Usage records

CyteOnto.usage is an AgentUsage that adds up the requests and tokens of the successful LLM calls of the instance, to describe and to split labels, including those of from_config(). Failed attempts are not counted. CyteOnto.model_pair_usage is a ModelPairUsage that records whether the primary or the fallback models were used.

print(cyto.usage.requests, cyto.usage.totalTokens)
print(cyto.model_pair_usage.to_status_dict())

model_pair_usage.embedding is the full embedding configuration, including its apiKey, so print or log to_status_dict(), which holds only the names. Models and providers explains when the fallback models are used.

class cyteonto.AgentUsage(*, agentName='', modelName='', requests=0, inputTokens=0, outputTokens=0, totalTokens=0, toolUsage=<factory>)

Tally of LLM calls made through a given agent.

agentName: str = ''

Name of the agent, CyteOnto for CyteOnto.usage.

modelName: str = ''

The model id.

requests: int = 0

Number of model requests, including the requests that follow tool calls.

inputTokens: int = 0

Number of input tokens.

outputTokens: int = 0

Number of output tokens.

totalTokens: int = 0

Number of input and output tokens.

toolUsage: dict[str, int] = {}

Number of calls of each tool, such as get_pubmed_abstracts.

record(model_name, run_usage, tool_counts=None)

Add the usage of one pydantic-ai run and its tool calls.

Return type:

None

merge(other)

Add the counts of other to this tally.

Return type:

None

class cyteonto.ModelPairUsage(*, llmTier='primary', embeddingTier='primary', llm, embedding)

Which model pair tiers were used during a run.

llmTier: str = 'primary'

primary until the fallback agent is asked to describe labels, then mixed. The type also allows fallback.

embeddingTier: str = 'primary'

Which models made the label embeddings that were matched, new or cached: primary, fallback, or mixed for both. Embedding the ontology terms does not count.

llm: cyteonto.LlmConfig

Required. The primary LLM, replaced by fallback_llm, if one was passed, when llmTier becomes mixed.

embedding: cyteonto.EmbdConfig

Required. The primary embedding configuration, replaced by the fallback configuration once labels are matched with its embeddings.

to_status_dict()

Return the tiers and the provider and model names as strings.

Return type:

dict[str, str]