trulens.core.schema.base¶

trulens.core.schema.base ¶

Common/shared serializable classes.

Attributes¶

MAX_DILL_SIZE `module-attribute` ¶

MAX_DILL_SIZE: int = 1024 * 1024

Max size in bytes of pickled objects.

Classes¶

Cost ¶

Bases: SerialModel, BaseModel

Costs associated with some call or set of calls.

Attributes¶

n_requests `class-attribute` `instance-attribute` ¶

n_requests: int = 0

Number of requests.

To increment immediately when a request is made.

n_successful_requests `class-attribute` `instance-attribute` ¶

n_successful_requests: int = 0

Number of successful requests.

To increment only after a response is processed and it indicates success.

n_completion_requests `class-attribute` `instance-attribute` ¶

n_completion_requests: int = 0

Number of completion requests.

To increment immediately when a completion request is made.

n_classification_requests `class-attribute` `instance-attribute` ¶

n_classification_requests: int = 0

Number of classification requests.

To increment immediately when a classification request is made.

n_classes `class-attribute` `instance-attribute` ¶

n_classes: int = 0

Number of class scores retrieved.

To increment for each class in a successful classification response.

n_embedding_requests `class-attribute` `instance-attribute` ¶

n_embedding_requests: int = 0

Number of embedding requests.

To increment immediately when an embedding request is made.

n_embeddings `class-attribute` `instance-attribute` ¶

n_embeddings: int = 0

Number of embeddings retrieved.

To increment for each embedding vector returned by an embedding request.

n_tokens `class-attribute` `instance-attribute` ¶

n_tokens: int = 0

Total tokens processed.

To increment by the number of input(prompt) and output tokens in completion requests. While the input part of this could be incremented upon a request, the actual count is not easy to determine due to tokenizer variations and instead is usually seen in the response. Also, we want to count only tokens for successful requests that incur a cost to the user.

n_stream_chunks `class-attribute` `instance-attribute` ¶

n_stream_chunks: int = 0

In streaming mode, number of chunks produced.

To increment for each chunk in a streaming response. This does not need to wait for completion of the responses.

n_prompt_tokens `class-attribute` `instance-attribute` ¶

n_prompt_tokens: int = 0

Number of prompt tokens supplied.

To increment by the number of tokens in the prompt of a completion request. This is visible in the response though and should only count successful requests.

n_completion_tokens `class-attribute` `instance-attribute` ¶

n_completion_tokens: int = 0

Number of completion tokens generated.

To increment by the number of tokens in the completion of a completion request.

n_cortex_guardrails_tokens `class-attribute` `instance-attribute` ¶

n_cortex_guardrails_tokens: int = 0

Number of guardrails tokens generated. This is only available for requests instrumented by the Cortex endpoint.

cost `class-attribute` `instance-attribute` ¶

cost: float = 0.0

Cost in [cost_currency].

This may not always be available or accurate.

Functions¶

__rich_repr__ ¶

__rich_repr__() -> Result

Requirement for pretty printing using the rich package.

Perf ¶

Bases: SerialModel, BaseModel

Performance information.

Presently only the start and end times, and thus latency.

Attributes¶

start_time `instance-attribute` ¶

start_time: datetime

Datetime before the recorded call.

end_time `instance-attribute` ¶

end_time: datetime

Datetime after the recorded call.

latency `property` ¶

latency

Latency in seconds.

start_ns_timestamp `property` ¶

start_ns_timestamp: int

EXPERIMENTAL(otel_tracing): Start time in number of nanoseconds since the epoch.

end_ns_timestamp `property` ¶

end_ns_timestamp: int

EXPERIMENTAL(otel_tracing): End time in number of nanoseconds since the epoch.

Functions¶

__rich_repr__ ¶

__rich_repr__() -> Result

Requirement for pretty printing using the rich package.

min `staticmethod` ¶

min()

Zero-length span with start and end times at the minimum datetime.

now `staticmethod` ¶

now(latency: Optional[timedelta] = None) -> Perf

Create a Perf instance starting now and ending now plus latency.

PARAMETER	DESCRIPTION
`latency`	Latency in seconds. If given, end time will be now plus latency. Otherwise end time will be a minimal interval plus start_time. TYPE: `Optional[timedelta]` DEFAULT: `None`

of_ns_timestamps `staticmethod` ¶

of_ns_timestamps(
    start_ns_timestamp: int,
    end_ns_timestamp: Optional[int] = None,
) -> Perf

EXPERIMENTAL(otel_tracing): Create a Perf instance from start and end times in nanoseconds since the epoch.

trulens.core.schema.base¶