The use case: We build our AI agents' Knowledge Base (KB) from structured local Markdown kept in a version-controlled source of truth (Obsidian/Git), then upload the .md files to the KB. The KB is a derived copy of that source so when the source changes, the KB needs to be updated to match.

The limitation: Today, re-sync is only available for Google Drive files. For files uploaded "from computer," the only way to update content is delete the old file and re-add the new one, manually, per file. That's fine for one small KB, but it doesn't scale. It invites drift (the exact thing a grounded agent can't afford), and it breaks down when you're running multiple agents whose knowledge updates regularly.

The ask (either would solve it):

Extend re-sync to uploaded/local files (match content by filename, pull the latest), or expose a KB document upsert endpoint in the v2 API (create/replace/delete an individual file document in a KB), so we can script the sync from our source of truth.

Why I believe this matters:

HL is actively pushing Conversation AI toward richer knowledge (files, tables, rich text, re-ranking). Teams that treat their KB as a maintained/structured asset, which is exactly the behavior that makes these agents good need a way to keep that KB current without hand-syncing. This closes the loop between a maintained knowledge source and the live agent.