feat(coach): expose structured snapshot sections + tool-trace provenance

The Coach snapshot builder returns the structured, post-degrade record it
serialises to JSON alongside the JSON string, so a single domain block can be
read out without re-parsing. Add an optional tool-call trace to the provenance
envelope (which retrieval tools ran + whether each found data) and mirror it in
the OpenAPI provenance schema.

Groundwork for on-demand retrieval: the retrieval tools slice one section out
of this record, reusing the builder's gates and I/O verbatim.

src/lib/ai/coach/snapshot.ts

@@ -83,6 +83,16 @@ import type {
 
 export interface CoachSnapshotResult {
   snapshotJson: string;
+  /**
+   * v1.20.0 (F1) — the structured, post-degrade snapshot record `snapshotJson`
+   * is serialised from. Keyed by domain block (`bloodPressure`, `weight`,
+   * `pulse`, `mood`, `compliance`, `glucose`, `sleep`, `sleepRhythm`, `labs`,
+   * `illness`, `derived`, `dayStrain`, `trajectory`, `weeklyContext`, …) plus a
+   * `scope` block. The coach tool executor reads a single domain block out of
+   * this so an on-demand retrieval tool returns exactly the numbers the legacy
+   * snapshot-stuffing path would have shown — identical builder, gates, and I/O.
+   */
+  sections: Record<string, unknown>;
   /**
    * Provenance built from snapshot keys actually present. Stays in
    * sync with the SNAPSHOT block so the source-chip row mirrors what
@@ -2352,6 +2362,13 @@ async function buildCoachSnapshotImpl(
 
   return {
     snapshotJson: JSON.stringify(compactSnapshot, null, 2),
+    // v1.20.0 (F1) — the structured, post-degrade snapshot record keyed by
+    // domain block (`bloodPressure`, `glucose`, `labs`, …). The coach tool
+    // executor slices a single domain block out of this so a retrieval tool
+    // returns exactly the numbers the legacy snapshot path would have shown —
+    // same builder, same gates, same I/O. `snapshotJson` is this record
+    // serialised; exposing the record avoids re-parsing it.
+    sections: compactSnapshot,
     provenance: {
       windows: Array.from(windows),
       metrics: Array.from(metrics),

src/lib/ai/coach/types.ts

@@ -338,6 +338,15 @@ export interface CoachProvenance {
    * on a conversation reload. Absent on turns that carry no suggestion.
    */
   suggestion?: CoachSuggestion;
+  /**
+   * v1.20.0 (F1) — the retrieval-tool trace for this turn: which tools the
+   * Coach called and whether each found data (`present`). Metadata only — no
+   * raw values beyond what `keyValues` already persists. Lets a conversation
+   * reload show "what I looked at" and lets the hallucination audit replay
+   * grounding. Absent on the legacy snapshot path + on turns that called no
+   * tools.
+   */
+  toolCalls?: ReadonlyArray<{ name: string; present: boolean }>;
 }
 
 /**

src/lib/openapi/routes/coach.ts

@@ -277,6 +277,17 @@ const coachProvenanceSchema = z
       .describe(
         "Load-bearing numbers the assistant surfaced, rendered in the collapsible evidence block. Hard-capped at 8 entries.",
       ),
+    toolCalls: z
+      .array(
+        z.object({
+          name: z.string(),
+          present: z.boolean(),
+        }),
+      )
+      .optional()
+      .describe(
+        "v1.20.0 — the retrieval-tool trace for this turn: which tools the Coach called and whether each found data. Metadata only (no values). Absent on the legacy snapshot path and on turns that called no tools.",
+      ),
   })
   .meta({
     id: "CoachProvenance",