* fix(ptu): hand the prune a plain delete filter the query builder can serialise
The bounded sweep built its predicate as a read-only mapping view, which the query
builder refuses to serialise, so the nightly job raised as soon as a config-declared
deployment was priced. The charges were already written by then, which is why the run
looked like it had produced its rows.
The in-memory table these tests run against accepts any mapping, so only a live run
caught it. A predicate builder now returns a plain dict and is asserted as one, and the
catch-up pass has a test covering a config-declared reservation.
* refactor(ptu): build the prune predicate in one shot
Both filter shapes are known upfront, so the bounded one is constructed
directly rather than by mutating a value already declared Final.
The catch-up test took two independent clock reads, which disagree across
UTC midnight; it now derives both the reservation start and the expected
last charged day from a single read, matching the three sibling tests.
* feat(ptu): accrue flat cost for PTU deployments declared in config.yaml
The flat-cost rollup reads deployments from LiteLLM_ProxyModelTable, and config.yaml
models never reach that table by design, so a PTU deployment declared there accrued no
flat cost at all while still billing its traffic per token. The provider bills the
reservation whichever file declared it.
The rollup now also reads the deployments the router holds that no database row owns,
identified by db_model, skipping the per-request credential clones that carry
original_model_id and reuse their source's PTU config under a fresh id. Registering such
a deployment zeroes its pricing, since reserved capacity already pays for the traffic it
serves, and leaving a rate unset falls back to the public cost map, which makes the double
charge the default rather than an opt-in.
The rules both halves apply now live in one module. The rollup's test for what it will
charge and the router's test for what to zero have to agree, or a deployment one accepts
and the other declines serves its traffic for free. That module also owns the fields the
write endpoints already zero, so the two paths cannot drift: tiered_pricing is emptied
rather than zeroed because its tiers outrank the rates beside them, the search context
table is written zeroed because an absent one means the provider default, and any further
rate the deployment itself declares is zeroed alongside the standing set.
The prune is bounded to the deployments a run scanned, but only for a run that priced a
config-declared deployment. Deciding a row is garbage on staleness alone stays correct
while every run derives its charges from the same table, so a database-only run sweeps
exactly as it did before; once one host's charges come from a file the others cannot read,
a row it never considered is not evidence of anything.
Behaviour change worth calling out: a zeroed deployment sorts ahead of an unpriced sibling
in QualityRouter's cost tiebreak, where an unset rate previously sorted last. Reserved
capacity really is the cheaper choice, but the ordering moves.
* refactor(ptu): drop a Final rebind and two redundant isinstance guards
The basedpyright budget rejected reassigning a Final in the datetime coercion and
two isinstance calls the router entry's own type already guarantees. Filtering the
built records rather than the raw entries removes both guards and leaves
_router_deployment as the single validator.
The field replaces the standard RPM/TPM checks for batch submissions, so a
key holder or team admin writing it could pick their own batch quota.
Mirrors the output-token-estimate admin gate: change-based, so resending
the stored value stays allowed, and enforced on key generate, update, bulk
team-key update, regenerate, and team new/update.
`ProxyConfig.get_config()` walked the parsed config and replaced every
`os.environ/<KEY>` string with `get_secret(value)` before anything initialized
the secret manager, so a key held only by the manager resolved to `None` and
that `None` was written back into the config. The later fallback in
`load_config` could not recover it, because the key now existed with a `None`
value.
Hoist the initialization into `get_config()`, ahead of the resolution pass, so
every entrypoint gets it: the CLI already did this itself, but the microservice
entrypoints (`gateway/main.py`, `backend/main.py`) uvicorn the app directly and
bypass the CLI. `load_config`'s own call is now redundant and is dropped, so
startup builds the manager once instead of building one and discarding it.
`get_config()` also runs on management-endpoint request paths, so this returns
early once a manager exists rather than rebuilding the client per request.
Also warn when a reference the manager would have been asked for resolves to
`None`. The reporter had no log line at all to work from. `get_secret` only
reaches the manager when reads are enabled and the name is in `hosted_keys`, so
`secret_manager_would_be_consulted` mirrors that gate and keeps the warning off
env-only references, which are expected rather than an error.
A Redis over-limit verdict now survives a failing rollback DECRBY instead of
escaping into the in-memory fallback and granting tokens the counter already
rejected; the unrolled increments expire with the TTL. pop_reservation now
falls through to the local record when the Redis pop succeeds but finds
nothing, so a reservation saved in memory after a transient Redis save
failure still refunds on cancel or completion.
In-memory grants now record an owner token, and a refund only debits local
counters when the popping worker is the one that granted them, so a terminal
response handled elsewhere can no longer shrink another worker's unrelated
fallback reservations. A Redis-granted refund that fails no longer falls back
to decrementing local counters either: the leaked Redis increments expire
with the TTL and only tighten the allowance.
Auth and pass-through failures reach post_call_failure_hook with the raw request body unstripped, so a client-supplied standard_logging_object could feed the new attribution fallback when the logging object carries none. Pop the key before the lift so only the logging object may supply it. Also coalesce a None recovered cost to 0.0 so the lift always overwrites any client-supplied response_cost, matching the merge base's clobber semantics.
Reserve-script failures now roll back the scopes already incremented before
re-raising into the in-memory fallback, so a partial redis outage no longer
leaks counter increments that shrink the shared allowance. Reservations
record which backend granted them, so a refund never debits redis counters
an in-memory grant did not charge. Terminal-status matching is now
case-insensitive because the Bedrock async-invoke retrieve path returns raw
AWS-cased statuses like Completed.
* fix(auth): resolve bare model names against wildcard deployments in model access groups
* test(e2e): cover model access group permission checks on keys and teams
`lite up` already patches ~/.claude/settings.json, but only for as long as it
runs in the foreground, and it restores the original file on exit. Users
proxying Claude Code through LiteLLM therefore have to re-wire it by hand after
every login.
--config-claude makes that write persistent. It reuses the settings shape
`lite up` writes (env.ANTHROPIC_BASE_URL plus an apiKeyHelper invocation),
preserves every unrelated key, creates the file when missing, and writes it
atomically with owner-only permissions. Plain `lite login` is unchanged.
Reaching the credential through apiKeyHelper rather than copying it into the
file means a later login refreshes it with no further action, and keeps the
short-lived CLI token out of settings.json entirely.
The shared parts of the settings-file handling move from up.py into a new
claude_settings.py, since up.py imports auth.py and so auth.py cannot import
up.py back. That module now also owns the registry of commands that can be
temporarily managing the file, so the persistent write refuses while either
`lite up` or `lite autoroute up` holds a backup it would later restore over
this write.
Because this write has no backup and no `lite down`, it is stricter than
`lite up` about the user's file: it writes through a symlinked settings.json
rather than replacing the link with a regular file, and it refuses rather than
silently discarding a non-object `env` value.
Also fixes the apiKeyHelper command itself: --base-url belongs to the
top-level `lite` group, so `lite auth print-token --base-url X` is rejected by
click with "No such option". Every settings file `lite up` has written carries
that malformed command, which makes the helper return nothing and every Claude
Code request lose its token. The existing tests only string-matched the
generated command, so the new tests parse it through the real CLI instead.
An OpenAPI-backed MCP tool whose upstream answered 401 came back as a
successful tool result carrying the upstream's rejection as its content, so a
caller saw {"error":"invalid_token"} presented as data and the gateway recorded
the request in its own spend log as call_mcp_tool | success.
Three layers each erased the outcome. The request function returned
response.text whatever the status, _handle_local_mcp_tool caught every exception
and returned it as ordinary TextContent, and both dispatch sites then stamped
isError=False unconditionally. Fixing only the first, which is the obvious fix,
changes nothing, because the two above it still map failure onto the
success-shaped value.
The status is now classified where the response is held: a 401 becomes
MCPUpstreamAuthError so the caller is told to re-authenticate, and every other
non-2xx becomes MCPOpenApiUpstreamError, which carries the status and drops the
upstream body rather than serving it as tool content. _handle_local_mcp_tool no
longer swallows, and the call_tool arm keeps the auth error's type. Nothing new
renders these: call_mcp_tool and call_tool_rest_api already turn them into an
isError result naming the status and into a real 401 with WWW-Authenticate, and
the OpenAPI path simply never reached them.
The result is now byte-identical to the regular MCP path for the same failure.
* feat(auto-router): scope shadow eval jobs to multiple keys
A shadow eval job now covers a set of keys instead of exactly one, and each
key carries its own max_turns budget, so one key exhausting its budget leaves
its siblings sampling. The existing job row already is the per-key unit
(api_key_id, max_turns, stopped_at, and the one-active-per-key-and-direction
partial unique index all live on it), so multi-key is grouping rather than
schema surgery: a new group_id column ties N sibling rows written atomically
by one create_many, the API's job id becomes the group id, and pre-existing
jobs backfill group_id = id so their ids keep resolving. The sampler hot path
is untouched; its test file has a zero-line diff
Results come back pooled plus a per-key breakdown and responses list every key
with its own budget, stop state and read-time labels. The dashboard is adapted
minimally to the new shapes (the picker stays single-key and submits a one-key
list); the multi-select picker and per-key table land in the stacked UI PR
* fix(shadow_eval): derive completed from spent budgets and record operator stops
* fix(shadow_eval): stamp stops atomically and freeze counts at the stamp
The stop endpoint wrote stopped_by and stopped_at as two separate updates, so
a failure between them left a job reading stopped while its unstamped legs
kept sampling, and the retry got 400 already stopped. One UPDATE now stamps
stopped_by and every missing stopped_at together, preserving the stopped_at a
leg earned from its own budget via COALESCE
Attempt counts now exclude attempts that land after a leg's stopped_at, so an
in-flight attempt finishing just after an operator stop can never push a
legacy pre-stopped_by job over its budget and flip it from stopped to
completed at read time
* fix(shadow_eval): backfill stopped_by so legacy stops never read as completions
* chore(ui): regenerate api types for the shadow eval stop fields
* fix(shadow_eval): let the stop statement pick one winner under racing stops
Two operators can both pass the derived-status guard in the race window. The
stop UPDATE now claims only legs with stopped_by still null and the endpoint
judges by its row count, so exactly one caller ever gets the 200 and the loser
gets the same already-stopped 400 a late caller gets
* refactor(shadow_eval): make the stop statement the whole state machine
The status guard ran before the UPDATE, so a stop racing the last budgeted
attempt still claimed the job and it read stopped forever instead of
completed. The statement now claims the job only while a leg still samples
inside the window with no stop recorded, and the endpoint reads once after
writing: a racing operator, a same-instant budget spend, and a repeat stop all
get the 400 naming the status the job actually holds. The pre-write guard and
the hand-built response go away
* chore(ui): regenerate api types for the stop route description
The flat-cost rollup reads deployments only from LiteLLM_ProxyModelTable, so a PTU
deployment declared in config.yaml never accrues flat cost. Those deployments live in
llm_router.model_list as plain dicts whose id sits in model_info rather than on the entry,
so they do not satisfy the shape _parse_ptu_model reads.
Adds a frozen record in that shape and a factory that maps a router entry onto it, leaving
_parse_ptu_model byte-identical so the existing cases stand as evidence of no behaviour
change. Nothing calls the factory yet; the caller lands with the loader union.
_decode_model_info also stops handing back valid JSON that is not an object. It decoded
a list or a scalar and returned it as a mapping, so the caller read fields off it and
raised, losing the whole run rather than the one bad deployment.
Both OpenAPI dispatch arms sourced the upstream credential only from the
deprecated global / BYOK mcp_auth_header and never from mcp_server_auth_headers,
so x-mcp-{alias}-authorization was silently dropped on spec_path servers and the
upstream API received no Authorization at all. The managed path already resolves
it through lookup_mcp_server_auth_in_headers, so the two had drifted.
_resolve_openapi_tool_auth now owns that resolution for both arms. A per-server
value is already a complete header value and is forwarded verbatim, while a BYOK
credential keeps its auth-type prefix, so the two are never conflated into
"Bearer Bearer <token>". The resolved credential is also handed to
resolve_openapi_upstream_auth, whose passthrough arm reads it through
_passthrough_token_from_mcp_auth_header and outranks the ContextVar.
server.py loses its inlined copy of the forwarded-header logic along with its
mcp_server is None guards, which are unreachable after the 503 raised above them.
Credit to the earlier analysis and approach in #33349, which this supersedes
against the current v2 credential resolver.
The model resync now mutates the router under MODEL_RECONCILE_LOCK, and the
agent resync shares the new AGENT_RECONCILE_LOCK with the periodic agent
reload, so a reconcile built from a pre-write DB snapshot can no longer evict
or duplicate what a read-through just registered. Every resync checks
should_load_db_object for its object type, keeping read-through consistent
with what the replica is configured to load, and the a2a raise sites tag
ProxyModelNotFoundError as non-retryable so an agent miss no longer burns the
model resync budget.
Resync registry misses with single-row DB fetches (guardrail by unique
name, agent by unique id or name, model by name then id) instead of
full-table loads, and bound them with a global budget of 20 resyncs per
5s window per registry that fails closed without negative-caching the
key.
Access group create/update now trust the reconcile outcome snapshot
captured under the reload lock instead of a post-lock router read, so a
concurrent reconcile can no longer surface a false degraded-serving 500.
Router.upsert_deployment restores the previously served deployment when
the replacement add fails under ignore_invalid_deployments, so a bad
update no longer silently drops a healthy deployment from serving.
