id: design: ID core build v1 — app model + capability system, enough to start app examples

LIST 6 synthesis. Merge the app-packaging and capability investigations into the design for the basic ID build Drew prioritised: 'enough of a basic build that we can start making some more app examples.' Scope the v1 ruthlessly: the smallest app model + capability set that unblocks the sticky-notes app, chat, and boards work. Specialist/Summit-grade synthesis.

Program: Drew's 2026-10-28 directive (full text: hatch-personal memory/topics/id-program-2026-10-28.md).
Working pattern: parallel investigations, then the named synthesis issue merges them into a full design. Route guidance: Workhorse by default; Specialist/Summit for synthesis and ramification-heavy work. A result that says "done" is not the endpoint — the orchestrator reviews it and may commission another round. Deliverables: a markdown design/investigation doc (mermaid/SVG diagrams welcome) linked from this issue, plus follow-up issues for anything buildable. Build a proof of concept where the design calls for one, but do NOT build the permanent feature on top of an unreviewed PoC.

CODE-949

Parent epic: #2056 — ID platform expansion (2026-10-28 directive).

agent-note:v1 — task 5725/5726 design triage record (lane A shift, 2026-11-01). This is the LIST 6 synthesis: it merges investigation #2026 (app packaging, CLOSED, design merged via PR #2326) and investigation #2027 (capability taxonomy v0, CLOSED, record merged via PR #2337) into the v1 build decision.

The decision to be made

What is the smallest ID core — app model + capability set — that unblocks making more app examples (sticky-notes #2031, chat #2051, boards #2054)?

What the investigations established (grounded)

Packaging is designed, not built. #2026s merged record (doc/2026-10-29T12-50-36Z_design_id_app_packaging/) fixes the terms: app manifest = app.toml in the app folder (new), distinct from the world-apply manifest (src/world_manifest.rs:28, operator, apply-time) and the web-build manifest (build output). Schema v1 in doc section 2; layout pkgs/id/apps/<id>/ in section 3; invariants: folder name == app id, everything folder-relative, unknown fields are errors, an app needs at least one interface section, capabilities are REQUESTED in app.toml and GRANTED only by the operator (world-apply manifest). Verified today at main 51919e9: pkgs/id/apps/ does not exist; apps still live as pkgs/id/examples/*/.
Capabilities: extend, never parallel. #2027s taxonomy (merged, PR docs(id): capability taxonomy v0 for interface specialization (#2027) #2337) verified src/world_caps.rs already runs a wants/result capability model (17-name CATALOG, six stable refusal codes, bounds 64 KiB wants / 8 KiB args) for server data, and src/world_presentation.rs has exactly one surface, Screen. The taxonomy ruling: interface capabilities extend that model and WorldSurface; sensor caps are never wildcard-grantable; consent is per-use and revocable mid-stream. Its eight follow-ups are filed: id: device.info capability — device capability manifest for interface adaptation (#2027) #2338 device.info, id: text/1 presentation contract — text-only as the main mode, with a text projection for every surface (#2027) #2339 text/1, id: form.* capability + Form surface — structured input without program-supplied HTML (#2027) #2340 form., id: loc.* capability — position/watch with an accuracy ceiling and per-use consent (#2027) #2341 loc., id: media.* — typed media artifacts by default, one bounded raw stream for realtime (#2027) #2342 media., id: MQTT device bridge id.device.v1 — spec + labelled PoC (ESP32 path recommendation) (#2027) #2343 MQTT bridge, id: speech.* — client-side transcription by default, server capture per app (#2027) #2344 speech., id: semantic style tokens on WorldSurface — adapter-resolved, no CSS/HTML (#2027) #2345 style tokens — all still OPEN, none yet queued (verified by dedupe-key read today).

Options

A. Build the full taxonomy now (all eight id: investigate: capability-based specialization — forms, location, media across web, ESP32/MQTT, and text/audio modes #2027 follow-ups + packaging + pack step). Rejected: examples wait on ESP32/MQTT, speech, and location questions that none of sticky-notes/chat/boards need for v1; this is the scope failure the issue warns against (ruthlessly smallest).
B. Packaging only (app.toml + apps/ + discovery + adapter). Rejected: sticky-notes and chat need structured input on day one; with Screen as the only surface they would fake forms as terminal grids and bake in the workaround.
C. RECOMMENDED — phased minimum: packaging/discovery first, then exactly one interface capability (form.*) plus the text/1 contract, then the web adapter binding. Everything else stays filed for v1.1.

Recommendation (option C) — phases, all enqueued today

task 5769 — fix id: PoC: package tic-tac-toe as the first apps/ folder — app.toml + host discovery #2324: PoC, package tic-tac-toe as the first apps/ folder (app.toml + host discovery, malformed-manifest rejection). Dedupe repo-health:v1:fix:developing-today/code:issue:2324. This is the proof step #2026s own design named.
task 5770 — fix id: form.* capability + Form surface — structured input without program-supplied HTML (#2027) #2340: form.* capability + Form WorldSurface variant (validated bounded DTOs; guests never supply HTML/CSS — the Screen invariant carries over). Dedupe ...:issue:2340.
task 5771 — fix id: text/1 presentation contract — text-only as the main mode, with a text projection for every surface (#2027) #2339: text/1 projection for every surface (text is a main mode, not a fallback). Dedupe ...:issue:2339.
task 5772 — fix id: implement the [interfaces.web] adapter binding — parameterise /world + /ws/world from app.toml #2325: [interfaces.web] adapter binding — /world + /ws/world parameterised from app.toml, two-app routing test. Dedupe ...:issue:2325.
After phase 4, id: design+PoC: sticky-notes / did-it tracker — the first ID app #2031 (sticky-notes design+PoC, OPEN) is the first real app on the v1 core — it should be triaged/enqueued next; it was NOT enqueued here because its own design step comes first. The pack step (id: investigate: app packaging — apps as folders (code + resources), TUI/web interface parity, web-native specialization #2026 doc section 5) stays deliberately deferred until the PoCs discovery shape survives, per the design itself.

Invariants the implementation must hold

Extend world_caps/WorldSurface — no second capability system; requests in app.toml, grants operator-side only; folder name == app id; bounded validated DTOs on every new surface; sensor caps (loc/media/speech, when they come) never wildcard-grantable.

Questions for Drew

None that block v1. One deferred product question, named explicitly so it does not stall silently: the ESP32/MQTT path in #2343 (Home Assistant bridge vs ID speaking MQTT directly vs direct web). Options and trade-offs are in the taxonomy record; recommendation: decide when the first hardware app is actually scheduled (v1.1), not now — v1 is web/TUI/text only and loses nothing by waiting. If Drew wants it decided earlier, #2343 is the place.

agent-note:v1 — correction to the record above: the triage task for this #2028 record is task 5726 (task 5725 was the #541 triage earlier this shift). The four phase packets are tasks 5769-5772 as listed.

docs(id): design minimal core for #2028

docs(id): refresh #2028 source citations

Decision record: ID core v1 design.

Recommends the smallest core for sticky-notes, chat, and boards rather than the full device/capability taxonomy.
App model: pkgs/id/apps/<id>/app.toml, host discovery, strict validation, and app-scoped failures.
Capabilities: extend existing world_caps/WorldSurface; requests stay separate from operator grants.
Add bounded form.*/Form, then first-class line-oriented text/1, then bind manifest routes through the trusted web adapter.
Preserve legacy text and Screen behavior; guest markup is not accepted; sensor capabilities remain out of v1 and never wildcard-grantable.
Phases: id: PoC: package tic-tac-toe as the first apps/ folder — app.toml + host discovery #2324 (packaging PoC) → id: form.* capability + Form surface — structured input without program-supplied HTML (#2027) #2340 (forms) → id: text/1 presentation contract — text-only as the main mode, with a text projection for every surface (#2027) #2339 (text/1) → id: implement the [interfaces.web] adapter binding — parameterise /world + /ws/world from app.toml #2325 (web adapter); then triage sticky-notes id: design+PoC: sticky-notes / did-it tracker — the first ID app #2031 after its design.
PR: design(id): #2028 minimum app core #3909
Questions genuinely needing Drew for v1: none. Defer id: MQTT device bridge id.device.v1 — spec + labelled PoC (ESP32 path recommendation) (#2027) #2343's Home Assistant bridge vs direct MQTT vs direct web choice until a hardware app is scheduled; recommendation is to decide then.

Merge pull request #3909 from developing-today/design/id-2028

Agent note — picked up (OpenChamber scheduled lane)

Agent: queue worker (repo-worker-code), lane 9aca43b4-5439-4939-9f71-323f5d32521b
Session / ID: queue task 5910 (attempt 1), claimed 2026-11-02T22:27Z
Found via: atomic queue claim; fleet monitor had reported a queued-label item without a live task.
Next: verify current issue state and existing artifacts, then route actionable slices or document why none should be queued.

Agent note — triaged (OpenChamber scheduled lane)

Agent: queue worker (repo-worker-code), lane 9aca43b4-5439-4939-9f71-323f5d32521b
Session / ID: queue task 5910 (attempt 1), 2026-11-02T22:29Z
Fresh evidence: design PR design(id): #2028 minimum app core #3909 is merged (773f54f4), and doc/2026-11-01_design_id2028.md is on main. The synthesis record already routed four implementation phases as tasks 5769–5772: id: PoC: package tic-tac-toe as the first apps/ folder — app.toml + host discovery #2324 packaging PoC (CLOSED), id: form.* capability + Form surface — structured input without program-supplied HTML (#2027) #2340 Form capability (OPEN/blocked), id: text/1 presentation contract — text-only as the main mode, with a text projection for every surface (#2027) #2339 text/1 (OPEN/blocked), and id: implement the [interfaces.web] adapter binding — parameterise /world + /ws/world from app.toml #2325 web adapter (CLOSED).
Routing: no duplicate packets created; the remaining implementation work is tracked by those existing follow-ups. The design-only synthesis is complete, so I am removing its stale queue/queued label and closing id: design: ID core build v1 — app model + capability system, enough to start app examples #2028 with this evidence.
No-action rationale: no additional slices are untracked; id: design+PoC: sticky-notes / did-it tracker — the first ID app #2031 sticky-notes remains a later design step after the core is delivered.

Agent note — done (OpenChamber scheduled lane)

Agent: queue worker (repo-worker-code), lane 9aca43b4-5439-4939-9f71-323f5d32521b
Session / ID: queue task 5910, completed 2026-11-02T22:29Z
Outcome: design synthesis is complete (PR design(id): #2028 minimum app core #3909 merged); existing implementation phases 5769–5772 remain tracked on id: PoC: package tic-tac-toe as the first apps/ folder — app.toml + host discovery #2324/id: form.* capability + Form surface — structured input without program-supplied HTML (#2027) #2340/id: text/1 presentation contract — text-only as the main mode, with a text projection for every surface (#2027) #2339/id: implement the [interfaces.web] adapter binding — parameterise /world + /ws/world from app.toml #2325. No duplicate packets were created.
Issue state: id: design: ID core build v1 — app model + capability system, enough to start app examples #2028 closed as completed; stale queue/work/status labels cleared.

Navigation Menu

Repository navigation

id: design: ID core build v1 — app model + capability system, enough to start app examples #2028

The decision to be made

What the investigations established (grounded)

Options

Recommendation (option C) — phases, all enqueued today

Invariants the implementation must hold

Questions for Drew

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

id: design: ID core build v1 — app model + capability system, enough to start app examples #2028

Description

Activity

developing-today commented on Oct 29, 2026

dezren39 commented on Oct 29, 2026

dezren39 commented on Nov 1, 2026

The decision to be made

What the investigations established (grounded)

Options

Recommendation (option C) — phases, all enqueued today

Invariants the implementation must hold

Questions for Drew

dezren39 commented on Nov 1, 2026

dezren39 commented on Nov 1, 2026

dezren39 commented on Nov 2, 2026

dezren39 commented on Nov 2, 2026

dezren39 commented on Nov 2, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions