Is Apache Answer the right tool for Circa?

Date: June 17, 2026 (decided) Context: Circa is the community SupportWire runs (joincirca.club). Apache Answer to be the durable, SEO-able knowledge base for the community — Q&A, badges, with participation gated to people who checked in to a Circa event physically.

Decision: Use Apache Answer — but as a knowledge base, not a warmth tool. The job is a long-term, SEO-able expertise asset for Circa, not "keep the community warm between events." Those are different jobs; this doc resolves to the knowledge-base job.

The job (settled)

Build a durable, searchable Q&A library that compounds over time and ranks in Google.
Participation (ask/answer/vote/badges) restricted to Circa attendees — scarcity + status.
Content readable online for SEO.

The model: public read, gated write

Read = public. Anyone and Google can find/read the Q&As. This is what makes it SEO-able and compounding. (The StackOverflow model.)
Write = gated to attendees. Only people who physically checked in to a Circa event can ask, answer, vote, earn badges.
Implication, stated plainly: the gate protects participation and reputation, not the content. Public content + gated contribution are the right trade for an SEO asset. You cannot have both private content and SEO — they're mutually exclusive, and SEO won.

Why Answer fits this job

Apache Answer = Q&A engine (StackOverflow). Transactional, structured, durable, searchable — exactly right for a knowledge base.
Badges/status reward the gated contributor base.
Cheap, one-click, low ops.

What this is NOT

Not the tool to "keep the community warm." Q&A is a library, not a living room — people don't feel belonging by filing questions. If warmth becomes a goal later, that's a separate decision (a forum / Discourse), not this.
Not chat. Rejected earlier as too noisy and unintentional.
Not a dedicated mobile app. Settled separately: no native app until Circa is large and daily-active enough that web is the demonstrated bottleneck. Don't relitigate.

Make it not look dead (the real risk)

An empty Q&A site looks dead, not calm. An unanswered question is worse than no platform.
Pre-seed 20–30 real Q&As before launch — genuine questions with genuine answers, not faked pairs (faked ones smell fake).
Line up 2–3 committed answerers from the attendee base who will respond within 24h.
Keep the attendee gate on write.

Success metric

60-day test on "% of questions answered within 24h" — if answerers don't show up, the asset doesn't compound and the decision was wrong.
Secondary, slower signal: organic search impressions on the public Q&A pages (the SEO payoff).

Slack message fix

The original line "keep the conversation going" is wrong for this job — Q&A is not conversation. Use: "give Circa attendees a durable, searchable place to trade expertise — public to the web, contributable only by members." Then it holds.

Setup log (as built)

Name: Ask Circa
Domain: ask.joincirca.club (CNAME → carrot-cobra.pikapod.net, DNS-only/grey cloud). TLS provisioned by PikaPods — confirmed live (HTTPS handshake succeeded).
Host: PikaPods (managed Apache Answer, trial). Kill in 30s if the test fails.
MCP: Answer's built-in MCP server is wired into Claude Code at user scope (ask-circa, SSE, https://ask.joincirca.club/answer/api/v1/mcp/sse). Status: connected. Tools load on next session restart — then Claude can seed/monitor Q&As directly.
Backups (open): do NOT reuse the backbackbackback creds (they own every prod DB backup). Create a dedicated ask-circa-backups bucket + a bucket-scoped R2 API token before enabling PikaPods S3 backups.

Next (before this is a real launch)

Seed 20–30 real Q&As via the MCP (genuine pairs, not faked).
Line up 2–3 committed answerers from the attendee base (respond within 24h).
Confirm public read (crawlable) + gated write (attendee-only).
Set up backups to the scoped ask-circa-backups bucket.
Start the 60-day test: % of questions answered within 24h.

One Thing to Act On: Before launch, pre-seed 20–30 real Q&As and confirm 2–3 attendees will answer within 24h. A Q&A site with unanswered questions is worse than no site — the seed and the answerers are the launch, not the install.

answer-fit-decision