|
| 1 | +# Goal 0047 — Node audience/complexity facet |
| 2 | + |
| 3 | +Owner-proposed 2026-08-13: categorize nodes by target user audience. |
| 4 | +Owner's framing sharpened in the same conversation: the market splits |
| 5 | +at the PRODUCT level (non-devs converge on Zapier-class tools, |
| 6 | +engineers on n8n-class ones) — a single-experience product loses one |
| 7 | +audience entirely. Mill already carries both altitudes structurally: |
| 8 | +the Quick Panel / seeded-examples / approve-deny surface is the |
| 9 | +consume-and-run altitude, the canvas / MCP / code-exec surface is the |
| 10 | +authoring altitude. This goal's job is to keep ONE product serving |
| 11 | +both (progressive disclosure bridging them), never to pick a side via |
| 12 | +labels — the positioning consequence belongs in SPEC's audience |
| 13 | +section at pickup. |
| 14 | +Session recommendation (to confirm or override at pickup): keep |
| 15 | +FUNCTION as the palette's primary grouping (goal 0001 design wave 3's |
| 16 | +9-group IA) and make audience/complexity a FACET, not a parallel |
| 17 | +taxonomy — the precedent pattern is progressive disclosure ("show |
| 18 | +advanced"), not "for business users" labels, which age badly and |
| 19 | +misfit real people who straddle audiences. The enterprise-shaped |
| 20 | +second half of the idea is governance, not display: WHO MAY USE a |
| 21 | +node class (code-execution, raw HTTP, MCP) is a guardrail/policy |
| 22 | +question — deferred until multi-user exists, but the same metadata |
| 23 | +enables it. |
| 24 | + |
| 25 | +## Plan (three tiers, only the first two in this goal) |
| 26 | + |
| 27 | +1. `NodeType` gains an audience/complexity metadata field (exact |
| 28 | + name/enum decided at pickup — e.g. `Complexity: basic|advanced`), |
| 29 | + set for every registered node, enforced by the existing |
| 30 | + node-standard machine checks (`TestNodeTypes`) so a new node can't |
| 31 | + omit it. `.claude/rules/node-standard.md` gains the item. |
| 32 | +2. Palette + ⌘K surface it as progressive disclosure (an |
| 33 | + "advanced" filter/toggle or search facet) — UX shape decided |
| 34 | + against the live palette, not speculated; ship only if the node |
| 35 | + count/scan cost justifies it at build time (anti-proliferation: |
| 36 | + the metadata is cheap; the toggle must earn its pixels). |
| 37 | +3. DEFERRED with trigger: policy-gated node availability per |
| 38 | + user/role — when Mill has multi-user or admin-managed |
| 39 | + distribution; the field from (1) is its enabler. Recorded here so |
| 40 | + it isn't re-invented. |
| 41 | + |
| 42 | +## Acceptance (checkable) |
| 43 | + |
| 44 | +- [ ] Every registered NodeType carries the new field; TestNodeTypes |
| 45 | + fails on omission; node-standard.md documents it. |
| 46 | +- [ ] The palette/⌘K facet shipped OR explicitly rejected-with-reason |
| 47 | + here after checking against the live palette's actual scan |
| 48 | + cost. |
| 49 | +- [ ] SPEC.md's node/palette section updated in the same change. |
0 commit comments