Skip to main content

Class: QueryGenerator

Defined in: packages/agentos/src/orchestration/pipeline/query/QueryGenerator.ts:136

Builds tier-appropriate prompts and generates LLM answers.

The generator selects a model (standard vs. deep) based on the query tier, constructs a system prompt with optional documentation context and research synthesis, then delegates to generateText for the actual LLM call.

Example​

const gen = new QueryGenerator({
  model: 'openai:gpt-4.1-mini',
  modelDeep: 'openai:gpt-4.1',
  provider: 'openai',
});

const result = await gen.generate('How does auth work?', 1, chunks);
console.log(result.answer);

Constructors​

Constructor​

new QueryGenerator(config): QueryGenerator

Defined in: packages/agentos/src/orchestration/pipeline/query/QueryGenerator.ts:147

Create a new QueryGenerator instance.

Parameters​

config​

QueryGeneratorConfig

Generator configuration specifying models, provider, and optional credential overrides.

Returns​

QueryGenerator

Methods​

generate()​

generate(query, tier, chunks, researchSynthesis?): Promise<GenerateResult>

Defined in: packages/agentos/src/orchestration/pipeline/query/QueryGenerator.ts:167

Generate an answer for the given query at the specified complexity tier.

Parameters​

query​

string

The user's original question.

tier​

QueryTier

The classified complexity tier (0–3).

chunks​

RetrievedChunk[]

Retrieved documentation chunks, sorted by relevance.

researchSynthesis?​

string

Optional research narrative (T3 only).

Returns​

Promise<GenerateResult>

A promise resolving to the generated answer, model used, and token usage.