billion-context-pi 0.1.72 → 0.1.73

package/dist/index.js

@@ -14,7 +14,7 @@ import { readFileSync as readFileSync6 } from "fs";
 import { homedir as homedir10 } from "os";
 import { join as join16 } from "path";
 
-// node_modules/acp-kernel/dist/chunk-QPVKTDYF.js
+// node_modules/acp-kernel/dist/chunk-VPRKYTHW.js
 import { createRequire } from "module";
 var require2 = createRequire(import.meta.url);
 function defaultCountTokens(text) {
@@ -227,6 +227,8 @@ function formatRanges(compressible, protectedRanges) {
       endRef: r.endRef,
       startNum: refNum3(r.startRef),
       endNum: refNum3(r.endRef),
+      startPos: r.startIndex ?? refNum3(r.startRef),
+      endPos: r.endIndex ?? refNum3(r.endRef),
       count: r.count,
       tokens: r.tokens,
       userMsgs: r.userMsgs ?? 0,
@@ -246,6 +248,8 @@ function formatRanges(compressible, protectedRanges) {
       endRef: r.endRef,
       startNum: refNum3(r.startRef),
       endNum: refNum3(r.endRef),
+      startPos: r.startIndex ?? refNum3(r.startRef),
+      endPos: r.endIndex ?? refNum3(r.endRef),
       count: r.count,
       tokens: r.tokens,
       userMsgs: 0,
@@ -259,13 +263,14 @@ function formatRanges(compressible, protectedRanges) {
       dangerous: false
     });
   }
-  entries.sort((a, b2) => a.startNum - b2.startNum);
+  entries.sort((a, b2) => a.startPos - b2.startPos || a.startNum - b2.startNum);
   const merged = [];
   for (const e of entries) {
     const last = merged[merged.length - 1];
-    if (last && e.startNum <= last.endNum + 1) {
+    if (last && e.startPos <= last.endPos + 1) {
       last.endRef = e.endRef;
       last.endNum = Math.max(last.endNum, e.endNum);
+      last.endPos = Math.max(last.endPos, e.endPos);
       last.count += e.count;
       last.tokens += e.tokens;
       last.userMsgs += e.userMsgs;
@@ -1519,7 +1524,7 @@ var ACP_STATUS_TOOL_OPENAI = {
     }
   }
 };
-var ACP_CACHE_TOOL_DESCRIPTION = "Prompt-cache reconciliation: grand ledger (total input/cached/output, overall hit rate) with every request's miss split into new content / compression re-pay / TTL expiry, plus per-fold economics (breakeven turns vs measured cadence). Read-only. Call when asked about cache hits, cache invalidation, or what compression costs.";
+var ACP_CACHE_TOOL_DESCRIPTION = `Prompt-cache reconciliation: grand ledger (total input/cached/output, overall hit rate) with every request's miss split into new content / compression re-pay / TTL expiry, plus per-fold economics (breakeven turns vs measured cadence). Defaults to a compact summary (totals + verdicts + anomalies only); pass detail="full" for every fold and line item. Read-only. Call when asked about cache hits, cache invalidation, or what compression costs.`;
 var ACP_CACHE_TOOL_OPENAI = {
   type: "function",
   function: {
@@ -1527,7 +1532,13 @@ var ACP_CACHE_TOOL_OPENAI = {
     description: ACP_CACHE_TOOL_DESCRIPTION,
     parameters: {
       type: "object",
-      properties: {}
+      properties: {
+        detail: {
+          type: "string",
+          enum: ["summary", "full"],
+          description: '"summary" (default): totals, verdicts, notable folds, anomalous requests only. "full": every retained fold and line item.'
+        }
+      }
     }
   }
 };
@@ -1588,6 +1599,62 @@ var ACP_TOOL_NAMES = /* @__PURE__ */ new Set([
   ACP_STATUS_TOOL_NAME,
   ACP_CACHE_TOOL_NAME
 ]);
+var COMPRESS_TOOL_GOOGLE = {
+  name: COMPRESS_TOOL_NAME,
+  description: COMPRESS_TOOL.description,
+  // `anyOf` is rejected by older API revisions, so `content` declares the
+  // object form; a JSON-encoded string of that array is still accepted by
+  // parseCompressInput, and the line form (a summary whose first line is
+  // 'm00150–m00220 optional topic') is documented in the description.
+  parameters: {
+    type: "object",
+    properties: {
+      topic: {
+        type: "string",
+        description: "Optional short title for the compressed range"
+      },
+      content: {
+        type: "array",
+        description: "One or more ranges to compress into separate summary blocks. Object form: {startId,endId,summary,topic?}. A JSON-encoded string of that array is also accepted; in the line form the summary begins with its own first line 'm00150\u2013m00220 optional topic', the rest being the summary markdown verbatim. REQUIRED \u2014 compress without content is invalid.",
+        items: {
+          type: "object",
+          properties: {
+            topic: { type: "string" },
+            startId: {
+              type: "string",
+              description: "mNNNNN ref at the start of the range"
+            },
+            endId: {
+              type: "string",
+              description: "mNNNNN ref at the end of the range"
+            },
+            summary: {
+              type: "string",
+              description: "Self-contained summary replacing the range"
+            }
+          },
+          required: ["startId", "endId", "summary"]
+        }
+      }
+    },
+    required: ["content"]
+  }
+};
+var DECOMPRESS_TOOL_GOOGLE = {
+  name: DECOMPRESS_TOOL_NAME,
+  description: DECOMPRESS_TOOL_OPENAI.function.description,
+  parameters: DECOMPRESS_TOOL_OPENAI.function.parameters
+};
+var SEARCH_CONTEXT_TOOL_GOOGLE = {
+  name: SEARCH_CONTEXT_TOOL_NAME,
+  description: SEARCH_CONTEXT_TOOL_OPENAI.function.description,
+  parameters: SEARCH_CONTEXT_TOOL_OPENAI.function.parameters
+};
+var ACP_STATUS_TOOL_GOOGLE = {
+  name: ACP_STATUS_TOOL_NAME,
+  description: ACP_STATUS_TOOL_OPENAI.function.description,
+  parameters: ACP_STATUS_TOOL_OPENAI.function.parameters
+};
 var ALWAYS_PROTECTED_TOOLS = ["compress"];
 var NEVER_PRESERVE_RECENT_TOOLS = [
   "decompress",
@@ -2012,6 +2079,73 @@ function computeTurnGroups(messages) {
   }
   return groups;
 }
+function computeIntegrityWithdrawals(messages, foldedIds) {
+  const remaining = new Set(foldedIds);
+  const withdrawn = /* @__PURE__ */ new Set();
+  const handledTurns = /* @__PURE__ */ new Set();
+  const handledPairs = /* @__PURE__ */ new Set();
+  const reasoningIds = /* @__PURE__ */ new Set();
+  const callIds = /* @__PURE__ */ new Set();
+  const callIdByMessageId = /* @__PURE__ */ new Map();
+  const resultIdByCallId = /* @__PURE__ */ new Map();
+  for (const m2 of messages) {
+    if (!m2.id) continue;
+    if (m2.contentType === "reasoning") reasoningIds.add(m2.id);
+    if (m2.role === "assistant" && m2.contentType === "tool-call") {
+      callIds.add(m2.id);
+      if (typeof m2.toolCallId === "string") {
+        callIdByMessageId.set(m2.id, m2.toolCallId);
+      }
+    }
+    if (m2.contentType === "tool-result" && typeof m2.toolCallId === "string") {
+      if (!resultIdByCallId.has(m2.toolCallId)) {
+        resultIdByCallId.set(m2.toolCallId, m2.id);
+      }
+    }
+  }
+  const groups = computeTurnGroups(messages);
+  let changed = true;
+  while (changed) {
+    changed = false;
+    for (let g2 = 0; g2 < groups.length; g2++) {
+      if (handledTurns.has(g2)) continue;
+      const group = groups[g2];
+      const foldHasReasoning = group.some(
+        (id) => remaining.has(id) && reasoningIds.has(id)
+      );
+      if (!foldHasReasoning) continue;
+      const keptHasCall = group.some(
+        (id) => !remaining.has(id) && callIds.has(id)
+      );
+      if (!keptHasCall) continue;
+      handledTurns.add(g2);
+      for (const id of group) {
+        remaining.delete(id);
+        withdrawn.add(id);
+      }
+      changed = true;
+    }
+    for (const m2 of messages) {
+      if (!m2.id || !callIds.has(m2.id)) continue;
+      const callId = callIdByMessageId.get(m2.id);
+      if (callId === void 0 || handledPairs.has(callId)) continue;
+      const resultId = resultIdByCallId.get(callId);
+      if (resultId === void 0) continue;
+      if (remaining.has(m2.id) === remaining.has(resultId)) continue;
+      handledPairs.add(callId);
+      remaining.delete(m2.id);
+      remaining.delete(resultId);
+      withdrawn.add(m2.id);
+      withdrawn.add(resultId);
+      changed = true;
+    }
+  }
+  return {
+    withdrawn,
+    splitTurnCount: handledTurns.size,
+    splitPairCount: handledPairs.size
+  };
+}
 function estimateTextTokens(text) {
   return Math.ceil(text.length / 4);
 }
@@ -2061,12 +2195,14 @@ function computeProtectedRefs(messages, state, config, countTokens = estimateTex
   return result;
 }
 function buildCompressibleRanges(messages, state, config, protectedZoneRefs, countTokens = estimateTextTokens) {
-  const compressibleMsgs = [];
+  let compressibleMsgs = [];
   const protectedMsgs = [];
   const protectedCallIds = collectProtectedToolCallIds(messages, config);
   let skipSinceCompressible = false;
   let skipSinceProtected = false;
+  let msgIndex = -1;
   for (const msg of messages) {
+    msgIndex++;
     const ref = state.messageRefs.byRaw[msg.id];
     if (!ref || ref === "BLOCKED") continue;
     if (isSyntheticOrPruned(msg, state)) {
@@ -2079,7 +2215,8 @@ function buildCompressibleRanges(messages, state, config, protectedZoneRefs, cou
         ref,
         gapBefore: skipSinceProtected,
         tokens: countMessageTokens(msg, countTokens),
-        tools: msg.toolName ? [msg.toolName] : []
+        tools: msg.toolName ? [msg.toolName] : [],
+        index: msgIndex
       });
       skipSinceProtected = false;
       skipSinceCompressible = true;
@@ -2091,16 +2228,35 @@ function buildCompressibleRanges(messages, state, config, protectedZoneRefs, cou
       continue;
     }
     compressibleMsgs.push({
+      id: msg.id,
       ref,
       gapBefore: skipSinceCompressible,
       tokens: countMessageTokens(msg, countTokens),
       chars: (msg.text ?? "").length,
       isTool: isToolMessage(msg),
-      isUser: msg.role === "user"
+      isUser: msg.role === "user",
+      index: msgIndex
     });
     skipSinceCompressible = false;
     skipSinceProtected = true;
   }
+  const unfoldedIds = computeIntegrityWithdrawals(
+    messages,
+    new Set(compressibleMsgs.map((info) => info.id))
+  ).withdrawn;
+  if (unfoldedIds.size > 0) {
+    let gapPending = false;
+    const kept = [];
+    for (const info of compressibleMsgs) {
+      if (unfoldedIds.has(info.id)) {
+        gapPending = true;
+        continue;
+      }
+      kept.push(gapPending ? { ...info, gapBefore: true } : info);
+      gapPending = false;
+    }
+    compressibleMsgs = kept;
+  }
   const compressible = [];
   let cur = null;
   for (const info of compressibleMsgs) {
@@ -2117,10 +2273,13 @@ function buildCompressibleRanges(messages, state, config, protectedZoneRefs, cou
         chars: info.chars,
         toolPct: info.isTool ? 100 : 0,
         textPct: info.isTool ? 0 : 100,
-        userMsgs: info.isUser ? 1 : 0
+        userMsgs: info.isUser ? 1 : 0,
+        startIndex: info.index,
+        endIndex: info.index
       };
     } else {
       cur.endRef = info.ref;
+      cur.endIndex = info.index;
       cur.count++;
       cur.tokens += info.tokens;
       cur.chars = (cur.chars ?? 0) + info.chars;
@@ -2147,10 +2306,13 @@ function buildCompressibleRanges(messages, state, config, protectedZoneRefs, cou
         endRef: info.ref,
         count: 1,
         tokens: info.tokens,
-        tools: [...info.tools]
+        tools: [...info.tools],
+        startIndex: info.index,
+        endIndex: info.index
       };
     } else {
       pcur.endRef = info.ref;
+      pcur.endIndex = info.index;
       pcur.count++;
       pcur.tokens += info.tokens;
       for (const t of info.tools) {
@@ -2183,6 +2345,10 @@ function mergeBatch(batch) {
     textPct: 100 - toolPct,
     userMsgs: batch.reduce((s, r) => s + (r.userMsgs ?? 0), 0)
   };
+  if (first.startIndex !== void 0 && last.endIndex !== void 0) {
+    merged.startIndex = Math.min(...batch.map((r) => r.startIndex ?? Infinity));
+    merged.endIndex = Math.max(...batch.map((r) => r.endIndex ?? -Infinity));
+  }
   if (batch.some((r) => r.dangerous === true)) {
     merged.dangerous = true;
   }
@@ -2765,49 +2931,11 @@ function applySingleRange(input) {
     );
   }
   {
-    const reasoningIds = /* @__PURE__ */ new Set();
-    const callIds = /* @__PURE__ */ new Set();
-    for (const m2 of input.messages) {
-      if (!m2.id) continue;
-      if (m2.contentType === "reasoning") reasoningIds.add(m2.id);
-      if (m2.role === "assistant" && m2.contentType === "tool-call")
-        callIds.add(m2.id);
-    }
-    const withdrawIds = /* @__PURE__ */ new Set();
-    let splitTurnCount = 0;
-    for (const group of computeTurnGroups(input.messages)) {
-      const foldHasReasoning = group.some(
-        (id) => effectiveMessageIds.has(id) && reasoningIds.has(id)
-      );
-      if (!foldHasReasoning) continue;
-      const keptHasCall = group.some(
-        (id) => !effectiveMessageIds.has(id) && callIds.has(id)
-      );
-      if (!keptHasCall) continue;
-      splitTurnCount++;
-      for (const id of group) withdrawIds.add(id);
-    }
-    const resultIdByCallId = /* @__PURE__ */ new Map();
-    for (const m2 of input.messages) {
-      if (!m2.id || m2.contentType !== "tool-result") continue;
-      if (typeof m2.toolCallId !== "string") continue;
-      if (!resultIdByCallId.has(m2.toolCallId)) {
-        resultIdByCallId.set(m2.toolCallId, m2.id);
-      }
-    }
-    let splitPairCount = 0;
-    for (const m2 of input.messages) {
-      if (!m2.id || m2.contentType !== "tool-call") continue;
-      if (typeof m2.toolCallId !== "string") continue;
-      const resultId = resultIdByCallId.get(m2.toolCallId);
-      if (resultId === void 0) continue;
-      if (effectiveMessageIds.has(m2.id) === effectiveMessageIds.has(resultId)) {
-        continue;
-      }
-      withdrawIds.add(m2.id);
-      withdrawIds.add(resultId);
-      splitPairCount++;
-    }
+    const {
+      withdrawn: withdrawIds,
+      splitTurnCount,
+      splitPairCount
+    } = computeIntegrityWithdrawals(input.messages, effectiveMessageIds);
     if (withdrawIds.size > 0) {
       for (const id of withdrawIds) effectiveMessageIds.delete(id);
       const beforeWithdraw = filteredIds.length;
@@ -3972,13 +4100,35 @@ function fmtTime(at) {
   const p = (x2) => String(x2).padStart(2, "0");
   return `${p(d2.getHours())}:${p(d2.getMinutes())}:${p(d2.getSeconds())}`;
 }
-function formatCacheReport(report, sessionLabel) {
+var HIT_HEALTHY_PCT = 90;
+var HIT_WATCH_PCT = 70;
+var ANOMALY_HIT_PCT = 85;
+var ANOMALY_MISS_TOK = 5e3;
+var TTL_SPIKE_TOK = 1e4;
+var SUMMARY_FOLD_TOPS = 3;
+var SUMMARY_ANOMALY_CAP = 20;
+function hitVerdict(pct2) {
+  if (pct2 >= HIT_HEALTHY_PCT) return "HEALTHY";
+  if (pct2 >= HIT_WATCH_PCT) return "WATCH";
+  return "INVESTIGATE";
+}
+function medianHitPct(lines) {
+  if (lines.length === 0) return null;
+  const xs = lines.map((l3) => l3.hitPct).sort((a, b2) => a - b2);
+  return xs[Math.floor(xs.length / 2)] ?? null;
+}
+function fmtIdleGap(ms) {
+  if (ms < 9e4) return `${Math.max(1, Math.round(ms / 1e3))}s`;
+  if (ms < 90 * 6e4) return `${Math.round(ms / 6e4)}m`;
+  if (ms < 36 * 36e5) return `${(ms / 36e5).toFixed(1)}h`;
+  return `${(ms / 864e5).toFixed(1)}d`;
+}
+function foldVerdict(f) {
+  return f.paidBack === null ? "?" : f.paidBack ? "PAID BACK" : "NOT PAID BACK";
+}
+function formatCacheReportFull(report) {
   const t = report.totals;
   const out = [];
-  out.push(
-    `ACP CACHE REPORT${sessionLabel ? ` (${sessionLabel})` : ""} \u2014 ${t.requests} requests`
-  );
-  out.push("");
   out.push("GRAND LEDGER");
   out.push(`  total input    ${fmtTok(t.input)} tok`);
   out.push(
@@ -4038,6 +4188,145 @@ function formatCacheReport(report, sessionLabel) {
   }
   return out.join("\n");
 }
+function formatCacheReportSummary(report) {
+  const t = report.totals;
+  const out = [];
+  out.push("GRAND LEDGER");
+  out.push(`  total input    ${fmtTok(t.input)} tok`);
+  out.push(
+    `  total cached   ${fmtTok(t.cached)} tok  (hit ${t.hitPct.toFixed(1)}% \u2192 ${hitVerdict(t.hitPct)})`
+  );
+  out.push(`  total output   ${fmtTok(t.output)} tok`);
+  out.push(
+    `  miss breakdown (input \u2212 cached = ${fmtTok(Math.max(0, t.input - t.cached))} tok):`
+  );
+  const avgNew = t.requests > 0 ? t.newContent / t.requests : 0;
+  out.push(
+    `    new content     ${fmtTok(t.newContent)} tok  (~${fmtTok(avgNew)}/req fresh append \u2014 not an invalidation)`
+  );
+  out.push(
+    `    compress re-pay ${fmtTok(t.compRepay)} tok  (${report.economics.folds} folds \u2014 re-billed prefix)`
+  );
+  const spikeIdx = [];
+  report.lines.forEach((l3, i) => {
+    if (l3.ttlRepay >= TTL_SPIKE_TOK) spikeIdx.push(i);
+  });
+  const spikes = [...spikeIdx].sort(
+    (a, b2) => (report.lines[b2]?.ttlRepay ?? 0) - (report.lines[a]?.ttlRepay ?? 0)
+  ).slice(0, SUMMARY_FOLD_TOPS).sort((a, b2) => a - b2);
+  const spikeText = spikes.map((i) => {
+    const l3 = report.lines[i];
+    if (!l3) return "";
+    let idle = "";
+    const prev = report.lines[i - 1];
+    if (prev && l3.at > prev.at) idle = `, idle ${fmtIdleGap(l3.at - prev.at)}`;
+    return `#${l3.seq} ${fmtTok(l3.ttlRepay)}${idle}`;
+  }).join(" \xB7 ");
+  out.push(
+    `    ttl/other       ${fmtTok(t.ttlRepay)} tok  (stable-prefix misses: TTL expiry / eviction${spikeText ? `; top spikes: ${spikeText}` : ""})`
+  );
+  out.push(
+    `  identity check   ${t.balanced ? "OK" : "BROKEN"} \u2014 ${fmtTok(t.input)} = ${fmtTok(t.cached)} + ${fmtTok(t.newContent)} + ${fmtTok(t.compRepay)} + ${fmtTok(t.ttlRepay)} (residual ${t.residual})`
+  );
+  const e = report.economics;
+  if (e.folds > 0) {
+    const p = report.profile;
+    out.push("");
+    out.push(`FOLD ECONOMICS (${e.folds} folds @ w=${p.w} r=${p.r} q=${p.q})`);
+    out.push(
+      `  gross saved ${fmtTok(e.grossSaved)} tok \xB7 repay cost ${fmtTok(e.repayCost)} tok \xB7 summary cost ${fmtTok(e.summaryCost)} tok \u2192 net ${e.netTokens >= 0 ? "+" : ""}${fmtTok(e.netTokens)} tok`
+    );
+    out.push(
+      `  verdict: ${e.paidBackCount} PAID BACK / ${e.notPaidBackCount} NOT PAID BACK / ${e.unobservedCount} unobserved`
+    );
+    if (e.notPaidBackCount > 0)
+      out.push(
+        `  NOT PAID BACK = measured post-fold cadence never reached n* (end-of-session / back-to-back folds \u2014 one-time cost, not data loss)`
+      );
+    if (report.folds.length <= SUMMARY_FOLD_TOPS * 2) {
+      for (const f of report.folds) {
+        const nstar = f.breakevenTurns === null ? "n/a" : f.breakevenTurns.toFixed(1);
+        const k = f.turnsToNextFold === null ? "\u2014" : String(f.turnsToNextFold);
+        const h = f.hPct === null ? "n/a" : `${f.hPct.toFixed(1)}%`;
+        out.push(
+          `  #${f.seq} ${fmtTime(f.at)} S=${fmtTok(f.S)} \u03C3=${fmtTok(f.sigma)} h=${h} T=${fmtTok(f.T)} \u0394C\u2081=${fmtTok(f.oneTimeCostUnits)} \u0394s=${fmtTok(f.perTurnSavingUnits)}/turn n*=${nstar} k=${k} \u2192 ${foldVerdict(f)}`
+        );
+      }
+    } else {
+      const largest = [...report.folds].sort((a, b2) => b2.S - a.S).slice(0, SUMMARY_FOLD_TOPS);
+      const worst = [...report.folds].sort((a, b2) => b2.oneTimeCostUnits - a.oneTimeCostUnits).slice(0, SUMMARY_FOLD_TOPS);
+      out.push(
+        `  largest folds: ${largest.map((f) => `#${f.seq} S=${fmtTok(f.S)} ${foldVerdict(f)}`).join(" \xB7 ")}`
+      );
+      out.push(
+        `  worst one-time: ${worst.map((f) => `#${f.seq} \u0394C\u2081=${fmtTok(f.oneTimeCostUnits)} (${foldVerdict(f)}${f.turnsToNextFold !== null ? `, k=${f.turnsToNextFold}` : ""})`).join(" \xB7 ")}`
+      );
+      const shown = /* @__PURE__ */ new Set([...largest, ...worst]);
+      if (report.folds.length > shown.size)
+        out.push(
+          `  \u2192 ${report.folds.length - shown.size} more folds omitted (detail:"full" lists every fold)`
+        );
+    }
+  }
+  if (report.lines.length > 0) {
+    out.push("");
+    const shownIdx = [];
+    report.lines.forEach((l3, i) => {
+      if (l3.hitPct < ANOMALY_HIT_PCT || l3.missed >= ANOMALY_MISS_TOK)
+        shownIdx.push(i);
+    });
+    const last = report.lines.length - 1;
+    if (!shownIdx.includes(last)) shownIdx.push(last);
+    const kept = shownIdx.slice(Math.max(0, shownIdx.length - SUMMARY_ANOMALY_CAP));
+    const anomalies = kept.filter((i) => {
+      const l3 = report.lines[i];
+      if (!l3) return false;
+      return l3.hitPct < ANOMALY_HIT_PCT || l3.missed >= ANOMALY_MISS_TOK;
+    }).length;
+    if (anomalies === 0) {
+      const med = medianHitPct(report.lines);
+      out.push(
+        `LINE ITEMS \u2014 no anomalies (${report.lines.length} requests, median hit ${med !== null ? med.toFixed(1) : "n/a"}%${report.linesOmitted > 0 ? `, ${report.linesOmitted} older outside window` : ""})`
+      );
+    } else {
+      out.push(
+        `LINE ITEMS (anomalies: hit<${ANOMALY_HIT_PCT}% or miss\u2265${fmtTok(ANOMALY_MISS_TOK)}):`
+      );
+      out.push(
+        "  #     time      input  cached    hit%      new    comp     ttl  fold"
+      );
+      for (const i of kept) {
+        const l3 = report.lines[i];
+        if (!l3) continue;
+        const fold = l3.foldSeq !== null ? `#${l3.foldSeq}` : "";
+        out.push(
+          `  ${String(l3.seq).padStart(4)}  ${fmtTime(l3.at)}  ${fmtTok(l3.input).padStart(7)}  ${fmtTok(l3.cached).padStart(7)}  ${l3.hitPct.toFixed(1).padStart(5)}%  ${fmtTok(l3.newContent).padStart(6)}  ${fmtTok(l3.compRepay).padStart(6)}  ${fmtTok(l3.ttlRepay).padStart(6)}  ${fold}`
+        );
+      }
+      const omitted = report.lines.length - kept.length;
+      if (omitted > 0) {
+        const rest = report.lines.filter((_2, i) => !kept.includes(i));
+        const med = medianHitPct(rest);
+        out.push(
+          `  \u2026 ${omitted} lines omitted (median hit ${med !== null ? med.toFixed(1) : "n/a"}%${report.linesOmitted > 0 ? `, ${report.linesOmitted} older outside window` : ""} \u2014 detail:"full" lists all)`
+        );
+      }
+    }
+  }
+  return out.join("\n");
+}
+function formatCacheReport(report, sessionLabel, opts) {
+  const t = report.totals;
+  const header = `ACP CACHE REPORT${sessionLabel ? ` (${sessionLabel})` : ""} \u2014 ${t.requests} requests`;
+  if (opts?.detail === "full") {
+    return [header, "", formatCacheReportFull(report)].join("\n");
+  }
+  return [
+    header + `  [summary \u2014 detail:"full" for every fold & line]`,
+    "",
+    formatCacheReportSummary(report)
+  ].join("\n");
+}
 function renderMessage2(m2) {
   const parts = [];
   switch (m2.contentType) {
@@ -12123,7 +12412,7 @@ function resolveActivePack(adapter, cwd, provider, modelId, resolver) {
 }
 var HOST_VERSION = (() => {
   try {
-    return "0.1.72";
+    return "0.1.73";
   } catch {
     return "dev";
   }
@@ -14172,7 +14461,9 @@ ${extra.join("\n")}` : base;
 }
 
 // src/cache-tool.ts
-var CacheParams = typebox_exports.Object({});
+var CacheParams = typebox_exports.Object({
+  detail: typebox_exports.Optional(typebox_exports.Union([typebox_exports.Literal("summary"), typebox_exports.Literal("full")], { description: '"summary" (default): totals, verdicts, notable folds, anomalous requests only. "full": every retained fold and line item.' }))
+});
 var num = (v2) => typeof v2 === "number" && Number.isFinite(v2) && v2 > 0 ? v2 : 0;
 var refNum2 = (ref) => Number(ref.replace(/\D/g, "")) || 0;
 function cacheSamples(entries) {
@@ -14207,26 +14498,26 @@ function foldEvents(state) {
     firstFoldStartTokens: prefixBefore(b2.startRef)
   }));
 }
-async function cacheReportText(runtime, ctx) {
+async function cacheReportText(runtime, ctx, detail = "summary") {
   const { state, entries } = await runtime.stateFor(ctx);
   const report = buildCacheReport(cacheSamples(entries), foldEvents(state));
-  return formatCacheReport(report, ctx.sessionManager.getSessionId());
+  return formatCacheReport(report, ctx.sessionManager.getSessionId(), { detail });
 }
 function makeCacheTool(runtime, overrides) {
   return applyToolPromptOverrides({
     name: "acp_cache",
     label: "ACP Cache Report",
-    description: "Prompt-cache reconciliation: grand ledger (total input/cached/output, session hit rate) with every request's miss split into new content / compression re-pay / TTL expiry, plus per-fold economics (one-time cost, breakeven turns vs measured cadence). Read-only.",
+    description: `Prompt-cache reconciliation: grand ledger (total input/cached/output, session hit rate) with every request's miss split into new content / compression re-pay / TTL expiry, plus per-fold economics (one-time cost, breakeven turns vs measured cadence). Defaults to a compact summary (totals + verdicts + anomalies only); pass detail="full" for every fold and line item. Read-only.`,
     promptSnippet: "acp_cache({})",
     promptGuidelines: [
       "Call when asked about cache hits, cache invalidation, or what compression costs.",
       "Read-only: reports numbers, never mutates context."
     ],
     parameters: CacheParams,
-    async execute(_toolCallId, _params, _signal, _onUpdate, ctx) {
+    async execute(_toolCallId, params, _signal, _onUpdate, ctx) {
       if (runtime.refused) return { details: void 0, content: [{ type: "text", text: runtime.refusalMessage ?? UNSUPPORTED_HOST_MESSAGE }] };
       try {
-        const text = await cacheReportText(runtime, ctx);
+        const text = await cacheReportText(runtime, ctx, params.detail === "full" ? "full" : "summary");
         return { details: void 0, content: [{ type: "text", text }] };
       } catch (e) {
         logThrow("cache", e, { sid: ctx.sessionManager.getSessionId() });
@@ -18714,11 +19005,12 @@ function makeCommands(runtime, pi) {
     {
       name: "acp-cache",
       options: {
-        description: "Prompt-cache reconciliation: grand ledger (input/cached/hit rate) with every request's miss split into new content / compression re-pay / TTL expiry, plus per-fold economics.",
-        handler: async (_args, ctx) => {
+        description: "Prompt-cache reconciliation: grand ledger (input/cached/hit rate) with every request's miss split into new content / compression re-pay / TTL expiry, plus per-fold economics. Append 'full' for the every-line listing.",
+        handler: async (args, ctx) => {
           let text;
           try {
-            text = await cacheReportText(runtime, ctx);
+            const detail = /(^|\s)(--)?full(\s|$)/.test(args ?? "") ? "full" : "summary";
+            text = await cacheReportText(runtime, ctx, detail);
           } catch (e) {
             ctx.ui.notify(e instanceof Error ? e.message : String(e), "error");
             return;
@@ -18868,7 +19160,7 @@ async function statusReport(runtime, ctx) {
   const sentTokens = estimateTokens(coreMessages, coveredIds, imageTokens) + systemPromptTokens;
   const viewSentTokens = adjustedTokenCount(runtime.core, coreMessages, state, config, sentTokens, imageTokens, systemPromptTokens);
   const turn = runtime.core.processTurn({ messages: coreMessages, state, config, tokenCount: anchorStale ? viewSentTokens : Math.max(viewSentTokens, realUsage?.tokens ?? 0) });
-  const versionStr = "0.1.72" ? `billion-context-pi@${"0.1.72"} \xB7 pack: ${resolveSurfaceMeta(runtime.adapter, ctx?.cwd ?? process.cwd(), ctx?.model?.provider, ctx?.model?.id).pack}` : void 0;
+  const versionStr = "0.1.73" ? `billion-context-pi@${"0.1.73"} \xB7 pack: ${resolveSurfaceMeta(runtime.adapter, ctx?.cwd ?? process.cwd(), ctx?.model?.provider, ctx?.model?.id).pack}` : void 0;
   let text = buildStatusPanel({
     version: versionStr,
     tokenCount: sessionTokens,
@@ -18889,6 +19181,94 @@ async function statusReport(runtime, ctx) {
   return text;
 }
 
+// src/live-only-tail.ts
+var MAX_LIVE_ONLY_TAIL = 8;
+function sigText(content) {
+  if (typeof content === "string") return content;
+  if (!Array.isArray(content)) return "";
+  const parts = [];
+  for (const block of content) {
+    const b2 = block;
+    if (b2?.type === "text" && typeof b2.text === "string") parts.push(b2.text);
+  }
+  return parts.join("\n");
+}
+function weakMessageSig(message) {
+  const m2 = message ?? {};
+  const role = typeof m2.role === "string" ? m2.role : "";
+  const customType = typeof m2.customType === "string" ? m2.customType : "";
+  const toolCallId = typeof m2.toolCallId === "string" ? m2.toolCallId : "";
+  const text = sigText(m2.content);
+  return [role, customType, toolCallId, String(text.length), text.slice(0, 32), text.length > 32 ? text.slice(-32) : ""].join("\0");
+}
+function entryAsLiveShape(entry) {
+  if (entry.type === "message") {
+    const msg = entry.message;
+    return { role: msg.role, customType: msg.customType, toolCallId: msg.toolCallId, content: msg.content };
+  }
+  if (entry.type === "custom_message") {
+    return { role: "custom", customType: entry.customType, content: entry.content };
+  }
+  return void 0;
+}
+function liveOnlyTail(entries, live) {
+  const persisted = entries.filter((e) => e.type === "message" || e.type === "custom_message").map((e) => entryAsLiveShape(e)).filter((shape) => shape !== void 0);
+  const max = Math.min(persisted.length, live.length);
+  let i = 0;
+  while (i < max && weakMessageSig(persisted[i]) === weakMessageSig(live[i])) i++;
+  if (i === max) {
+    const tail = live.slice(max);
+    return tail.length > 0 && tail.length <= MAX_LIVE_ONLY_TAIL ? tail : null;
+  }
+  if (i === max - 1 && i === persisted.length - 1 && i === live.length - 1) {
+    const liveMsg = live[i];
+    if (liveMsg && typeof liveMsg.role === "string" && liveMsg.role === "user") {
+      const before = sigText(persisted[i]?.content);
+      const after = sigText(liveMsg.content);
+      if (before && after && after.startsWith(before)) {
+        const suffix = after.slice(before.length).trim();
+        if (suffix.length > 0) {
+          return [{ role: "user", content: [{ type: "text", text: suffix }], timestamp: Date.now() }];
+        }
+      }
+    }
+  }
+  return null;
+}
+
+// src/system-passthrough.ts
+function isSystem(message) {
+  return typeof message === "object" && message !== null && message.role === "system";
+}
+function carryHostSystemMessages(rebuilt, input) {
+  const inputSystems = input.filter(isSystem);
+  if (inputSystems.length === 0) return rebuilt;
+  const rebuiltSystems = rebuilt.filter(isSystem);
+  if (rebuiltSystems.length === 0) return [...inputSystems, ...rebuilt];
+  let changed = false;
+  let sysIdx = 0;
+  const patched = rebuilt.map((message) => {
+    if (!isSystem(message)) return message;
+    const source = inputSystems[sysIdx++];
+    if (!source) return message;
+    const out = message;
+    const src = source;
+    let next = null;
+    for (const key of Object.keys(src)) {
+      if (out[key] === void 0 && src[key] !== void 0) {
+        next ??= { ...out };
+        next[key] = src[key];
+      }
+    }
+    if (next !== null) {
+      changed = true;
+      return next;
+    }
+    return message;
+  });
+  return changed ? patched : rebuilt;
+}
+
 // src/degeneration.ts
 var DEFAULT_DEGENERATION_GUARD = { enabled: true, minRun: 200 };
 var MIN_VALID_MIN_RUN = 8;
@@ -19480,7 +19860,7 @@ async function autoInstallLatest(latest, extDirOverride) {
       "--no-fund",
       "--no-save"
     ];
-    const prevVersion = (await readPackageJson(join15(extDir, "package.json")))?.version ?? "0.1.72";
+    const prevVersion = (await readPackageJson(join15(extDir, "package.json")))?.version ?? "0.1.73";
     const { code, stderr } = await runNpmImpl(installArgs(latest), { cwd: npmDir, timeout: 6e4 });
     if (code !== 0) {
       logWarn("update", {
@@ -19498,7 +19878,7 @@ async function autoInstallLatest(latest, extDirOverride) {
     }
     const verify = await verifyInstall(npmDir, latest);
     if (!verify.ok) {
-      const rollbackTo = SEMVER_RE.test(prevVersion) ? prevVersion : "0.1.72";
+      const rollbackTo = SEMVER_RE.test(prevVersion) ? prevVersion : "0.1.73";
       logWarn("update", { event: "auto-install-verify-failed", latest, reason: verify.reason, rollbackTo });
       const rb = await runNpmImpl(installArgs(rollbackTo), { cwd: npmDir, timeout: 6e4 });
       logInfo("update", { event: "rollback", from: latest, to: rollbackTo, ok: rb.code === 0 });
@@ -19573,7 +19953,7 @@ async function checkForUpdate(autoUpdate, notify) {
     }
     const latest = await fetchLatestVersion(tag);
     if (!latest) return;
-    const current = runtimeVersion ?? "0.1.72";
+    const current = runtimeVersion ?? "0.1.73";
     const hasUpdate = isVersionNewer(latest, current);
     debug.event("update-check", {
       current,
@@ -19775,7 +20155,7 @@ function wireSessionLifecycle(pi, runtime, standDownIfProxied) {
     setDelegatePolicy(DEFAULT_DELEGATE_POLICY);
     const sid = ctx.sessionManager.getSessionId();
     const modelInfo = ctx.model;
-    logInfo("session", { event: "start", sid, cwd: ctx.cwd, debug: runtime.adapter.debug ?? null, version: true ? "0.1.72" : null, model: modelInfo?.id ?? null, modelApi: modelInfo?.api ?? null, contextWindow: modelInfo?.contextWindow ?? null });
+    logInfo("session", { event: "start", sid, cwd: ctx.cwd, debug: runtime.adapter.debug ?? null, version: true ? "0.1.73" : null, model: modelInfo?.id ?? null, modelApi: modelInfo?.api ?? null, contextWindow: modelInfo?.contextWindow ?? null });
     let delegateStoodDown = false;
     try {
       await runtime.reloadConfig(ctx.cwd);
@@ -20031,6 +20411,16 @@ ${rendered.text}${example}`);
           ctx.ui.notify(`[ACP] compress failed ${outcome.count}\xD7 this turn \u2014 nudge paused until the next user message (emergency truncation still active).`);
         }
       }
+      const liveTail = liveOnlyTail(entries, event.messages);
+      if (liveTail && liveTail.length > 0) {
+        rebuilt.push(...liveTail);
+        logInfo("live-only-tail", { sid, event: "appended", tail: liveTail.length, outMsgs: rebuilt.length });
+      }
+      const withHostSystem = carryHostSystemMessages(rebuilt, event.messages);
+      if (withHostSystem !== rebuilt) {
+        rebuilt = withHostSystem;
+        logInfo("system-passthrough", { sid, event: "carried", systems: event.messages.filter((m2) => m2.role === "system").length, outMsgs: rebuilt.length });
+      }
       debug.event("context-out", { outMsgs: rebuilt.length, injected: turn.nudge?.shouldInject ?? false, emergency: turn.nudge?.breakdown?.emergencyOverride === 1 });
       const updateCheck = checkForUpdate(runtime.adapter.autoUpdate ?? true, (msg) => {
         if (ctx.hasUI) ctx.ui.notify(msg);