CostEstimate

@forge/monorepo

---

@forge/monorepo / backend/src / CostEstimate

Type Alias: CostEstimate

CostEstimate = object

Defined in: backend/src/usage/index.ts:82

Properties

cachedInputTokens?

readonly optional cachedInputTokens?: number

Defined in: backend/src/usage/index.ts:97

How much of inputTokens the caller expects to be served from cache — task #247.

Optional, and its absence means "assume none", which is what this did unconditionally before. That default is deliberately the conservative one for a spend limit: over-reserving refuses a turn slightly early, under-reserving admits a turn that then exceeds the ceiling, and only the second one costs money.

A seam rather than a prediction. The platform cannot know a hit rate before the call, and inventing one would make every reservation wrong in a new way; a caller that has just observed 97% on the previous turn of the same conversation genuinely does know something, and this is where it says so.

---

inputTokens

readonly inputTokens: number

Defined in: backend/src/usage/index.ts:84

---

maxOutputTokens

readonly maxOutputTokens: number

Defined in: backend/src/usage/index.ts:85

---

modelId

readonly modelId: string

Defined in: backend/src/usage/index.ts:83