mirror of
https://github.com/BerriAI/litellm.git
synced 2026-10-12 03:48:56 +00:00
0338498067
* fix(router): preserve native baseline identity and accounting * fix(router): preserve injected system caches in native baselines * fix(router): prepare native baselines through shared request owners * fix(router): capture native baseline fields from the provider schema * refactor(router): reuse native provider parameter discovery * fix(router): keep long native baselines and abstain after compaction Message history no longer counts against the settings snapshot budget, so long and non-ASCII native sessions keep modeled baselines. Selected-tier compaction now abstains because the baseline would otherwise inherit the compacted history. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(router): judge implicit caching against the selected request The implicit-cache guard compared the selected response's cache usage with the projected baseline's breakpoints, so selected-tier cache markers made a usable unmarked baseline plan look like unexplained caching. The guard now checks the selected wire request. Also removes a stamp-reuse branch that could never run because routing clears the stamp first; every pass already captures caller settings from fresh kwargs. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
59 lines
2.3 KiB
Python
59 lines
2.3 KiB
Python
from typing import Final
|
|
|
|
from litellm.router_utils.baseline_request import baseline_request, capture_baseline_parameters
|
|
|
|
|
|
def test_baseline_snapshot_owns_nested_caller_settings_and_overrides_routed_settings() -> None:
|
|
reasoning: Final = {"effort": "medium"}
|
|
snapshot: Final = capture_baseline_parameters({"reasoning": reasoning, "verbosity": "low"})
|
|
assert snapshot is not None
|
|
reasoning["effort"] = "high"
|
|
projected: Final = baseline_request(
|
|
{"messages": [{"role": "user", "content": "hello"}], "reasoning": reasoning, "verbosity": "high"},
|
|
snapshot,
|
|
{"verbosity": "medium"},
|
|
)
|
|
assert projected == {
|
|
"messages": [{"role": "user", "content": "hello"}],
|
|
"reasoning": {"effort": "medium"},
|
|
"verbosity": "low",
|
|
}
|
|
|
|
|
|
def test_oversized_snapshot_fails_closed_before_json_validation() -> None:
|
|
assert capture_baseline_parameters({"output_config": {"format": "x" * 5_000_000}}) is None
|
|
|
|
|
|
def test_snapshot_retains_extra_body_settings_but_no_credentials() -> None:
|
|
assert capture_baseline_parameters({"api_key": "private", "extra_body": {"verbosity": "low"}}) == {
|
|
"extra_body": {"verbosity": "low"}
|
|
}
|
|
|
|
|
|
def test_chat_projection_applies_extra_body_after_top_level_parameters() -> None:
|
|
snapshot: Final = capture_baseline_parameters({"verbosity": "high", "extra_body": {"verbosity": "low"}})
|
|
assert snapshot is not None
|
|
assert baseline_request({}, snapshot, {}) == {"verbosity": "low"}
|
|
|
|
|
|
def test_baseline_projection_keeps_caller_tools_and_request_parameter_precedence() -> None:
|
|
from litellm.router import Router
|
|
|
|
deployment: Final = {
|
|
"tools": [{"type": "function", "function": {"name": "configured"}}],
|
|
"tool_choice": "required",
|
|
"max_tokens": 64,
|
|
}
|
|
caller: Final = {
|
|
"tools": [{"type": "function", "function": {"name": "caller"}}],
|
|
"tool_choice": "auto",
|
|
"max_tokens": 128,
|
|
}
|
|
actual_request: Final = dict(caller)
|
|
Router._merge_tools_from_deployment({"litellm_params": deployment}, actual_request)
|
|
snapshot: Final = capture_baseline_parameters(caller)
|
|
assert snapshot is not None
|
|
assert baseline_request({"tools": [{"name": "routed-only"}], "max_tokens": 4}, snapshot, deployment) == {
|
|
**deployment,
|
|
**actual_request,
|
|
}
|