Class: EchoLLM

LLM stand-in. Answers with answer (a string, or a function of the query and documents), or else with the top document's content. Every call is recorded in calls, so tests can assert on what the pipeline sent.

Constructors

Constructor

new EchoLLM(options?): EchoLLM;

Parameters

Parameter	Type	Description
options?	{ answer?: string | ((arg0, arg1) => string); fallback?: string; }	-
options.answer?	string | ((arg0, arg1) => string)	-
options.fallback?	string	Answer when there are no documents and no answer

Returns

EchoLLM

Properties

Property	Type	Description
answer	string | ((arg0, arg1) => string)	The fixed answer, or a function of (query, docs) that returns it.
calls	object[]	-
fallback	string	Returned when there are no documents and no answer.
metadata	object	-
metadata.name	string	-
metadata.type	string	-
metadata.version	string	-

Methods

generate()

generate(
   query,
   docs?,
   options?
): Promise<string>;

Parameters

Parameter	Type	Description
query	string	-
docs?	any[]	-
options?	{ signal?: AbortSignal; }	-
options.signal?	AbortSignal	-

Returns

Promise<string>

---

generateStream()

generateStream(
   query,
   docs?,
   options?
): AsyncGenerator<{
  done: boolean;
  token: string;
}, void, unknown>;

Streams the same answer word by word, then a done chunk.

Parameters

Parameter	Type	Description
query	string	-
docs?	any[]	-
options?	{ signal?: AbortSignal; }	-
options.signal?	AbortSignal	-

Returns

AsyncGenerator<{ done: boolean; token: string; }, void, unknown>

Yields

---

getCacheKey()

getCacheKey(role): string;

Cache identity for createRagPipeline({ cache }); none when answer is a function, since its output cannot be identified.

Parameters

Parameter	Type	Description
role	string	-

Returns

string