Aksara

The problem

Otorita IKN produces a lot of institutional knowledge across its ten Eselon-1 units: regulations, technical reports, academic studies and field notes. The old system kept it all but made it hard to use. There was no search, so staff asked colleagues for files over chat. Uploading meant a long two-step form with no drafts, so most knowledge never got uploaded, and managers had no view of what existed or which units were contributing.

Three groups share Aksara: the public, browsing published knowledge with no login; staff, uploading and reading internal material; and verifiers, curators and administrators, who approve and organise it. Much of the material isn't public, and some is limited to the directorate that owns it.

What I built

Every document gets a permanent public code

Each document has a private internal ID from the moment it's a draft. On its first publish it also gets a public code, and that code becomes its public address. I chose this because a public link has to be permanent and citable, and a link built from the title would break the moment someone fixed a typo. Codes are never reused, and a document that's unpublished and republished gets its old code back.

One function decides who can open a restricted document

There are three access levels: public, internal to OIKN, and restricted, checked on the server for every request against a person's work unit, rank, roles and any temporary access they were granted. One function decides who can open a restricted document without asking, called from every place a document is served, so the rule can't drift between the listing, search and file downloads. I first gave that power to a broader admin role, then narrowed it to one superadmin role once I saw how sensitive it was.

Review follows fixed steps with a full history

A document moves from draft to verification, then curation, then published, or it's returned or rejected. Only the system changes that status, never a manual edit. Verifiers and curators each work a split-screen queue with three choices: approve, return with a required note, or reject. Every step lands on a timeline nobody can edit, and a revised, already-published document stays live until the revision passes review again.

Search that forgives typos and keeps the link

I shipped database full-text search with typo tolerance and a "did you mean" suggestion, behind a single interface so a larger search engine can replace it later without touching the pages. Every filter, sort and page number lives in the address bar, so a filtered result can be bookmarked, shared or refreshed.

National legal records are federated, never copied

National legal records live in a system OIKN already runs, called JDIH, so Aksara never copies its documents. I store only light metadata, refresh it with a daily job, and fetch the real content from JDIH's API when someone opens a record, cached briefly server-side. Copying JDIH's store instead would have created a second copy of the same regulation that drifts out of date. The integration ended up bigger than planned: it now includes an auto-classifier and its own full-text search, on top of the metadata sync the original spec called for.

Result

I built Aksara alone, starting in July 2026, and it is live at aksara.ikn.go.id. The public portal is open to any visitor, and every document is checked on the server against three access levels. Documents are filed under 43 organizational units, 10 knowledge domains and 92 more specific topic subcategories. Beyond the original plan, the JDIH integration grew its own auto-classifier and search index, and I added a gazetteer and an achievements module. A native mobile app now shares the same backend.