List traces with filtering and pagination. Returns trace metadata including latency, cost, and observation/score counts.

Get a specific trace with full details including all observations and scores.

Delete a specific trace. This action is irreversible.

List observations (generations, spans, events) with cursor-based pagination. Includes usage metrics, costs, and latency.

Get a specific observation with all details including input/output, usage, costs, and timing.

Create a score for a trace or observation. Supports numeric, boolean, and categorical score types.

List scores with filtering. Supports both trace and observation scores.

Get a specific score by ID.

Delete a specific score. This action is irreversible.

Create a score configuration that defines validation rules for scores. Supports numeric ranges, categorical values, and boolean types.

List all score configurations in the project.

Get a specific score configuration by ID.

Create a new dataset for evaluation. Datasets contain items with expected inputs/outputs for testing LLM applications.

List all datasets in the project.

Get a dataset by name including its items and run names.

Create or upsert a dataset item. Use id parameter for idempotent updates.

List items in a dataset with optional filtering.

Get a specific dataset item by ID.

Delete a specific dataset item. This action is irreversible.

Create a dataset run item linking a trace/observation to a dataset item for evaluation.

List all runs for a specific dataset.

Get a specific dataset run by name including its run items.

List all sessions. Sessions group multiple traces from the same user interaction.

Get a specific session by ID including its traces.