feat: graph context injection — the session's prior knowledge, steered in at session start (vision-context-injection.md)

One shot per chat, fired at the first turn's onFinish: the session's
topic (its only typed message) is embedded as an EmbeddingGemma retrieval
query; the semantically closest graph propositions are selected under the
session's token budget (pure budget, no similarity threshold — the
decided doctrine) and delivered as a marker-only steer through the wake
pipeline, queuing behind pending questions like every sub-agent result.

- contracts: ChatSettings.graphContextTokens (default 1024, max 4096,
  step 128, 0 = off); GraphContextData + graphContextText; the
  graph-context JobKind + buildGraphContextSteer (id steer_ctx_<chatId>)
- core/graph-context.ts: brute-force cosine top-K, greedy chars/4 fill
  under budget, most-recent UnderstandingLevel fold, count cap 64,
  parse-don't-cast at the marker seam
- server: embedQuery (role-asymmetric query format, same warm singleton
  — vector-space parity by construction); the runner (startGraphContext-
  Injection one-shot gates; runGraphContextJob validates ledger parity);
  wake branches (chatIdOf/jobIdOf, finalizeJob done, recoverStaleJobs
  re-deliver or contextRespawn); abortJob kills context jobs SILENTLY
  (nothing was promised in the lesson); turn-service trigger seam
- adapter: data-graph-context flattens to the canonical block (malformed
  parts have no model face)
- UI: the budget slider (home dock + /chat/new draft, the session-data
  strip); the quiet SystemLine "~537 · 10 propositions"; dock kind
  "Knowledge context"; model-picker PUT now spreads settings (a model
  pick can no longer clear the sibling override)

Live-smoked end to end on the real embedder + real GLM: seeded ledger
(10 propositions), session on "modal logic: possible worlds and the
necessity box" — job in the dock, queued behind the prober's question,
delivered on the answer, SystemLine in the transcript, wake turn reads
the block and continues the probe. Gates: tsc, lint, 459/459, build.

components/chat.tsx

@@ -21,7 +21,8 @@ import { SubagentSteerCard } from "@/components/artifacts/steer-card";
 import { Composer } from "@/components/composer";
 import { JobsContext } from "@/components/jobs-context";
 import { StatsBar } from "@/components/stats-bar";
-import type { ChatSettings } from "@/src/core";
+import { graphContextOfMessage, type ChatSettings } from "@/src/core";
+import { formatTokens } from "@/components/usage-format";
 import { PlanBlock, PlanPopup, latestTree } from "@/components/plan-block";
 import { PlanMarginalia } from "@/components/plan-window";
 import { WaypointsIcon } from "lucide-react";
@@ -81,6 +82,16 @@ const useIsoLayoutEffect = typeof window === "undefined" ? useEffect : useLayout
 
 /** A wake-pipeline steer message: user-role delivery whose metadata marks it
  * as a sub-agent result — rendered as an artifact card, never a bubble. */
+/** The knowledge-context card's one-liner: the numbers the observability
+ *  doctrine wants visible at the seam — tokens and proposition count
+ *  (parse-don't-cast; a malformed part still renders a quiet generic line). */
+function graphContextLine(message: { parts: ReadonlyArray<{ type: string; data?: unknown }> }): string {
+	const data = graphContextOfMessage(message);
+	return data
+		? `Knowledge context — ~${formatTokens(data.tokens)} · ${data.propositions.length} propositions`
+		: "Knowledge context";
+}
+
 /** The quiet system line (custom-message scheme rendering): centered, muted,
  *  small — the transcript's voice for events that are not speech (wake
  *  scaffolding, the learner's plan verdict). */
@@ -540,7 +551,14 @@ export function Chat({
 									// PROTOCOL" §4). Key fallback: legacy transcripts (pre id-hygiene)
 									// can still hold empty ids — the index keeps keys unique for them.
 									<SubagentSteerCard key={m.id || `m-${mi}`} message={m} />
-								) : m.role === "user" && m.parts.some((p) => p.type === "data-probe-handoff") ? (
+								) : m.role === "user" && m.parts.some((p) => p.type === "data-graph-context") ? (
+								// The prior-knowledge injection's delivery (vision-context-injection.md):
+								// quiet system line, the numbers only — the block itself was for the
+								// model (adapter flattening); the part data stays the audit trail.
+								<SystemLine key={m.id || `m-${mi}`} testId="graph-context">
+									{graphContextLine(m)}
+								</SystemLine>
+							) : m.role === "user" && m.parts.some((p) => p.type === "data-probe-handoff") ? (
 								// The probe→teach handoff wake (custom-message scheme): quiet
 								// system line, never a user bubble — the wake is scaffolding,
 								// the "Planning" divider marks the seam, and the model-facing