/**
 * questions/ports/question.generator.port — inline question generator port.
 *
 * @deprecated Inline question generation is superseded by the async
 * QuestionerQueue path (`questionerEnqueue`). Will be removed once the inline
 * fallback in `opportunity.discover.ts` is retired.
 *
 * Protocol-level read contract for decision-question generation. Implementations
 * live in the backend (see `QuestionGeneratorService`, which delegates to the
 * QuestionerAgent's `discovery` mode) and are injected into the protocol via
 * `ProtocolDeps`/`ToolContext`. The protocol module never constructs its own
 * LLM-bound generator — callers inject one (or `undefined` to opt out).
 *
 * IND-547: canonical question generator port.
 */
import type { DiscoveryQuestionInput } from "../../shared/schemas/discovery-question.schema.js";
import type { QuestionGenerationResult } from "../domain/question.schema.js";
export interface QuestionGeneratorReader {
    /**
     * Run the question generator over a single discovery turn.
     *
     * @param input  Discovery turn payload (query + negotiation digests + chat context).
     * @param options.signal  Optional AbortSignal. When aborted (deadline reached or
     *   upstream cancel) the in-flight LLM call is cancelled and `null` is returned —
     *   discovery still emits its response, just without questions.
     * @returns The structured result, or `null` when generation failed,
     *   guardrails dropped all candidates, the underlying LLM threw, or
     *   the call was aborted.
     */
    generate(input: DiscoveryQuestionInput, options?: {
        signal?: AbortSignal;
    }): Promise<QuestionGenerationResult | null>;
}