reward: a seed rule nobody edited follows the seed
reward
The rewards rules engine. Programs of rules, claims that staff verify, an
append-only points ledger, levels, badges, budgets and caps. admin.hanzo.ai
configures it; cloud feeds it events and mounts it as the reward capability at
/v1/reward.
e, err := reward.Mount(app, reward.Config{
DB: db, // hanzoai/orm: the shared SQL, or SQLite
Credits: credit.Client{Base: commerceURL}, // commerce's POST /v1/billing/credit
Mail: sender, // optional
})
How it works
A rule reads one event from the typed catalog, tests its fields, and says
what an approved claim does: points, a badge, credit, an email. An event arriving at
Evaluate makes a claim for every active rule it satisfies, inside the rule's
limits and the program's budget. A manual rule (the default) waits for staff, who
Approve, Reject or Reverse it, each audited. Nothing is edited: claims, steps,
ledger rows and badge rows are written once under ids that are hashes of what they
are about, so every act is exactly-once on SQLite and on the shared SQL alike.
| Package | Is |
|---|---|
catalog |
the typed events, under five pillars: use, share, community, contribute, expert |
rule |
rules, and the program that holds them with budget, levels, badges, templates |
claim |
a claim, and the steps that decide it |
ledger |
the append-only points ledger, legacy backfill, the trail of staff acts |
engine |
Evaluate, Approve, Reject, Reverse, Backfill, the configuration |
profile, emblem |
public profiles, and the live README badge |
credit |
the client for commerce's one credit path |
http |
the routes |
See LLM.md for the design and the routes.
License
MIT OR Apache-2.0.