Interface: JudgmentRow

Defined in: src/core/agent/findings/types.ts:259

A SECOND SOURCE's reading of one result (9.104.0, .findings({ judge })): a calibrated classifier (agentfootprint/classify) asked what the result is worth for the proposition the model declared before the call — or, when the call declared none, for the user's question (against says which). Written beside the model's own StandingRow, never merged with it and never served in its place: foldLedger(...).standingOf stays the model's reading, foldLedger(...).judgments is the judge's. A disagreement between the two is a FACT of the record, resolved by nobody.

Every field is the provider's own data or a measurement around the call — probabilities as sent (never renormalised), confidence as sent, testsSubject the provider's probability that the result tests the proposition at all, usage when the provider reported it, latencyMs measured by the adapter. Nothing here is inferred by the library.

Properties

against

readonly against: "question" | "proposition"

Defined in: src/core/agent/findings/types.ts:268

What the result was judged AGAINST: the call's declared proposition, or the user's question.

---

clipped?

readonly optional clipped?: true

Defined in: src/core/agent/findings/types.ts:277

Set when the result text was cut at JUDGE_RESULT_CHARS before the judge saw it.

---

confidence

readonly confidence: number

Defined in: src/core/agent/findings/types.ts:271

---

iteration

readonly iteration: number

Defined in: src/core/agent/findings/types.ts:279

The iteration whose dispatch landed the result.

---

judge

readonly judge: object

Defined in: src/core/agent/findings/types.ts:266

The classifier's port name and the provider's resolved model string.

model

readonly model: string

name

readonly name: string

---

kind

readonly kind: "judgment"

Defined in: src/core/agent/findings/types.ts:260

---

latencyMs

readonly latencyMs: number

Defined in: src/core/agent/findings/types.ts:275

---

probabilities

readonly probabilities: Readonly<Record<Standing, number>>

Defined in: src/core/agent/findings/types.ts:270

---

source

readonly source: "judge"

Defined in: src/core/agent/findings/types.ts:264

---

standing

readonly standing: Standing

Defined in: src/core/agent/findings/types.ts:269

---

testsSubject?

readonly optional testsSubject?: number

Defined in: src/core/agent/findings/types.ts:273

The provider's probability that the result tests the proposition / question at all.

---

toolCallId

readonly toolCallId: string

Defined in: src/core/agent/findings/types.ts:262

The RESULT judged — the tool call's id, the same key StandingRow uses.

---

toolName

readonly toolName: string

Defined in: src/core/agent/findings/types.ts:263

---

turn?

readonly optional turn?: number

Defined in: src/core/agent/findings/types.ts:286

The conversation turn the row was filed in (AgentState.turnNumber) — stamped by the one writer while an honesty layer is armed (the inputs layer, the answer layer), absent otherwise. The ledger crosses turns on a continued conversation, and iteration restarts at 1 every run.

---

usage?

readonly optional usage?: object

Defined in: src/core/agent/findings/types.ts:274

inputTokens

readonly inputTokens: number

outputTokens

readonly outputTokens: number