@juspay/neurolink
Version:
Universal AI Development Platform with working MCP integration, multi-provider support, voice (TTS/STT/realtime), and professional CLI. 58+ external MCP servers discoverable, multimodal file processing, RAG pipelines. Build, test, and deploy AI applicatio
32 lines (31 loc) • 1.24 kB
TypeScript
/**
* Autoresearch task executor — bridges TaskManager with the
* autoresearch experiment loop.
*
* Each tick:
* 1. Loads/creates a ResearchWorker for the task's tag
* 2. Gets the phase-appropriate tool filter
* 3. Calls NeuroLink.generate() with research tools + prompt
* 4. Advances phase based on which tools the AI called
* 5. Returns a TaskRunResult
*
* Workers are cached by tag to avoid re-initialization on each tick.
* Forwards the NeuroLink emitter to each worker for lifecycle events.
*/
import type { AutoresearchEmitter, NeuroLinkExecutable, Task, TaskRunResult } from "../types/index.js";
/**
* Clear all cached workers. Called by TaskManager.shutdown().
*/
export declare function clearWorkerCache(): void;
/**
* Execute one autoresearch tick for a task.
*
* Returns a TaskRunResult-shaped object.
* If the task is missing autoresearch config, returns an error result
* instead of throwing (so the scheduler can record the failure).
*
* @param emitter - Optional emitter to forward autoresearch lifecycle events
*/
export declare function executeAutoresearchTick(task: Task & {
autoresearch?: unknown;
}, neurolink: NeuroLinkExecutable, emitter?: AutoresearchEmitter): Promise<TaskRunResult>;