Answerwell

Make institutional knowledge findable

Answerwell adds structured Q&A to Confluence: questions, accepted answers, follows, tags, moderation, and expert routing. Questions and answers use native pages and comments, so the collaboration model remains tied to the host platform's content and permissions.

The engineering challenge is the boundary between native state and app-owned metadata. Events may be duplicated or reordered, content may become inaccessible, and search indexing is asynchronous. Reconciliation and permission checks have to keep lists, answer status, and notifications consistent with what a viewer can actually read.

Validate assumptions on the host platform

The implementation was deployed to real development and staging sites and exercised through multiple test personas, notification flows, forged-call checks, and uninstall/reinstall behavior. Native content survives uninstall while app storage is removed, and the product treats that distinction explicitly.

A separate benchmark used a 300-question test space to measure real REST read paths. Moving vote sorting into CQL reduced the warm p90 from 1.76 seconds to 0.60 seconds. These timings exclude the Forge bridge, app-data invocation, and rendering; they measure the data path rather than the complete user experience.

Release state

The app is awaiting marketplace and business setup. The Atlassian Questions import path was exercised end to end; an additional legacy importer remains a labeled preview because it needs validation against real existing data. Answerwell remains a working product name.