[Idea]: manual annotations

[FlorianHeigl]

What are you trying to do?

I'm trying to find a workflow for multiple volunteers to comment on PDF files (annotations on specific spots)
It doesn't matter if it's the processed content or the original file.

What feels difficult or missing today?

Currently, one can run an AI chat on the topic and save that, and it'll have references to certain spots.
It's not possible (as far as I see) to do that myself. There should be a way to highlight a text sequence and pass that into the current notebook or to have it sent to the notebook chat as a "focus".

What outcome would help?

I would love to be able to have the processed or original documemt on the left of the "wysiwyg" editor and be able to write something like "line 83 makes no sense because" on the right, in some click-based way. like without copying, just highlight+reference as a button.

How do you handle this today?

I look at it and copy, at the moment I'm also still in uploading/embedding stuff.
I don't see (at the moment) how to create and store actual annotations which is a huge issue.
I also don't see how I can have the model break apart larger work items (say a long PDF)

Additional context, examples, or possible directions

My context is covid-related research papers. Currently interested volunteers all have to do the manual work looking for flaws in the papers. I wanna merge AI based preprocessing and human comments/notes. One of the factors is breaking down work, so i.e. you might have two people go through a certain chapter and they ought to be able to compare their impressions somehow...
but the other obvious thing would be IT security guides (CISA, BSI, ...) which can be 1000s of pages and need to be split up, but one also needs to be able to say "that item here is outdated, they have a mistake there"

If you need me to flesh this out a lot more, I can try. My limitation is that I don't understand how user interfaces are really built, so I am stuck with very vague terms for some of it. At worst, I can upload a few 100 files and do more practical testing and then add here. I already opened the discussion because now I still have a neutral way of describing it, once I 've done more daily work in Open Notebook I'll not be able to think outside it's box as good as before.

How would you like to participate?

• I can add use cases or join the product/design discussion.
• I can help explore the technical approach.
• I can help implement an approved work item.
• I can test or validate a proposed solution.

Before posting

• I searched existing Discussions and Issues for related proposals.

[lfnovo]

This is one of the richest ideas in the queue — and I want to acknowledge the meta-move: opening it before daily use reshapes your framing is exactly the right instinct. Fresh eyes are valuable.

Reading it, I see four distinct needs:

1. Manual annotations anchored to a passage — highlight a span in the processed/original document and attach your own comment, without copying. Today only the AI can produce passage-anchored output; humans can't. That's the core gap, and it's real.
2. Highlight → send to chat as focus — same reading surface, different action: select a passage and make it the anchor of a conversation.
3. Multiple volunteers comparing impressions — constrained by a deliberate posture: Open Notebook is single-user first (PDR-001 in the decision records). Multi-user has its own discussion (Multi-user support — what should it mean for Open Notebook? #1254); I'd keep collaboration there, and design annotations so the data model doesn't preclude it later.
4. Breaking huge documents into work units (chapters of a 1000-page guide) — [Idea]: Textbook Slicing #1230 (Textbook Slicing) is circling the same need; connecting the two.

On direction: (1) and (2) align well with where the product is heading — passage-level references are a recurring theme (e.g. #294), and deepening the Source reading experience is very much on the map. No commitment on when/how yet, but this is exploration worth investing in.

Since you offered to add use cases: when two volunteers review the same chapter, what does "comparing their impressions" concretely look like — seeing each other's annotations side by side, a merged view, or a discussion thread per annotation? That answer shapes whether annotations are notes-with-anchors or something more conversational.

Status: exploring — (1)+(2) as the core; collaboration routed to #1254; work-breakdown connected to #1230.

[FlorianHeigl]

I'll keep this as a TODO and come back with feedback after consulting with a few people. roughly a month to get feedback, need to see if I can find screenshots of some legal annotation tools which have implemented this already in some (unknown to me) manner.