx's default trust store (SSL_CERT_DIR) unless SSL is configured
- the evicted client closer asks the HTTP/2 spread whether a request is in flight
- LITELLM_HTTP2 set to 0, no or off, and a quoted "false" http2 setting, opt out of HTTP/2
- the wire test's held-search barrier waits on an event instead of polling
* fix(http2): count proxy-mounted transports when checking an evicted client for in-flight requests
* test(http2): move the HTTP/2 eviction cases that open a loopback server into the integration wire tests
* fix(http2): read an unquoted http2: 0 as off and free the stream slot of a response dropped without closing
* fix(http2): free the stream slot of a partly read response dropped without closing
* test(http2): check that the evicted client and its stream close after a partly read response is dropped
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: mrinal <mrinal@berri.ai>
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(proxy): preserve reasoning fields on streaming cache hits
* fix(proxy): validate cached streaming delta fields
Signed-off-by: jesco-absolut <team@srswti.com>
* fix(caching): replay cached audio, images, reasoning and citations in stream order
A cached answer replayed as a stream now ends with its own empty finish chunk, so a chunk that
carries only audio or images keeps its payload and role. Reasoning fields lead on a chunk with no
text, and annotations arrive once on the last text slice, matching a live stream's order.
* fix(caching): replay a cached tool call after the text, like a live stream
Only the role, reasoning fields and provider fields lead a replayed answer. Tool calls now arrive on
their own chunk after the last text slice, and logprobs ride the first text chunk, so a cached
answer with reasoning and a tool call keeps a live stream's reasoning, text, tool call order.
* fix(caching): replay the reasoning before a cached tool call that has no text
A cached answer that thinks and then only calls a tool now replays its reasoning on its own chunk before the
tool call, the order a live stream sends, instead of one chunk carrying both.
* fix(caching): keep blank-only text when replaying a cached answer with reasoning
A cached answer that thinks, sends only whitespace as text, and then calls a tool now replays that whitespace
between the reasoning and the tool call, instead of dropping it.
---------
Signed-off-by: jesco-absolut <team@srswti.com>
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* feat(proxy): replace the pre-migration row check with a rolled-back dry run of the pending migrations
* feat(proxy): roll back each migration of the dry run on its own and baseline-check databases without a ledger
* test(proxy): cover the dry run CLI and PrismaManager wrapper without a database
* fix(proxy): refuse the no-ledger dry run for a custom migration folder like boot does, and probe the rolled-back indexes directly
* ci(codecov): stop carrying forward the roi-postgres and redis-compat flags
Their carried-forward sessions were measured against older proxy_cli.py revisions (633 and 689 line maps against 710 today) and mark docstring lines of the merged file as missed, which failed codecov/patch on this PR at 70% while every unit shard reports the lines hit
---------
Co-authored-by: devin-ai-integration[bot] <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix: drain shutdown callbacks and persist pending tag spend
* fix: catch shutdown timeouts on Python 3.10
* fix: share shutdown callback budget with HTTP draining
* fix: type shutdown cancellation tasks and document its clock
* test: retain callbacks after bounded worker cancellation
* test(proxy): inspect the lifespan body after the SIGTERM wrapper split
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(utils): keep the success callback enqueue on its own task
The synchronous enqueue spawned the logging worker loop inside the awaited
call and ran callbacks that per-call loops used to drop, which hung the SDK
callback tests on the autouse worker flush. The shutdown worker flush covers
the queued callback on its own, so the one-tick task is restored
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(logging_worker): keep the queue's unfinished count exact when a full queue rejects a callback
The aggressive clear ran the rejected callback and marked it done against
the queue it was never put in, so flush() could return while a real
callback was still running and the shutdown drain cancelled it early. The
rejected callback is now re-enqueued once the clear made room, and only
runs uncounted when the queue is still full
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(logging_worker): build the aggressive-clear run set immutably
Final tuple instead of an appended list, with the re-enqueue in a
_requeue helper, and Final annotations on the overflow regression test
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(logging): track overflow callbacks through shutdown
* chore(logging): explain shared callback ownership
* test(logging): isolate deferred callback completion
---------
Co-authored-by: Joshua Valluru <326636767+joshua-berri@users.noreply.github.com>
Co-authored-by: joshua <joshua@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(guardrails): scan and rewrite streamed /v1/completions in post_call pipelines
Gives OpenAITextCompletionHandler a process_output_streaming_response that
assembles the buffered TextCompletionResponse chunks per choice index at end
of stream, runs the guardrail over the assembled text, and writes text
rewrites back into the first text-carrying chunk of each choice. Tests and
live proof still pending.
* fix(guardrails): scan and rewrite streamed /v1/completions under post_call guardrails
Assemble each choice's streamed text once the stream ends, run the guardrail on it, write a mask back into the buffered chunks, and end the stream with the guardrail's error on a block. User key metadata now goes through the shared metadata bucket so applied guardrails reach spend logs.
* test(guardrails): cover streamed /v1/completions under post_call guardrails
* fix(guardrails): keep Gray Swan key metadata, scan a cut-short text stream, and pass non-chunk frames through
* fix(guardrails): replace blocked text completion streams like chat, keep Gray Swan's vendor payload, scan text streams in the content filter
* fix(guardrails): release an unfinished choice's held text and omit usage the upstream never sent
The content filter's streaming hook held back the end of a choice until its
finish_reason, which litellm sends only for the first choice of an n>1
stream, so choice 1 lost its last characters on text and chat streams. A
closing chunk now releases every unfinished choice when the stream ends.
A blocked text stream's closing chunk reported 0/0 usage when the upstream
sent none; it now carries usage only when the upstream reported it.
* fix(guardrails): send no usage on the content filter's end-of-stream chunk
* fix(guardrails): keep the usage chunk last and log the masked text of streamed text completions
* fix(guardrails): stream Responses API events through the content filter as they arrive
* fix(guardrails): treat a null text completion choice index as choice zero when assembling streams
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
