youngkingdom/dify
1
0
Fork 0
mirror of https://github.com/langgenius/dify.git synced 2026-05-06 10:28:10 +08:00
Code Issues Packages Projects Releases Wiki Activity

feat: enhance tencent trace integration with LLM core metrics (#27126)

Browse Source 
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
Co-authored-by: crazywoola <100913391+crazywoola@users.noreply.github.com>
This commit is contained in:
XlKsyt
2025-10-29 15:53:30 +08:00
committed by GitHub
parent 82890fe38e
commit 1e9142c213
13 changed files with 609 additions and 57 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
api/core/model_runtime/entities/llm_entities.py
View File

@ -38,6 +38,8 @@ class LLMUsageMetadata(TypedDict, total=False):
prompt_price: Union[float, str]
completion_price: Union[float, str]
latency: float
time_to_first_token: float
time_to_generate: float
class LLMUsage(ModelUsage):
@ -57,6 +59,8 @@ class LLMUsage(ModelUsage):
total_price: Decimal
currency: str
latency: float
time_to_first_token: float | None = None
time_to_generate: float | None = None
@classmethod
def empty_usage(cls):
@ -73,6 +77,8 @@ class LLMUsage(ModelUsage):
total_price=Decimal("0.0"),
currency="USD",
latency=0.0,
time_to_first_token=None,
time_to_generate=None,
)
@classmethod
@ -108,6 +114,8 @@ class LLMUsage(ModelUsage):
prompt_price=Decimal(str(metadata.get("prompt_price", 0))),
completion_price=Decimal(str(metadata.get("completion_price", 0))),
latency=metadata.get("latency", 0.0),
time_to_first_token=metadata.get("time_to_first_token"),
time_to_generate=metadata.get("time_to_generate"),
)
def plus(self, other: LLMUsage) -> LLMUsage:
@ -133,6 +141,8 @@ class LLMUsage(ModelUsage):
total_price=self.total_price + other.total_price,
currency=other.currency,
latency=self.latency + other.latency,
time_to_first_token=other.time_to_first_token,
time_to_generate=other.time_to_generate,
)
def __add__(self, other: LLMUsage) -> LLMUsage: