03 · KNOWLEDGE
services/
wiki/services/{service-name}.md
One page per service. The "what calls what" view — upstream, downstream, known failure modes, and the running list of incidents the service has been involved in.
A service page documents its callers, its callees, and how it tends to break.
What a page looks like
--- type: service name: checkout-api status: reviewed owner: payments-team upstream: [user-service, inventory-api] downstream: [orders-mart, fraud-scoring] incidents: [2026-03-14-checkout, 2026-04-12-checkout] tags: [tier-1, customer-facing] last_updated: 2026-04-26 --- ## Purpose Owns the customer-facing checkout flow … ## Dependencies Calls [[user-service]] for auth, [[inventory-api]] for stock. Read by [[orders-mart]] and [[fraud-scoring]] … ## Known failure modes - [[connection-pool-exhaustion]] — 2 incidents - [[downstream-slow-call]] — 1 incident ## Incident history - [[2026-04-12-checkout]] — SEV2, 31min - [[2026-03-14-checkout]] — SEV2, 47min
Sections — in order
- Purpose — what this service does in one paragraph.
- Dependencies — what it calls, what calls it. Every named service is a wiki-link.
- Known failure modes — bullet list linking to
root-causes/pages. - Incident history — newest first.
- Owners — link to the owning team's page.
- Notes — open questions, runbooks, anything else.
Rules
- File name is the service slug in kebab-case, e.g.
checkout-api.md. - Type-specific frontmatter:
upstream,downstream,incidents. - Every service named in
upstream/downstreammust have a page or a stub. - Known failure modes link out to
root-causes/— don't restate the cause inline. - Unknown owner? Write
owner: unknownand flag it under Notes.