Proposal 077: Swarm Broadcast Assistance¶

Based on:

doc/project/20-memos/swarm-broadcast-assistance.md (promoted memo)
doc/project/20-memos/whisper-corpus-composition.md
doc/project/40-proposals/003-question-envelope-and-answer-channel.md
doc/project/40-proposals/013-whisper-social-signal-exchange.md
doc/project/40-proposals/039-crisis-space-seed-v1.md
doc/project/40-proposals/066-inquirium-assistant-channel.md
doc/project/40-proposals/069-corpus.md
doc/project/40-proposals/073-agent-orchestration-organ.md

Status¶

Draft

Date¶

2026-07-03

Executive Summary¶

A node user may open a communication window to the swarm and broadcast an issue that matters to them — a personal problem, a technical blockage, a safety concern, or a request for orientation. The swarm treats the broadcast as a shared assistance request, not a point-to-point message: it may gather relevant context, route sub-questions to capable nodes, hire a Corpus deliberation, request clarification, and converge on a helpful response or action plan — synthesized from many partial contributions with traceable sources, explicit uncertainty, and a clear distinction between facts, hypotheses, and recommendations.

This proposal promotes the assistance memo into an architectural direction by grounding it in mechanisms that now exist: the P003 question-envelope / answer-channel transport, Corpus procurement and deliberation (P069), bounded Agent sessions (P073), the Inquirium assistant channel with its epistemic and mode-rigor guarantees (P066), and the crisis boundary (P039, P066 assistant-crisis-candidate).

Assistance is the path Whisper deliberately is not: Whisper correlates weak signals into association; assistance is a scoped, user-initiated call for help. The two compose (an association room may raise an assistance request or hire expertise), but they never merge.

Context and Problem Statement¶

The primitives for collective help exist, but no contract ties them into an assistance flow: P003 can carry questions to selected responders, Corpus can convene topic experts, Agent can run bounded unattended reasoning, and the assistant channel can hold the human's hand — yet "I need help with X, swarm" has no envelope, no scope control, no synthesis contract, and no explicit crisis boundary. The memo held the vision; implementation pressure (the Whisper→Corpus composition, the crisis-candidate path) now justifies fixing the semantics.

Goals¶

Define the assistance request as a first-class, scoped artifact riding the existing question-envelope family (P003), not a new transport.
Let the user control scope and intensity: private / federation-local / wider; whether follow-up questions are allowed; urgency class.
Define the synthesis contract: multi-source response with traceable contributions, explicit uncertainty, and fact / hypothesis / recommendation separation (inheriting Inquirium's epistemic block semantics).
Define continuity: an assistance case stays open until resolved or safely handed off, with facts recording each transition.
Keep the crisis boundary hard: assistance may nominate a crisis (crisis-candidate), but crisis authority stays host-owned (P066), and data custody under crisis stays with Crisis Space (P039).
Support both sponsored and gift-priced help (price 0 is normatively the gift-economy regime — see the composition memo).

Non-Goals¶

Not an emergency dispatch system: acute personal emergencies prefer the local help-mode path (P013 boundary); assistance is not the default first response to a likely cardiac arrest.
Not a replacement for Corpus: Corpus is one hireable resolution step, not the assistance flow itself.
Not a reputation or scoring system for helpers (see the anti-Goodhart guard in the composition memo: recognition is a trace, not a currency).
Not swarm gossip: requests go to scoped, selected responders (P003), never to "everyone" implicitly.

Proposed Model¶

1. Assistance request envelope¶

An assistance-request.v1 rides the P003 question-envelope family with an assistance profile: stable id, scope (private | federation-local | wider), urgency class, follow-up policy, TTL after which the asker no longer promises to accept late answers, and an optional coarse context/class: community-gift marker (never identifying an association room or its members — the privacy guard from the composition memo).

2. Decomposition and routing¶

The receiving side may decompose the issue and route sub-questions to capable nodes using existing discovery (offer catalog, Corpus topic resolution) and transport (P003 selected-responder, Artifact Delivery). Decomposition and routing run as bounded Agent sessions (P073): human-in-the-loop gates effects, budgets, and acceptance — never individual reasoning turns.

3. Synthesis contract¶

The response need not come from one node. The synthesized answer carries: contribution references (traceable sources), explicit uncertainty, and the fact / hypothesis / recommendation distinction. Stance is never authoritative; plurality is preserved when viable alternatives exist (Inquirium P066 epistemic semantics apply end to end).

4. Continuity and handoff¶

An assistance case is a fact-backed lifecycle: opened, decomposed, responded, clarified, resolved, or handed off (to a human, a service, or the crisis path). Every transition leaves a fact; silent closure is not a state.

5. Crisis boundary¶

If assistance processing detects crisis indicators, it emits a crisis-candidate to the host crisis layer (P066) — it never declares crisis itself. Crisis data custody follows Crisis Space (P039). Assistance may continue alongside a crisis as its network-facing support surface, but crisis authority and assistance choreography never merge.

6. Economic modes¶

A request may be answered under sponsorship (an asker or supporter contracts paid steps, e.g. a Corpus procurement) or as gift work (zero-price offers). Providers may position themselves for community work in the offer catalog; the concrete offer-field shape lands in the owning contracts when picked up.

Trade-offs¶

Reusing the P003 question-envelope family avoids a new transport, but it means the assistance-specific semantics must be explicit as a profile, lifecycle, and synthesis contract rather than hidden in delivery mechanics.
Calling this a "broadcast" gives users a clear mental model of asking the swarm for help, but the protocol must keep "broadcast" scoped: selected responders, signed scope, and TTL are part of the contract, not optional operator etiquette.
Letting Agent sessions decompose and route work makes assistance useful before every step is human-driven, but effect commitment, budgets, and final acceptance stay behind human-in-the-loop gates.
Sponsor-as-asker is the simplest MVP identity model for association-room requests. A room-scoped asker identity is more symmetric, but it belongs after association-room lifecycle and room membership semantics are concrete.

Failure Modes and Mitigations¶

Failure	Risk	Mitigation
Assistance used as emergency dispatch	Delayed real-world response	Hard non-goal; urgency class routes acute cases to the local help-mode path; crisis-candidate to host layer.
Scope creep of a broadcast	Sensitive issue leaks wider than intended	Scope is part of the signed envelope; responders outside scope refuse; TTL bounds late spread.
Community marker deanonymizes a group	Association-room existence leaks	Marker stays coarse (`community-gift`), optional, never carries room/topic/member identity.
Helper reputation farming via cheap tasks	Gift economy corrupted by Goodhart	No fungible score; recognition only as signed, non-aggregable acknowledgment facts.
Unattended agents commit effects	Autonomy exceeds budgeted trust	P073 bounded contracts; HIL gates effects/budgets/acceptance; deliberation output is a signed answer, not an action.

Open Questions¶

No unresolved questions remain for this proposal's MVP contract.

~~Does assistance-request.v1 need its own schema, or is a P003 question-envelope profile (like selected-responder) sufficient for MVP?~~ Resolved: MVP uses a P003 question-envelope assistance profile. A separate assistance-request.v1 may be introduced only after the lifecycle diverges enough to need its own contract family.
~~What is the minimal case read-model (open assistance cases, states, handoffs) and where does it live — assistant channel surface or a dedicated host capability?~~ Resolved: the minimal assistance case read-model lives behind a dedicated host capability. Assistant Channel may consume and present it, but does not own the assistance state.
~~How does an association room formally raise an assistance request — sponsor-as-asker only (MVP), or a room-scoped asker identity later?~~ Resolved: MVP uses sponsor-as-asker. Room-scoped asker identity is a later extension once association-room identity surfaces are stronger.

Implementation Tracker¶

Status values: todo, in-progress, partial, done, deferred.

ID	Task	Status	Notes
P077-001	Define the assistance profile on the P003 envelope family	todo	Scope, urgency, follow-up policy, TTL, coarse `context/class` marker; MVP uses a P003 `question-envelope` profile rather than a separate `assistance-request.v1` schema.
P077-002	Assistance case lifecycle facts and read-model	todo	opened/decomposed/responded/clarified/resolved/handed-off; every transition is a fact.
P077-003	Crisis-candidate emission from assistance processing	todo	Rides P066 `assistant-crisis-candidate`; never declares crisis; host crisis layer owns status.
P077-004	Corpus-as-resolution-step choreography	todo	Side-room first, per the composition memo; signed answer posted back into the case.
P077-005	Gift/sponsored economic modes on assistance steps	todo	Price 0 = gift regime; provider community-affinity positioning in the offer catalog.

Next Actions¶

Define the P003 question-envelope assistance profile and schema-gate expectations.
Define the dedicated host-capability assistance case read-model.
Wire the advisory cross-references (013, 039, 069, 073) — done at promotion time.
Revisit after the Whisper association-room lifecycle slice lands, so the room→assistance path can grow beyond sponsor-as-asker when needed.