Interface: LLMCallOptions

Defined in: src/core/LLMCall.ts:82

Properties

contextBudget?

readonly optional contextBudget?: object

Defined in: src/core/LLMCall.ts:104

Per-slot context budgets, in characters (8.11.0). The LLMCall twin of AgentOptions.contextBudget — two slots here, since an LLMCall has no tools slot.

Each slot warns (and emits agentfootprint.context.budget_pressure) when it composes over its budget. Nothing is truncated — the full content still reaches the LLM; the budget is a signal, not a limiter. Defaults: systemPrompt 4000, messages 10000.

messages?

readonly optional messages?: number

systemPrompt?

readonly optional systemPrompt?: number

---

costBudget?

readonly optional costBudget?: number | { onExceed: "warn" | "halt"; usd: number; }

Defined in: src/core/LLMCall.ts:125

Cumulative USD budget per run. When provided along with pricingTable, LLMCall emits agentfootprint.cost.limit_hit with action: 'warn' the first time cumulative USD crosses the budget. Execution continues — consumers choose whether to abort by listening to the event.

The object form { usd, onExceed } is accepted for symmetry with Agent, but onExceed must be 'warn' here: halting means "stop at the next iteration boundary", and one call has no next boundary. 'halt' is refused at build rather than silently ignored.

---

groupTranslator?

readonly optional groupTranslator?: GroupTranslator<unknown>

Defined in: src/core/LLMCall.ts:143

Optional per-COMPOSITION translator (UI-agnostic). See core/translator.ts. When attached, runner.getUIGroup() invokes it with the LLMCall's GroupMetadata (kind 'LLMCall', id, name, empty members[], plus extra.slots with the three slot ids — system-prompt, messages, tools — so Lens can render the slot cards inside an LLMCall card without inspecting buildTimeStructure). Returns undefined when omitted.

---

id?

readonly optional id?: string

Defined in: src/core/LLMCall.ts:87

Stable id used for topology + events. Default: 'llm-call'.

---

maxTokens?

readonly optional maxTokens?: number

Defined in: src/core/LLMCall.ts:93

Optional max output tokens.

---

model

readonly model: string

Defined in: src/core/LLMCall.ts:89

Model to request from the provider.

---

name?

readonly optional name?: string

Defined in: src/core/LLMCall.ts:85

Human-friendly name shown in events/metrics. Default: 'LLMCall'.

---

pricingTable?

readonly optional pricingTable?: PricingTable

Defined in: src/core/LLMCall.ts:113

Pricing adapter. When set, LLMCall emits agentfootprint.cost.tick after every LLM response with per-call and cumulative USD. Run-scoped — the cumulative resets on each .run().

---

provider

readonly provider: LLMProvider

Defined in: src/core/LLMCall.ts:83

---

structureRecorders?

readonly optional structureRecorders?: readonly StructureRecorder[]

Defined in: src/core/LLMCall.ts:133

Optional build-time recorders threaded into footprintjs's flowChart() factory. Each recorder observes per-node build events (onStageAdded / onSubflowMounted / etc.) for this LLMCall's internal chart (Initialize + slot mounts + CallLLM). When omitted, no build-time observation is wired up.

---

temperature?

readonly optional temperature?: number

Defined in: src/core/LLMCall.ts:91

Optional sampling temperature.