billion-context 0.1.103 → 0.1.104-pr.700.529

package/dist/index.js

@@ -51956,7 +51956,11 @@ ${reasons.map((r) => `- ${r}`).join("\n")}` : "";
     const shrinkRatio = preContext > 0 ? r.tokensCompressed / preContext : 0;
     const foldPoint = [...ranges].sort((a, b2) => refNum(a.startRef) - refNum(b2.startRef))[0]?.startRef ?? "unknown";
     ctx.session.lastCompress = { at: Date.now(), shrinkRatio, foldPoint, blocks: r.blocksCreated, tokensCompressed: r.tokensCompressed };
-    ctx.log(`[acp-compress-obs] shrink ${Math.round(shrinkRatio * 100)}% (~${r.tokensCompressed}/${preContext} tok) foldPoint=${foldPoint} blocks=${r.blocksCreated}`);
+    ctx.session.stats.pendingFoldUsage = true;
+    const anchorTok = res.state.blocks.reduce((n, b2) => n + (b2.active ? Math.ceil(b2.summary.length / 4) : 0), 0);
+    const postCtx = Math.max(0, preContext - r.tokensCompressed);
+    const ceiling = postCtx > 0 ? Math.floor(100 * anchorTok / postCtx) : 0;
+    ctx.log(`[acp-compress-obs] shrink ${Math.round(shrinkRatio * 100)}% (~${r.tokensCompressed}/${preContext} tok) foldPoint=${foldPoint} blocks=${r.blocksCreated} anchor\u2248${anchorTok} tok (${res.state.blocks.filter((b2) => b2.active).length} active blocks, sys excluded) postCtx\u2248${postCtx} \u2192 next-request cache ceiling \u2265${ceiling}%`);
     const warn = r.warnings.length > 0 ? ` ${r.warnings.join("; ")}` : "";
     let msg = `[Compressed ${detail} \u2192 ${r.blocksCreated} block(s), ~${r.tokensCompressed} tokens saved.${warn}]`;
     const maxShrink = maxShrinkPerCompress();
@@ -53490,8 +53494,10 @@ function recordUsage(ctx, usage, round) {
   if (typeof out === "number") ctx.session.stats.outputTokens += out;
   const hitPct = typeof cached === "number" && total > 0 ? Math.round(cached / total * 100) : 0;
   warnCacheCollapse(ctx.session, total, cached ?? 0);
+  const foldNew = ctx.session.stats.pendingFoldUsage === true;
+  if (foldNew) ctx.session.stats.pendingFoldUsage = false;
   ctx.log(
-    `[acp-usage] round ${round} input=${total} cached=${cached ?? 0} (cache hit ${hitPct}%)`
+    `[acp-usage] round ${round} input=${total} cached=${cached ?? 0} (cache hit ${hitPct}%)${foldNew ? " fold=new" : ""}`
   );
 }
 async function* runCompressLoop(upstream, ctx, requestBody, requestOptions, adapter, systemPrompt, signal) {
@@ -56770,6 +56776,10 @@ function applyUsageSample(session, sample, protocol) {
     session.stats.lastInputTokens = Math.max(0, total - (session.stats.compressCreditTokens ?? 0));
     warnCacheCollapse(session, total, sample.cachedTokens ?? 0);
     session.hostContextTokens = total + (session.hostCreditTokens ?? 0);
+    const hit = sample.cachedTokens === void 0 || total <= 0 ? void 0 : Math.round(100 * (sample.cachedTokens ?? 0) / total);
+    const foldNew = session.stats.pendingFoldUsage === true;
+    if (foldNew) session.stats.pendingFoldUsage = false;
+    log("info", `[${session.id}] [plugin] [acp-usage] input=${total} cached=${sample.cachedTokens ?? "n/a"}${hit === void 0 ? "" : ` (cache hit ${hit}%)`} ctx=${session.hostContextTokens}${foldNew ? " fold=new" : ""}`);
   }
   if (sample.outputTokens !== void 0) session.stats.outputTokens += sample.outputTokens;
 }