Koha is a great project (another great open project is Evergreen)... but last I checked it doesnt really do document management. Outside of the librarian/information professional communities there seems to be a lack of understanding about information categorization and classification.
There are fairly large differences in how you handle the storage, metadata and retrieval options for something that could 'circulate' in a traditional library and a component piece (article/paper). Traditionally there have been specialized resources for source level objects (books/journals) and there were specialized resources for digging deeper and locating chapters/articles/papers/abstracts within source objects.
I commonly see researchers checking a library catalog for a known article title, when the OPAC only includes journal title data
Maybe this is something new with the blurring of publishing on the web? There has been a lot of work on federated searching as a a solution to this... but in most implementations my experience has been that patrons lose the ability to use the individual resources to fullest effect and get a results list that leads to further confusion.
In any case, Koha would be great for setting up a lending library, and could be used as a framework for managing the metadata (title, author, year) with a link to a file... but in terms of actually having searchable text/OCR and additional tools it would be a poor choice (something like LibraryThing would be simpler if you werent worried about checkin/checkout, patron records, late fees etc)