Interface: MockProviderOptions

Defined in: src/adapters/llm/MockProvider.ts:36

Properties

chunkDelayMs?

readonly optional chunkDelayMs?: LatencyMs

Defined in: src/adapters/llm/MockProvider.ts:91

For stream(): delay between successive chunks (ms). Pass a single number for a fixed delay or a [min, max] tuple for a uniformly random delay per chunk (e.g. [30, 80] for typing-like cadence). Default 30ms.

Has no effect on complete().

delayMs?

readonly optional delayMs?: LatencyMs

Defined in: src/adapters/llm/MockProvider.ts:82

Alias for thinkingMs. Kept for back-compat with prior revisions.

name?

readonly optional name?: string

Defined in: src/adapters/llm/MockProvider.ts:37

replies?

readonly optional replies?: readonly MockReply[]

Defined in: src/adapters/llm/MockProvider.ts:62

Scripted replies for multi-turn / tool-using agents. Each entry is consumed in order — iteration 1 reads replies[0], iteration 2 reads replies[1], and so on. Use Partial to inject toolCalls:

mock({
  replies: [
    { toolCalls: [{ id: '1', name: 'lookup', args: { id: 42 } }] },
    { content: 'Found it: refunds take 3 business days.' },
  ],
});

Exhaustion semantics: if the agent calls the LLM more times than there are replies, complete() / stream() throw a clear error. This makes mock-script bugs loud, not silent. Tune the agent's maxIterations to bound the call count.

Takes precedence over reply and respond when set.

reply?

readonly optional reply?: string

Defined in: src/adapters/llm/MockProvider.ts:39

Fixed response content. Overrides respond when set.

respond?

readonly optional respond?: (req) => string | Partial<LLMResponse>

Defined in: src/adapters/llm/MockProvider.ts:71

Build the response from the request. Returns either a plain string (renders as content with no tool calls) or a partial LLMResponse so consumers can simulate tool calls + multi-turn loops without needing a separate scripted() helper.

Default: echoes the last user message.

Parameters

req

LLMRequest

Returns

string | Partial<LLMResponse>

stopReason?

readonly optional stopReason?: string

Defined in: src/adapters/llm/MockProvider.ts:93

Fixed stop reason to return. Default 'stop'.

thinkingMs?

readonly optional thinkingMs?: LatencyMs

Defined in: src/adapters/llm/MockProvider.ts:80

Simulated wall-clock delay per request (ms). Pass a single number for a fixed delay or a [min, max] tuple for a uniformly random delay (e.g. [3000, 8000] for "real LLM" thinking time). Default 0 (instant).

Aliased via delayMs for backward compatibility.

usage?

readonly optional usage?: Readonly<{ cacheRead?: number; cacheWrite?: number; input?: number; output?: number; }>

Defined in: src/adapters/llm/MockProvider.ts:95

Override usage counts returned. Default: chars/4 heuristic.