Repair repo face swarm speech

notes/fresh-workspace-handoff.md

@@ -164,7 +164,7 @@ Git history and smoke artifacts carry routine proof. `state/evidence.jsonl` shou
 
 ## Likely Next Bounded Move
 
-- Keep the repo Face swarm paused until the heartbeat timeout/cadence boundary is repaired. On 2026-06-09, a health pass cleared stale `GameCult Local Orchestrator` locks, restarted the VoidBot stack so bot/worker `dist` entrypoints were fresh, verified Postgres/Qdrant/Ollama/Discord/local backup health, and regenerated the swarm dashboard/CultMesh snapshot. Scheduled repo-face-heartbeats then repeatedly exceeded the orchestrator window and killed the `gpt-5.4` state-projector child, so `state/agent-swarm-paused.json` is set to `"paused": true`. Do not thaw it until state projection has a bounded turn budget, cheaper/deferred path, or other coherent timeout fix, and a scheduled pulse proves clean.
+- The repo Face swarm is unpaused again as of 2026-06-09. The repair isolated the expensive `gpt-5.4` state-projector path behind local `REPO_FACE_STATE_PROJECTOR_ENABLED=false`, limited local heartbeat selection to one job per tick, pointed `REPO_FACE_HEARTBEAT_DEFAULT_CHANNEL_ID` back at Aquarium (`1501196543150264332`), and disabled idle cooling for the wake test. Epiphany posted in Aquarium through the registered webhook, so the room is no longer silent. Keep watching scheduled `GameCult Local Orchestrator` pulses before raising cadence or re-enabling the state projector.
 - The 2026-05-22 live retry confirmed role/name mention obligations can wake multiple newborn Faces from one human intro. Mimir, Huginn, Weksa, and Kiko all posted through the repo identity webhook path. If a Face speaks with its name but no avatar, inspect `.voidbot/private/repo-discord-identities.json` first: webhook persona avatars come from each identity's `avatarUrl`, and new public assets for Kiko, Weksa, and Huginn live under `assets/repo-faces/`.
 - Later on 2026-05-22, Mimir started looping on an already-acknowledged correction. Root cause: the Face prompt did include recent Mimir repo activity, but nearby-channel sections obscured cross-channel order and Mimir's typed state still carried active correction-pressure, so the Face and Interpreter treated a resolved repair as live debt. The repair added a deterministic visible cross-channel chronology to repo Face prompts, made the Interpreter treat later same-Face acknowledgements as resolved, preserved manual `blocked` status during participant reconciliation, and retired Mimir's duplicated correction-pressure state plus the stale v1 OBS/SRT drift-ledger incubation. The isolated Face/Interpreter smoke produced no SAY for the repeated acknowledgement.
 - Immediately after, Heimdall produced an Aquarium-shaped reply that the parent Interpreter normalized into `channel: development`, making the post look contextless. Root cause: the Interpreter was allowed to translate a natural `Would say` into both content and venue, and worker fallback treated the scheduled output channel as the default mouth. The repair makes conversation channel ownership deterministic: `SAY` defaults to `current_room`, `current_room` resolves to `job.guildContext.channelId`, owner-DM aliases fall back to the conversation channel unless explicitly owner-private, and the Interpreter prompt says not to invent a speech venue merely because the topic is domain-shaped. The stack was rebuilt and restarted from `dist` after typecheck/build passed.
@@ -175,7 +175,7 @@ Git history and smoke artifacts carry routine proof. `state/evidence.jsonl` shou
 - The overnight 2026-05-23 run showed a softer version of the previous loop: Faces did not collapse into nonsense, but the Aquarium room stayed captured by the Nibu-origin/faction lore seam for hours. Existing `saturationScore` lived on private thought/incubation surfaces, and the repetition prompt mostly caught copied openings/endings, not semantic topic attractors. The heartbeat prompt now projects current-room topic saturation from repeated raw-transcript terms and estimates whether the attractor is home-adjacent or another steward's gravity well from each Face's identity/jurisdiction text. The state projector should turn foreign attractors into character-specific crowding, neglect, boredom, rivalry, handoff, or pivot pressure; the Interpreter diagnoses non-owner same-topic politeness as missing personal stake, ownership, freshness, handoff, or pivot rather than applying a topic ban or token blacklist.
 - Void's broader moderation authority had become venue leakage: recent-history projection stripped `channelId`, leaving the configured `publicSpeechTarget` as the obvious place for replies, so Aquarium-shaped talk could land in `#general`. The rumination context now preserves message `channelId` and projects `recentConversationTarget`; prompt doctrine says source-channel replies belong in that source channel, while `publicSpeechTarget` is for unanchored herald/artifact announcements.
 - The Nibu/Void turf-crossing discussion exposed a social-state projection gap: existing bonds/status reads project well once typed state already contains them, but the live pipeline did not own "this recent event should become relationship/status pressure." The repair adds bounded raw relationship-pressure evidence to the state packet before the projector: recent messages that touch the Face, peer Faces, existing social targets, jurisdiction/domain terms, or status/territory/consultation language. The projector may turn that evidence into tentative felt pressure, and the Interpreter may persist a bond/status/mood note only when the projected memory or Face output shows a character-specific reaction. This should let resentment, rivalry, consultation offense, and territorial tension emerge without manual state edits.
-- The swarm is currently paused through `state/agent-swarm-paused.json` with `"paused": true`. It had been restarted deliberately on 2026-05-24 after global doctrine inheritance landed, but the 2026-06-09 health pass showed scheduled Face turns can overrun the orchestrator window inside the `gpt-5.4` state projector. Treat this as an ownership/cadence problem, not a prompt-polish problem: repair the heartbeat budget/projection path before resuming ordinary turns.
+- The swarm brake currently reads `state/agent-swarm-paused.json` with `"paused": false`. If scheduled Face turns start overrunning again, use that brake rather than letting the orchestrator churn. The known 2026-06-09 timeout source was the model state-projector child; local runtime currently keeps it disabled while the deterministic fallback and Interpreter/worker speech rails are watched.
 - Direct Face `UPDATE REQUEST` and `BIFROST TOPIC` packets are ignored while `REPO_FACE_BIFROST_ENABLED=false`. Do not debug missing Bifrost jobs as a live failure unless that flag has been deliberately enabled again.
 - Live model-facing prompt prose has started moving out of executable code. The shared loader is `packages/shared/src/prompt-template.ts`; extracted prompt surfaces live under `prompts/`, including repo Face turn/reviewer prompts, provider wrapper prompts, situational social read prompts, source-grounding reminders, consensus-feed prompts, and repo Face model-smoke scenarios. Code should project data and select prompt names; prompt files should own behavioral prose.
 - Identity collapse has started. Active bot, worker, MCP repo-identity tools, and Face CTB scheduler paths now load through the canonical Face registry and treat repo Discord identities as compatibility projections. Next bounded move is to finish removing remaining direct conceptual dependence on repo identities, then split the Face turn loop into scheduler, context assembler, prompt renderer, candidate parser, policy gate, and transport adapter.