Skip to content

Feature request: background / async mode for biblio_docling

Problem

biblio_docling (Docling PDF parsing) can be slow for large or multi-page documents. Since MCP tool calls are synchronous, the calling agent blocks until parsing completes — it cannot do any other work while waiting.

Proposed solution

Add a background execution mode to biblio_docling so it can run without blocking the agent. Options:

  1. Fire-and-forget with polling: biblio_docling(..., background=True) returns immediately with a job ID. A companion tool biblio_docling_status(job_id) lets the agent check progress or retrieve results later.
  2. Callback/notification pattern: run the parsing in a subprocess or async task and post a notification (e.g. worklog capture) when done, so the agent can pick up results in a subsequent turn.
  3. Queue integration: route the request through the existing worklog task queue so it runs as a background task with Telegram notification on completion.

Why it matters

  • Agents working on literature ingestion workflows often need to parse multiple PDFs — sequential blocking calls compound the wait time.
  • The agent could be doing useful work (writing notes, processing already-parsed results) while Docling churns through a PDF.
  • Aligns with the broader pattern of making long-running MCP tools non-blocking.

Source context: pixecog

PixEcog (pixecog): Neuropixels and ECoG dataset and analysis

Recent commits: 

6429574 untrack
614785c remove CI builds/ from tracking, add to gitignore
63f0884 pipeio notebook test

README:

type: readme

Quick Start for Collaborators

Follow this checklist to get started with Pixecog documentation and workflows.

🐀 Pixecog Project — Compact Overview

Core principles

  • One immutable BIDS raw dataset (raw/) as the canonical baseline
  • Each analysis pipeline ha