Greptile review on #26756 (P2): if `attempt_db_reconnect` itself raises
(e.g. lock cancellation, timer error, unexpected internal failure), the
original `httpx.ReadError` / transport error was lost — `failure_handler`
and `db_exceptions` alerts then logged the reconnect exception instead of
the actual DB transport problem, masking the root cause.
Wrap the reconnect call in a try/except. On reconnect failure, re-raise
the *original* `first_exc` and chain the reconnect error as `__cause__`
so it remains visible for debuggability without becoming the primary
exception observers see.
Adds `test_call_with_db_reconnect_retry_preserves_original_error_when_reconnect_raises`
asserting (a) the propagated exception is the original transport error
and (b) the reconnect exception is attached as `__cause__`.
Two related fixes layered on top of the existing reconnect plumbing:
1. Restore reconnect-and-retry on `PrismaClient.get_generic_data` (issue
#25143). 1.83.x lost the transport-reconnect-and-retry-once branch that
1.82.6 had on this method, so transient `httpx.ReadError` flaps now
surface immediately as `db_exceptions` alerts. `_update_config_from_db`
fans out four concurrent `get_generic_data` reads, so a single transport
blip used to mark four alerts and a stale config window.
Adds `call_with_db_reconnect_retry` to `litellm/proxy/db/exception_handler.py`
— a single canonical "try DB read, on transport error reconnect once and
retry once" wrapper. Mirrors the inline pattern in
`auth_checks._fetch_key_object_from_db_with_reconnect` so we have one
implementation rather than three drifting copies, and gives future read
paths a clean opt-in.
2. Fix the `_engine_confirmed_dead` flag-reset bug in
`_run_reconnect_cycle`. The flag was cleared before `_do_heavy_reconnect()`
ran, so any failure inside the heavy reconnect (timeout, missing
DATABASE_URL, recreate failure) left the flag False — and the next
attempt could silently demote to the lightweight path even though the
engine was genuinely dead. Move the reset into the success branch so the
flag stays True across heavy-reconnect failures and the next attempt
re-enters the heavy branch.
Tests:
- `tests/test_litellm/proxy/db/test_exception_handler_reconnect_retry.py`
(new) — 9 tests covering the helper's contract: happy path, retry on
transport error, no retry on data-layer errors, propagation when reconnect
fails, propagation after second transport error, `hasattr` guard for
partial mocks, fresh-coroutine-per-call invariant, explicit timeout
override, default timeouts read off the prisma_client.
- `tests/test_litellm/proxy/db/test_prisma_self_heal.py` — adds:
- `test_get_generic_data_retries_on_transport_error_for_config_table`
- `test_get_generic_data_propagates_when_reconnect_fails`
- `test_engine_confirmed_dead_persists_across_failed_heavy_reconnect`
(regression test for the flag-reset bug).
All 16 self-heal tests + 9 helper tests + 535 auth/exception-handler tests
pass locally.
- server.py: drop the redundant server_id append in
_get_filtered_mcp_servers_from_mcp_server_names. iter_known_server_prefixes
already yields server_id unconditionally, so the manual append (and its
misleading comment) was a no-op duplicate.
- utils.py: rewrite the SHORT_MCP_TOOL_PREFIX docstring to accurately
describe the collision behaviour. The previous wording said collisions
were 'cosmetic only', but a natural-hash collision IS a routing-correctness
issue, which is precisely why we already added _assign_unique_short_prefix
to rehash deterministically. The new comment cross-references that path.
- utils.py: restrict the first character of the short prefix to [A-Za-z]
via a 52-char alphabet for position 0 only. The remaining two positions
still use the full base62 alphabet. This keeps prefixes valid identifiers
on every backend and gives 52*62*62 = 199_888 distinct prefixes (still
comfortably more than any realistic deployment).
- tests: add coverage proving the first character of the prefix is always
alphabetic across many server_ids and rehash attempts.
Co-authored-by: Mateo Wang <mateo-berri@users.noreply.github.com>
The test was flaking on unrelated asyncio ERROR records (e.g. "Unclosed
client session" from background tasks in other tests). Restrict the
assertion to records emitted by LiteLLM loggers so the test only fails
on errors actually produced by the code under test.
Co-authored-by: Mateo Wang <mateo-berri@users.noreply.github.com>
Two MCP servers can natural-hash to the same three-character base62
prefix. With 62**3 = 238_328 slots the birthday bound is ~488 servers
for 50% collision probability, so a single proxy hosting more than
~100 MCP servers has a non-trivial chance of seeing a collision in
practice — and a collision means tool names from two different servers
share a routing key, causing silent mis-routing.
Mitigation:
- compute_short_server_prefix(server_id, attempt=N) folds an attempt
counter into the SHA-256 seed, so rehashes are deterministic and
produce a fresh three-char prefix space per attempt.
- New MCPServer.short_prefix field caches the resolved (post-dedup)
prefix on the model so it stays stable across the process lifetime.
- MCPServerManager._assign_unique_short_prefix walks attempts 0..N
until it finds a prefix not already used by another server in the
combined registry. Logs an INFO line when a rehash happens so
operators have a breadcrumb if it ever does.
- Wired into every registration path: load_servers_from_config,
add_server, update_server, reload_servers_from_database. The
database reload path also carries the previously-resolved prefix
forward so reloads don't churn it.
- get_server_prefix prefers the cached short_prefix when set, so the
resolved value (not the raw natural hash) is used everywhere.
- iter_known_server_prefixes yields the cached short_prefix too, so
reverse-lookup tolerance covers the rehashed form.
No-op when LITELLM_USE_SHORT_MCP_TOOL_PREFIX is disabled — the field
stays None and behaviour is unchanged.
Co-authored-by: Mateo Wang <mateo-berri@users.noreply.github.com>
Adds LITELLM_USE_SHORT_MCP_TOOL_PREFIX. When enabled, tool / prompt /
resource / resource-template names emitted from MCP servers are prefixed
with a deterministic three-character base62 ID derived from the server's
server_id (SHA-256 → base62) instead of the (potentially long)
alias / server_name. This keeps namespaced tool names well under the
60-character upper bound enforced by some model APIs while still letting
us distinguish MCP-routed tools from local tools.
Behavioural notes:
- Default off — when the env var is unset, the long-prefix behaviour
is unchanged. The plan is to flip the default in a future release
and remove the gate after a deprecation window.
- Prefix derivation is deterministic, so it is stable across processes,
workers and restarts without any persistence layer.
- Reverse-lookup is tolerant: _create_prefixed_tools registers every
known prefix form (alias / server_name / server_id / short ID) in
the routing map and _get_mcp_server_from_tool_name resolves any of
them. Old clients holding cached long-prefixed names continue to
route correctly even after the flag is enabled.
- _get_allowed_mcp_servers_from_mcp_server_names accepts the short
prefix in /mcp/{server_name}-style URLs.
- The OpenAPI tool-listing path now filters by the active server
prefix instead of server.name so spec-backed servers benefit too.
Co-authored-by: Mateo Wang <mateo-berri@users.noreply.github.com>
* Unify cost calc in success_handler dict and typed branches
* Trim verbose comments and docstrings
---------
Co-authored-by: Michael Riad Zaky <michaelr@Mac.localdomain>
Co-authored-by: Michael Riad Zaky <michaelr@Michaels-MacBook-Air.local>
Claude 3.5 Sonnet v2 reached EOL on Bedrock 2026-03-01, returning the same
404 EOL error as 3.7 Sonnet. Sonnet 4.5 supports both InvokeModel and
Converse APIs on Bedrock, so use the same model for both routes.
AWS Bedrock has reached end-of-life for `claude-3-7-sonnet-20250219-v1:0`,
returning 404s with "This model version has reached the end of its life."
Update test references to `claude-sonnet-4-5-20250929-v1:0` (same capability
surface: thinking, tools, prompt caching, PDF input, vision, computer use).
The bedrock/invoke pass-through tests stay on Sonnet 3.5 since Sonnet 4.5
is converse-only on Bedrock.
Companion to the prior commit. process_items only converted empty
`items: {}` to `{"type": "object"}`. But anyOf branches like
`{"type": "array"}` (no items field at all) were untouched, so after
convert_anyof_null_to_nullable stripped the null branch and added
nullable, the array branch was sent to Vertex as
`{"type": "array", "nullable": true}` — which Vertex rejects with
INVALID_ARGUMENT (`any_of[0].items: missing field`).
Make process_items synthesize `items: {"type": "object"}` for any
`type == "array"` schema where items is missing or empty.
Also:
- Convert test_gemini_tool_calling_working_demo to a hermetic mock
test asserting items is present on the array branch in the sent
body. Was previously a real-network call to Vertex and was the
test the user reported still failing in CI.
- Add unit test test_build_vertex_schema_array_branch_missing_items_in_anyof
covering the missing-items shape directly.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Cache provider config lookups for Vertex Anthropic messages so repeated requests reuse the same config object and preserve credential cache state. Add a regression test to catch any future loss of config reuse.
Made-with: Cursor
* fix(caching): preserve prompt_tokens_details through embedding cache round-trip
The embedding caching layer was dropping prompt_tokens_details (including
image_count) because CachedEmbedding had no field for usage metadata and
the cache retrieval code reconstructed Usage without it. This caused
inconsistent responses where the first call returned image_count but
cached responses did not, breaking cost tracking for multimodal embeddings.
Add prompt_tokens_details to CachedEmbedding, persist per-item details
during cache storage, aggregate them on retrieval, and merge them in
combine_usage() for partial cache hits.
* style: apply Black formatting to caching files
* fix(caching): address Greptile review — cyclic import, guarded construction, nested dict merge
Move PromptTokensDetailsWrapper to inline import to resolve CodeQL cyclic
import warning. Guard PromptTokensDetailsWrapper construction with
try/except to handle unexpected cached keys. Add recursive dict merging
in _merge_prompt_tokens_details for nested fields like
cache_creation_token_details.
Route vector store search `extra_body` into provider transformers and handle Bedrock `retrievalConfiguration` explicitly so only intended provider-specific fields are forwarded.
Made-with: Cursor
convert_anyof_null_to_nullable was stripping the items field from array
branches inside anyOf when a sibling null branch was present, leaving
{"type": "array"} without items. Vertex requires items whenever
type == "array" (even inside anyOf) and rejects the call with
INVALID_ARGUMENT.
Leave the (possibly empty) items in place so the downstream process_items
step can convert {} to {"type": "object"}, which is what Vertex wants.
Also:
- Update test_build_vertex_schema expected output, which was codifying
the broken shape.
- Convert test_gemini_tool_calling_not_working to a hermetic mock test
that asserts the request body sent to Vertex includes items inside
the callbacks anyOf array branch. The previous form made a real
network call and was flaky in CI.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds a CLI flag (`--timeout_worker_healthcheck`, env `TIMEOUT_WORKER_HEALTHCHECK`)
that forwards to uvicorn's `timeout_worker_healthcheck` Config kwarg (added in
uvicorn 0.37.0). Lets operators raise the supervisor's worker-ping timeout above
the default 5s when triaging workers being killed and respawned under load.
The helper introspects `uvicorn.Config.__init__` and only sets the kwarg if
supported, otherwise prints a warning - so the existing uvicorn>=0.32.1,<1.0.0
floor pin is unaffected. Gunicorn and Hypercorn paths are unchanged (the uvicorn
supervisor isn't running there); the value is also not passed to the helper at
all on those paths so the "uvicorn too old" warning never fires spuriously.
* Use auth key name if there are no app id in in headers or in extra_data
* use key alias instead of key name
* Fix
* last priority key alias
* Fix
* Add tests
* [Feat] Day-0 support for GPT-5.5 and GPT-5.5 Pro (#26449)
* feat(openai): day-0 support for GPT-5.5 and GPT-5.5 Pro
Add pricing + capability entries for the new GPT-5.5 family launched by
OpenAI on 2026-04-24:
- gpt-5.5 / gpt-5.5-2026-04-23 (chat): $5/$30/$0.50 per 1M
input/output/cached input
- gpt-5.5-pro / gpt-5.5-pro-2026-04-23 (responses-only): $60/$360/$6
per 1M input/output/cached input
Other fees (long-context >272k, flex, batches, priority, cache
discounts) follow the same ratios as GPT-5.4, with context window
retained at 1.05M input / 128K output.
No transformation / classifier code changes are required:
OpenAIGPT5Config.is_model_gpt_5_4_plus_model() already matches 5.5+ via
numeric version parsing, and model registration is driven from the
JSON. The existing responses-API bridge for tools + reasoning_effort
(litellm/main.py:970) already covers gpt-5.5-pro.
Tests:
- GPT5_MODELS regression list now covers gpt-5.5-pro and dated variants
- New test_generic_cost_per_token_gpt55_pro cost-calc test
- Updated test_generic_cost_per_token_gpt55 for long-context fields
* fix(openai): mirror reasoning_effort flags onto gpt-5.5 dated variants
gpt-5.5-2026-04-23 and gpt-5.5-pro-2026-04-23 were missing the
supports_none_reasoning_effort, supports_xhigh_reasoning_effort, and
supports_minimal_reasoning_effort flags that their non-dated
counterparts define. Reasoning-effort routing in OpenAIGPT5Config is
fully capability-driven from these JSON flags — since an absent flag
is treated as False for opt-in levels (xhigh), users pinning to a
dated snapshot would silently lose xhigh support and diverge from the
base alias on logprobs + flexible temperature handling.
Copy the flags onto both dated variants so every dated snapshot
inherits the base model's reasoning-effort capability profile.
Adds a parametrized regression test that asserts
supports_{none,minimal,xhigh}_reasoning_effort parity between each
dated variant and its non-dated counterpart, preventing future drift
when new snapshots are added.
* [Feat] Add azure/gpt-5.5 + azure/gpt-5.5-pro entries (+ dated variants) (#26361)
* feat(azure): add azure/gpt-5.5 + azure/gpt-5.5-pro entries (+ dated variants)
Azure variants of OpenAI's GPT-5.5 family. Microsoft has not yet
shipped GPT-5.5 on Azure OpenAI (latest GA on the Foundry models page
is GPT-5.4 as of 2026-04-24), but adding the entries day-0 mirrors the
established precedent for azure/gpt-5.4* (which were in the cost map
before the Azure rollout) so cost tracking and capability flags work
the moment customers deploy.
Schema follows the existing azure/gpt-5.4* shape:
- Same base/long-context pricing as openai/gpt-5.5*: $5/$30 chat,
$60/$360 pro per 1M, with priority tier 2x base
- Azure variants drop the flex/batches keys (Azure has no flex tier)
but keep priority pricing, matching gpt-5.4* precedent
- mode=chat for the thinking model, mode=responses for pro
reasoning_effort capability flags mirror the OpenAI variants exactly
since Azure proxies the same API contract: minimal rejection on both
chat and pro, low/none rejection on pro. Once #26456 (which sets
supports_low_reasoning_effort + minimal=false on openai/gpt-5.5*)
lands, OpenAI and Azure flag profiles align.
Tests pin entry presence + pricing for all four Azure variants and
verify the live-API-derived reasoning_effort flags.
* test: register supports_low_reasoning_effort in cost-map JSON schema
azure/gpt-5.5-pro and azure/gpt-5.5-pro-2026-04-23 added in this branch
carry supports_low_reasoning_effort=false. The strict
'additionalProperties: false' schema in
test_aaamodel_prices_and_context_window_json_is_valid rejected the new
key. Register it alongside the other supports_*_reasoning_effort
entries.
Note: the runtime side of this flag (code that reads it) lands in
#26456. Until that PR merges the flag is inert for both Azure and
OpenAI pro entries, but having the schema accept it lets cost-map
tests pass on either merge order.
* Use sanitize deep copy style to replace deepcopy usage
* Added test checking error is not happening anymore
* Added warning log when json copy failed
* Reduce to one change
* Fix spaces
---------
Co-authored-by: Ido Lavi <ido@noma.security>
Co-authored-by: yuneng-jiang <yuneng@berri.ai>
Co-authored-by: Mateo Wang <277851410+mateo-berri@users.noreply.github.com>
Co-authored-by: TomAlon <tom@noma.security>
tool_calls on assistant messages were translated to OllamaToolCall format
but never copied into the outgoing OllamaChatCompletionMessage, so Ollama
received {role: assistant, content: ''} with no tool_calls. The model
then had no record of having made a tool call, causing it to re-issue
the identical call on every turn (infinite loop).
Similarly, tool_call_id on role:tool messages was silently dropped.
Ollama uses this field to resolve the tool name from conversation history.
Also add tool_call_id to OllamaChatCompletionMessage TypedDict.
Fixes#26094
* fix(proxy): invoke post-call guardrails on pass-through endpoint responses (#20270)
Wire post_call_success_hook into non-streaming pass-through response path,
gated on explicit guardrail config (opt-in only, no backwards-compat break).
- Call post_call_success_hook after reading non-streaming response body
- Build enriched hook_data with guardrails metadata and litellm_logging_obj
at call site (avoids mutation of _parsed_body which is shared by logging)
- Handle ModifyResponseException with provider-agnostic error envelope,
post_call_failure_hook, and defensive try/except
- Strip stale content-length when guardrail modifies response body
- Move ModifyResponseException to litellm.exceptions to break cyclic import;
re-export from custom_guardrail for backwards compat
- Add call_type fallback in UnifiedLLMGuardrails for pass-through endpoints
using CallTypes.pass_through.value enum
* test: add unit tests for pass-through post-call guardrails
5 tests covering the post-call guardrail invocation on pass-through endpoints:
- post_call_success_hook fires when guardrails configured
- post_call_success_hook skipped when no guardrails (backwards compat)
- ModifyResponseException returns 200 with provider-agnostic error
- UnifiedLLMGuardrails resolves call_type from logging_obj for pass-through
- ModifyResponseException re-export from custom_guardrail stays in sync
Bedrock enforces non-increasing TTL ordering across cache_control blocks
(tools → system → messages). The tool cache_control TTL was being
unconditionally dropped to the default 5m, while system blocks preserved
the user-specified TTL for Claude 4.5+ models. This mismatch caused
"a ttl='1h' block must not come after a ttl='5m' block" errors when
users set ttl='1h' on both tools and system.
Converse path: add_cache_point_tool_block() now accepts a model param
and preserves TTL for Claude 4.5+, matching _get_cache_point_block().
Invoke path: _remove_ttl_from_cache_control() now also processes tools
(was only processing system and messages).
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(redis): cache GCP IAM token to prevent async event loop blocking
## Problem
GCPIAMCredentialProvider.get_credentials() calls _generate_gcp_iam_access_token
on every Redis connection establishment. This function performs synchronous HTTP
and gRPC calls (google-auth + google-cloud-iam) which block Python's asyncio
event loop while running.
Under concurrent load (e.g. connection pool warm-up, parallel health checks),
multiple connections are established simultaneously, each triggering an
independent blocking IAM token refresh. These refreshes serialise behind each
other inside the single-threaded event loop, causing individual Redis spans to
take 20-25 seconds instead of milliseconds.
Observed in production via Datadog APM: a single INCRBYFLOAT Redis span took
25.6 seconds (90% of a 28.4s trace), with GCP metadata + GenerateAccessToken
gRPC calls visible inside the span. This cascaded into aiohttp SocketTimeoutError
on upstream LLM API calls — not because the upstream was slow, but because the
event loop was frozen and the 30-second sock_read timer fired on a connection
that was never given CPU time.
## Fix
Add a module-level token cache (dict keyed by service account, value is
(token, expiry_monotonic)). _get_cached_gcp_iam_token() returns the cached
token on cache hit (no I/O), and refreshes only when expired using
double-checked locking so only one thread performs the network round-trip.
GCP IAM tokens are valid for 1 hour; the cache TTL is set to 55 minutes
(_GCP_IAM_TOKEN_TTL_SECONDS = 3300) to refresh safely before expiry.
The cache is shared across all GCPIAMCredentialProvider instances for the same
service account, so N concurrent Redis connections on the same pod share a
single token and avoid N concurrent blocking refreshes.
get_credentials_async() already used asyncio.to_thread (non-blocking), and is
updated to call _get_cached_gcp_iam_token so it also benefits from caching.
## Tests
- Updated existing test that expected a fresh token on every call to reflect
the new caching behaviour.
- Added tests for: cache hit (no redundant I/O), cache expiry and refresh,
and cache sharing across multiple provider instances.
- Added autouse fixture to clear the module-level cache between tests.
Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>
* refactor(redis): remove unused Optional import from _redis_credential_provider.py
* refactor(redis): improve documentation for GCPIAMCredentialProvider class
Updated the docstring for the GCPIAMCredentialProvider class to clarify its purpose and the caching mechanism for GCP IAM tokens. The changes enhance readability and maintainability by providing a more concise explanation of the token caching strategy and its benefits for Redis authentication.
* refactor(redis): improve documentation for GCPIAMCredentialProvider class
Updated the docstring for the GCPIAMCredentialProvider class to clarify its purpose and the caching mechanism for GCP IAM tokens. The changes enhance readability and maintainability by providing a more concise explanation of the token caching strategy and its benefits for Redis authentication.
---------
Co-authored-by: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>
user