Welcome. You're joining a team building Token Factory — read
what that is first (5 minutes), then come back here.
¶ Day 1: understand the shape
- Read What is Token Factory? and the
system architecture.
- Skim the Enterprise MVP — it's the
4-week sprint most engineering energy goes into right now — and the
NeoCloud edition for where it all goes.
- Open the status board. That's the honest picture of done vs not.
- Read the ADRs — five short
documents that answer 80% of "why did we…" questions.
Clone the repos from the repo map. The three
amazee.ai repos are running production systems — treat their main/prod
branches accordingly.
Then follow local development to get the
existing gateway stack running on your machine — it's the fastest way to
feel what the control plane does: create a team, issue a key, watch spend
appear.
- Two tracks, one platform. A Mirantis enterprise team is driving a 4-week
MVP for channel partners (Accenture, Dell); the NeoCloud track builds toward
IREN. The editions model is how we keep them from
forking. When in doubt: build the capability, package per edition.
- What's already built. The gateway/governance layer is shipped software
(amazee.ai) with paying customers — we extend it, we don't rewrite it.
- What's genuinely new. The runtime
and registry pillars are
greenfield, and where most hiring is happening.
- The one non-negotiable: nothing talks to an inference engine except
through the gateway. No key, no tokens.
- Feature statuses live in
src/data/capabilities.ts on this site — update
them in the same PR that changes reality. "In progress" means someone is
actively working on it now — a ticket merely existing does not.
- Work is tracked in Jira (project TKF); the site links tickets from the
status board.
- Architecture changes get an ADR — short, in
docs/decisions/.
See Team & Ownership for the current owner table,
the Jira board (TKF),
and the access checklist. Short version: runtime → Dmitrii Shishliannikov /
Ryan Zhang; everything else → Michael Schmid.