Documents · AI + codeClient project

Journal-ready formatting: 3–5 hours → 2–5 minutes

A team member sends a draft; the system returns a document laid out for the chosen journal plus a check report. Another journal takes seconds, with no second AI pass.

ClientAn agency that prepares journal publications

  • Used every day
  • 75+ documents in two weeks
  • The client adds journals itself
How it works
2–5 mininstead of 3–5 hours of manual formatting
Pythonn8nOpenAIDocker
stepautomations / document-formatting-botVideo
A team member sends a draft in Telegram and minutes later gets a document formatted for the chosen journal.Watch on YouTube

01Before

Every journal has its own rules: margins, fonts, headings, citation and reference styles, required sections. The team reformatted every document by hand, hours each. If one journal said no, everything started again for the next, and a missed error in the references sent the document straight back.

02What changed

  • Hours become minutes

    a document formatted for the journal arrives in 2–5 minutes instead of 3–5 hours of manual work.

  • Another journal in seconds

    a button under the finished file rebuilds it for the next journal with no second AI pass.

  • Checked before submission

    references, citations and journal requirements are checked automatically, with doubtful spots highlighted right in the file.

  • Journals added in-house

    a new journal is added from a link to its guidelines, and the system builds its profile itself.

03How it works

  1. A team member sends a draft in Telegram

  2. The AI parses the document's structure once

  3. Code lays it out for the journal and checks it

  4. The finished file and report arrive in minutes

stepautomations / document-formatting-bot01 / 06

01The draft on the left, the same document after the system on the right: the journal's margins, font and style, numbered sections; a citation with no source in yellow

Client feedback

The client says the system took the biggest routine burden off the team and saved them a huge amount of time. A second project followed, and we're discussing a third.

Where else it worksThe same approach, built for your task

Here it's journals. “AI parses the document once, code builds it to any standard” works wherever the same text has to be submitted in different formats.

  • 01Tenders and grants

    one application, fitted to each platform's or fund's rules, with required sections checked.

  • 02Reports on a client's template

    your data in the format your bank, investor or head office asks for.

  • 03Documents to national standards

    layout, numbering, lists and references, all automatic.

  • 04Publishers and media

    a manuscript in the house style, with citations and sources checked.

Input via Telegram, as here, or a web form, email or your own system. I'll build it for your formats.

Under the hoodTechnical details for specialists
  • AI once, then code

    the document's structure is stored inside the file, so every further journal is built without the model: fast, free and identical every time.

  • Numbers stay untouched

    table values come from the original, not from the model's output.

  • Word stays editable

    tables, charts, formulas and diagrams carry over as native Word objects, not images; tested on 45 real files.

Stack

n8nPython (FastAPI, python-docx, lxml)DockerOpenAICrossref (DOI)LibreOfficeTelegram bot

Sample data: a fictional document, authors and journals processed by the same service the client uses.

Show me the processyou want off your plate.

Describe the task in your own words. I'll reply personally and tell you what can be automated right away.

  1. You describe the task
  2. I reply personally
  3. You get a written plan
INTAKE WORKFLOW: ACTIVE
Email, Telegram or phone. Any channel that works for you.

I reply personally, usually the same day. This form runs on the same n8n stack I build for clients.