Remove graph replacement from modeling schema

This commit is contained in:
Meta Vens 2026-05-18 04:43:27 +01:00
commit 0112c88b4e
5 changed files with 18 additions and 6 deletions

View file

@ -241,9 +241,8 @@ pub fn epiphany_role_launch_output_schema(role_id: ThreadEpiphanyRoleId) -> serd
"statePatch".to_string(),
serde_json::json!({
"type": "object",
"description": "Required reviewable thread/epiphany/update patch for modeling. Use only graphs, graphFrontier, graphCheckpoint, scratch, investigationCheckpoint, observations, and evidence. The patch must include at least one durable modeling field, not observations/evidence alone.",
"description": "Required reviewable thread/epiphany/update patch for modeling. Use only graphFrontier, graphCheckpoint, scratch, investigationCheckpoint, observations, and evidence. Graph growth belongs in memoryPatchCandidates, not statePatch.graphs. The patch must include at least one durable modeling field, not observations/evidence alone.",
"anyOf": [
{"required": ["graphs"]},
{"required": ["graphFrontier"]},
{"required": ["graphCheckpoint"]},
{"required": ["scratch"]},
@ -264,6 +263,18 @@ pub fn epiphany_role_launch_output_schema(role_id: ThreadEpiphanyRoleId) -> serd
"additionalProperties": true
}),
);
map.insert(
"memoryPatchCandidates".to_string(),
serde_json::json!({
"type": "array",
"description": "Optional typed EpiphanyMemoryPatchCandidate records for graph or memory growth. Proposed domains, nodes, and edges are reviewed against the memory graph during roleAccept.",
"items": {
"type": "object",
"required": ["id", "profile", "status"],
"additionalProperties": true
}
}),
);
}
required.push("frontierNodeIds");
required.push("statePatch");

View file

@ -403,8 +403,9 @@ growth was accepted or rejected. `thread/epiphany/roleResult` findings also
expose typed `memoryPatchCandidates`, so the operator can inspect proposed graph
growth before acceptance. Legacy Modeling `statePatch.graphs` is now rejected by
policy and the specialist prompt routes graph growth through
`memoryPatchCandidates`; graph replacement remains only a non-normal legacy
surface outside Modeling acceptance. The next cleanup is removing remaining
`memoryPatchCandidates`; the Modeling output schema also no longer advertises
`statePatch.graphs`; graph replacement remains only a non-normal legacy surface
outside Modeling acceptance. The next cleanup is removing remaining
operator/UI assumptions that graph growth comes from thread-state graph
replacement instead of the unified memory graph.

View file

@ -96,7 +96,7 @@ remembers doctrine; the branch remembers the blade.
- The heartbeat initiative pass is landed as a bounded tool seam. `state/agent-heartbeats.msgpack` tracks Self, Face, Imagination, Eyes, Body, Hands, Soul, and Life as Ghostlight-style initiative participants with arena, participant kind, speed, next-ready time, reaction bias, interrupt threshold, load, status, constraints, personality cooldown, mood cooldown, effective cooldown, adaptive pacing, history, and pending turns through `epiphany-core::EpiphanyHeartbeatStateEntry` and the native `epiphany-heartbeat-store` binary. Role dossiers set baseline timing; appraisal mood/anxiety bends it through urgency, arousal, thought pressure, guardedness, and reaction intensity, so Hands-like work pressure and high-need anxious lanes recover sooner without being heartbeaten while a prior turn is running. `epiphany-heartbeat-store pump` computes pressure from external urgency, mood/anxiety, reaction intensity, thought pressure, and current pending load, then chooses tempo plus target concurrency: calm can launch zero and sleep slow; alarm can fill most active lanes. Idle turns are for rumination: light thought shuffling, role-quality attention, and candidate selfPatch pressure. Sleep/dream cycle passes are the intended distillation window for durable self-memory and doctrine. JSON heartbeat state and Python wrapper state are gone; general CultCache schema sync, polyglot domain loading, and debug display tools belong in CultLib. This is a callable scheduler seam, not yet an always-on daemon; a heart valve, not a whole circulatory god.
- The first Ghostlight-derived timing slice is landed in Epiphany, but Ghostlight is reference lineage rather than a sibling runtime to preserve. `epiphany-heartbeat-store init --profile ghostlight-scene --scene-id <id> --scene-participant <id|name|speed|reaction|threshold|constraints>` creates a typed CultCache MessagePack scene heartbeat store whose participants are `arena=scene`, `participantKind=character`; `tick` emits `ghostlight.initiative_schedule.v0` receipts with `scene_turn` actions and local-affordance basis. The generic Epiphany maintenance lanes are not auto-patched into scene stores.
- The first Void-derived routine slice is landed in Epiphany, but VoidBot is reference lineage rather than a runtime dependency. `epiphany-heartbeat-store routine --store .\state\agent-heartbeats.msgpack --artifact-dir .\.epiphany-heartbeats --agent-store .\state\agents.msgpack` reads typed role dossiers, computes bounded memory resonance, maintains incubation themes, runs analytic and associative cognition lanes, writes bridge syntheses/saturation/tension, projects the active thought cluster through each role's Ghostlight-shaped personality vectors, derives participant-local reactions, applies personality/mood timing, advances the sleep/dream cycle, updates the typed heartbeat store, and emits an auditable `epiphany.void_routine.v0` receipt. The routine now also carries anti-loop physiology imported from Void's calmer brain: `noveltyToSelf`, `noveltyToRoom`, source coverage, saturation pressure, refractory cooling, and explicit permission to let a live unsaturated thought deepen without forcing novelty theater every pass. This mutates only heartbeat physiology fields: project truth and role memory mutation remain on their reviewed surfaces.
- The next memory foundation must not build separate engines for repo graph and agent mind. Heartbeat cognition, role self-memory, repo architecture/dataflow, evidence scars, incubation, agency pressure, and context packets should converge through `EpiphanyMemoryGraph` profiles so one typed graph owns anchors, lifecycle, summaries, freshness, context cuts, and Qdrant embedding manifests. Repo, agent, and heartbeat profile producers now map accepted graph truth, reviewed role dossier memories/goals/values, and provisional heartbeat short-term/incubation/agency/candidate pressure into memory graph documents without scanner/model work. Heartbeat import deliberately does not mark durable promotion. Compose now merges profile snapshots into one validated graph and rejects duplicate authority; `epiphany-memory-graph compose` writes a composed typed CultCache graph from typed source graph stores; `epiphany-memory-graph refresh` can invoke live agent-memory plus heartbeat-cognition producers from their typed stores into one graph store; `epiphany-core` has a native typed `EpiphanyThreadState` CultCache store seam; the bridge mirrors loaded Codex-thread Epiphany state into `state/thread-state.msgpack`; CultNet advertises logical `EpiphanyThreadState` plus `EpiphanyMemoryGraph` snapshot contracts; memory graph nodes/edges/summaries project into deterministic typed embedding documents; `epiphany-memory-graph manifest` writes the embedding manifest back to the typed graph store; `epiphany-memory-graph index` fully rebuilds a Qdrant cache collection through the shared semantic cache transport; `epiphany-memory-graph semantic-context` uses ordered Qdrant IDs to seed typed graph context while falling back to graph traversal when the cache is absent; native role worker launch documents now carry typed `memoryContext` instead of full `graphs`; typed runtime role results can persist `memoryPatchCandidates` for graph/memory growth proposals; `epiphany-core` reviews/appends candidate domains/nodes/edges against the typed graph with lifecycle receipts; `epiphany-memory-graph review-candidate` / `apply-candidate` expose that law for typed candidate files; roleResult findings expose typed `memoryPatchCandidates` for pre-acceptance operator review; bridge Modeling `roleAccept` applies runtime-spine memory patch candidates into `state/memory-graph.msgpack` through the same graph law, bootstrapping from accepted thread graphs when needed; roleAccept responses expose typed `memoryPatchReviews`; and Modeling `statePatch.graphs` is rejected by policy/prompt so graph growth uses memory patch candidates. The next clean slice is removing remaining operator/UI assumptions that graph growth comes from thread-state graph replacement instead of the unified memory graph.
- The next memory foundation must not build separate engines for repo graph and agent mind. Heartbeat cognition, role self-memory, repo architecture/dataflow, evidence scars, incubation, agency pressure, and context packets should converge through `EpiphanyMemoryGraph` profiles so one typed graph owns anchors, lifecycle, summaries, freshness, context cuts, and Qdrant embedding manifests. Repo, agent, and heartbeat profile producers now map accepted graph truth, reviewed role dossier memories/goals/values, and provisional heartbeat short-term/incubation/agency/candidate pressure into memory graph documents without scanner/model work. Heartbeat import deliberately does not mark durable promotion. Compose now merges profile snapshots into one validated graph and rejects duplicate authority; `epiphany-memory-graph compose` writes a composed typed CultCache graph from typed source graph stores; `epiphany-memory-graph refresh` can invoke live agent-memory plus heartbeat-cognition producers from their typed stores into one graph store; `epiphany-core` has a native typed `EpiphanyThreadState` CultCache store seam; the bridge mirrors loaded Codex-thread Epiphany state into `state/thread-state.msgpack`; CultNet advertises logical `EpiphanyThreadState` plus `EpiphanyMemoryGraph` snapshot contracts; memory graph nodes/edges/summaries project into deterministic typed embedding documents; `epiphany-memory-graph manifest` writes the embedding manifest back to the typed graph store; `epiphany-memory-graph index` fully rebuilds a Qdrant cache collection through the shared semantic cache transport; `epiphany-memory-graph semantic-context` uses ordered Qdrant IDs to seed typed graph context while falling back to graph traversal when the cache is absent; native role worker launch documents now carry typed `memoryContext` instead of full `graphs`; typed runtime role results can persist `memoryPatchCandidates` for graph/memory growth proposals; `epiphany-core` reviews/appends candidate domains/nodes/edges against the typed graph with lifecycle receipts; `epiphany-memory-graph review-candidate` / `apply-candidate` expose that law for typed candidate files; roleResult findings expose typed `memoryPatchCandidates` for pre-acceptance operator review; bridge Modeling `roleAccept` applies runtime-spine memory patch candidates into `state/memory-graph.msgpack` through the same graph law, bootstrapping from accepted thread graphs when needed; roleAccept responses expose typed `memoryPatchReviews`; and Modeling `statePatch.graphs` is rejected by policy/prompt/schema so graph growth uses memory patch candidates. The next clean slice is removing remaining operator/UI assumptions that graph growth comes from thread-state graph replacement instead of the unified memory graph.
- Manual Codex-run rumination still has an explicit aftercare rule until Epiphany owns the full sleep consolidator. In the intended Epiphany cycle, sub-agents ruminate when idle and distill when they sleep; in this supervising Codex thread, a heartbeat/routine vigil is only physiology until a separate closing pass reviews the receipts and decides whether map, handoff, evidence, or role self-memory should change.
- The Face public-surface pass is landed as a bounded lane. Face's role dossier lives in `state/agents.msgpack`; `epiphany_specialists.toml` gives it a VoidBot-heartbeat-derived prompt stripped of moderation authority; `state/face-discord.toml` and the native `epiphany-face-discord` binary enforce that Face may interact only through #aquarium. Missing channel id or token writes candidate chat artifacts instead of posting elsewhere. If a Face has a persona name/avatar, `post` uses the shared Discord webhook pipe so each Epiphany can speak with its own nickname and avatar without a separate bot identity.
- The Void memory bridge is now native enough to keep the useful organs attached. `state/void-memory.toml` and `epiphany-void-memory` can check Void's Docker Postgres state spine, query Qdrant Discord-history and source/lore collections with the configured Ollama embedding model, and fetch raw archive context windows from Void's file-backed Discord archive. Native `epiphany-mvp-status` includes a `voidMemory` block so Aquarium/backend clients can inspect whether the mouth and memory are actually wired.

Binary file not shown.

View file

@ -297,4 +297,4 @@ current_status:
freshness propagation, summary-versus-child context cuts, and a validated
CultCache store entry; the native `epiphany-memory-graph` binary can
status/validate/query/smoke the store without Qdrant.
next_action: Continue the memory graph build unless a more urgent Codex starvation regression appears. Repo, agent, and heartbeat profile producers now map accepted architecture/dataflow graph truth, reviewed role dossier memory/goals/values, and heartbeat cognition pressure into `EpiphanyMemoryGraph` documents; compose merges profile snapshots into one validated graph; `epiphany-memory-graph compose` writes from typed graph stores; `epiphany-memory-graph refresh` can invoke live agent-memory and heartbeat-cognition producers from their typed stores; `epiphany-core` has a native typed `EpiphanyThreadState` CultCache store seam; the bridge mirrors loaded Codex-thread Epiphany state into `state/thread-state.msgpack`; CultNet advertises logical `EpiphanyThreadState` plus `EpiphanyMemoryGraph` snapshot contracts; memory graph nodes/edges/summaries project into deterministic typed embedding documents; `epiphany-memory-graph manifest` writes the embedding manifest back to the typed graph store; `epiphany-core::semantic_cache` owns shared Qdrant/Ollama transport; `epiphany-memory-graph index` fully rebuilds Qdrant from typed graph documents while keeping the manifest in the graph store; `epiphany-memory-graph semantic-context` asks Qdrant for graph IDs only, preserves hit order, then resolves context from typed graph state with a no-cache fallback; native role worker launches now carry typed `memoryContext` instead of full `graphs`; typed runtime role results can persist `memoryPatchCandidates`; `epiphany-core` can review/apply append-only candidate graph growth with lifecycle receipts; `epiphany-memory-graph review-candidate` / `apply-candidate` expose that law for typed candidate files; roleResult findings expose typed `memoryPatchCandidates` for pre-acceptance operator review; bridge Modeling `roleAccept` applies runtime-spine memory patch candidates into `state/memory-graph.msgpack` through the same graph law; roleAccept responses expose typed `memoryPatchReviews`; and Modeling `statePatch.graphs` is rejected by policy/prompt so graph growth uses memory patch candidates. The next clean slice is removing remaining operator/UI assumptions that graph growth comes from thread-state graph replacement instead of the unified memory graph.
next_action: Continue the memory graph build unless a more urgent Codex starvation regression appears. Repo, agent, and heartbeat profile producers now map accepted architecture/dataflow graph truth, reviewed role dossier memory/goals/values, and heartbeat cognition pressure into `EpiphanyMemoryGraph` documents; compose merges profile snapshots into one validated graph; `epiphany-memory-graph compose` writes from typed graph stores; `epiphany-memory-graph refresh` can invoke live agent-memory and heartbeat-cognition producers from their typed stores; `epiphany-core` has a native typed `EpiphanyThreadState` CultCache store seam; the bridge mirrors loaded Codex-thread Epiphany state into `state/thread-state.msgpack`; CultNet advertises logical `EpiphanyThreadState` plus `EpiphanyMemoryGraph` snapshot contracts; memory graph nodes/edges/summaries project into deterministic typed embedding documents; `epiphany-memory-graph manifest` writes the embedding manifest back to the typed graph store; `epiphany-core::semantic_cache` owns shared Qdrant/Ollama transport; `epiphany-memory-graph index` fully rebuilds Qdrant from typed graph documents while keeping the manifest in the graph store; `epiphany-memory-graph semantic-context` asks Qdrant for graph IDs only, preserves hit order, then resolves context from typed graph state with a no-cache fallback; native role worker launches now carry typed `memoryContext` instead of full `graphs`; typed runtime role results can persist `memoryPatchCandidates`; `epiphany-core` can review/apply append-only candidate graph growth with lifecycle receipts; `epiphany-memory-graph review-candidate` / `apply-candidate` expose that law for typed candidate files; roleResult findings expose typed `memoryPatchCandidates` for pre-acceptance operator review; bridge Modeling `roleAccept` applies runtime-spine memory patch candidates into `state/memory-graph.msgpack` through the same graph law; roleAccept responses expose typed `memoryPatchReviews`; and Modeling `statePatch.graphs` is rejected by policy/prompt/schema so graph growth uses memory patch candidates. The next clean slice is removing remaining operator/UI assumptions that graph growth comes from thread-state graph replacement instead of the unified memory graph.