fix(turn-metrics): exact per-turn placement via turnId, decode-only TPS, step-level retention

Placement
- Chunks carry their producing turn: StoredChunk.turnId (backend [email protected]) for
  committed rows (persisted in the IDB cache; old cached rows are upgraded on
  sync) and the live event's turnId for provisional ones (the optimistic echo is
  tagged on turn-start).
- interleaveTurnMetrics attributes every group to its turn (turnId, else the
  turn owning its stepId) and puts the turn total after the turn's LAST visible
  group. No user-message counting: sending, streaming, steering, images, errored
  turns and chat-limit unloading can no longer shift metrics between turns — at
  any point during a turn, not just after it ends. A turn cut in half by the
  chat limit keeps its total under its visible tail.
- Legacy rows (no turnId) fall back to an order-preserving, count-guarded match
  that leaves a turn blank rather than show another turn's numbers.
- A turn that errors re-syncs metrics (the backend now persists them).

TPS
- Step TPS = outputTokens / decodeMs only (no genTotalMs fallback — it includes
  time-to-first-token). Turn TPS sums tokens and decode time over timed steps
  only. Tool execution is excluded by the backend.

Display
- Retention = this turn's first-step cache reads / previous turn's last-step
  prompt (step-level; turn sums made it pin at 100%).
- Turn row shows output tokens + input processed (with cached), instead of a
  misleading 'total tok' that summed every step's full prompt.

1 files changed, 15 insertions, 7 deletions

diff --git a/src/features/conversation-cache/logic.ts b/src/features/conversation-cache/logic.ts
index 3cb23e2..344a061 100644
--- a/src/features/conversation-cache/logic.ts
+++ b/src/features/conversation-cache/logic.ts
@@ -5,27 +5,35 @@ import type { ConversationCacheIndexEntry, ReconcileResult } from "./types";
  * Merge authoritative seq-keyed chunks with cached chunks.
  *
  * Deduplicates by `seq`, produces seq-monotonic order.
- * `toAppend` = the incoming chunks whose `seq` is not already in `cached`
- * (exactly what to persist). Idempotent; tolerant of out-of-order/overlapping `incoming`.
+ * `toAppend` = the incoming chunks whose `seq` is not already in `cached`,
+ * plus cached rows UPGRADED with a `turnId` the incoming copy carries and the
+ * cached copy lacks (exactly what to persist — the store upserts by seq). Idempotent; tolerant of out-of-order/overlapping `incoming`.
  */
 export function reconcileCache(
   cached: readonly StoredChunk[],
   incoming: readonly StoredChunk[],
 ): ReconcileResult {
-  const seen = new Set<number>();
+  const bySeq = new Map<number, StoredChunk>();
   for (const chunk of cached) {
-    seen.add(chunk.seq);
+    bySeq.set(chunk.seq, chunk);
   }
 
   const toAppend: StoredChunk[] = [];
   for (const chunk of incoming) {
-    if (!seen.has(chunk.seq)) {
+    const have = bySeq.get(chunk.seq);
+    if (have === undefined) {
       toAppend.push(chunk);
-      seen.add(chunk.seq);
+      bySeq.set(chunk.seq, chunk);
+    } else if (have.turnId === undefined && chunk.turnId !== undefined) {
+      // Upgrade a cached row written before the server stamped `turnId`
+      // (seqs are immutable, so only this provenance field can differ).
+      const upgraded = { ...have, turnId: chunk.turnId };
+      toAppend.push(upgraded);
+      bySeq.set(chunk.seq, upgraded);
     }
   }
 
-  const merged = [...cached, ...toAppend].sort((a, b) => a.seq - b.seq);
+  const merged = [...bySeq.values()].sort((a, b) => a.seq - b.seq);
   return { merged, toAppend };
 }