Skip to main content
PATCH
Update dataset item

Authorizations

Authorization
string
header
required

Organization API key, generated in studio.sidenet.ai. Backend only — never in a browser.

Path Parameters

id
string
required

Dataset id.

Example:

"ds_orders_v3"

itemId
string
required

Dataset item id.

Example:

"dsi_2c7e91b4"

Body

application/json
input

The prompt the agent is given for this item. A string is one user message; an array of { role, content } messages sets up a multi-turn prompt whose last message is the turn the agent answers.

Example:

"Can I get a refund on an order I placed yesterday?"

groundTruth

The expected answer, for scorers that compare against one (answer similarity, faithfulness, …). A string is the usual shape; an object is accepted for structured outputs.

Example:

"Yes — orders can be refunded within 14 days. Ask for the order number."

requestContext
object | null

Per-item run settings. threadId pins the item to a conversation thread (several items sharing one thread form one conversation; run those with maxConcurrency 1). variables fills the {{placeholders}} in the agent's prompt blocks for this item, merged over the run-level variables leaf by leaf. No other keys are accepted.

metadata
object | null

Free-form labels for the item (a category, a source ticket, …). Passed to scorers as additional context.

Example:
toolMocks
object[] | null

Canned tool results for this item. When the agent calls toolName with args, the run serves output instead of executing the tool — so a side-effecting tool never runs during an experiment and the item is deterministic. toolName is the name the agent sees (the one in a result's toolCalls). args are matched by deep equality (key order ignored, array order significant); set matchArgs: "ignore" to match on the name alone, e.g. for a sub-agent whose prompt is free text. Several mocks for the same tool are consumed in call order. A mocked tool called with other arguments, or more often than it has mocks, fails the item (error.code TOOL_MOCK_MISMATCH / TOOL_MOCK_EXHAUSTED). Tools with no mock run live unless unmockedToolPolicy is deny. null removes every mock.

Maximum array length: 50
Example:
unmockedToolPolicy
enum<string> | null

What the run does with a tool call this item has no mock for: allow runs the real tool, deny fails the item (error.code TOOL_MOCK_NOT_DECLARED) before it runs. Overrides the run-level unmockedToolPolicy of POST /v1/experiments; the default is allow.

Available options:
allow,
deny
Example:

"deny"

Response

The updated item

item
object