nolo-cli 0.82.0-alpha.6 → 0.82.1

package/{chunk-E36HZARC.js → chunk-DNCEALKX.js}

@@ -6,13 +6,13 @@ import {
 } from "./chunk-DWLEOOGR.js";
 import {
   addContentAction
-} from "./chunk-ZZCJUXUI.js";
+} from "./chunk-V4MJDUSR.js";
 import {
   upload
-} from "./chunk-YFADVSZE.js";
+} from "./chunk-H62HCIUP.js";
 import {
   getRuntimeServerContext
-} from "./chunk-3RAG2NAD.js";
+} from "./chunk-UKWGR5SS.js";
 import {
   API_ENDPOINTS
 } from "./chunk-LCBAMY2M.js";
@@ -367,6 +367,7 @@ var truncateToolOutputForContext = (text, contextWindow) => {
 // packages/ai/context/planCompression.ts
 var MIN_COMPRESS_COUNT = 5;
 var ACTIVE_SUMMARY_TAIL_KEEP_COUNT = 2;
+var COMPACTION_RETENTION_CUSHION_RATIO = 0.18;
 var MIN_TRIGGER_PENDING_COUNT = MIN_COMPRESS_COUNT + ACTIVE_SUMMARY_TAIL_KEEP_COUNT;
 var getMessageTokenCount = (msg) => {
   const content = serializeMessageContent(msg.content) || "";
@@ -409,13 +410,14 @@ var normalizeContextUsageRatio = (value) => {
   const ratio = value > 1 ? value / 100 : value;
   return ratio >= 0 && ratio <= 1 ? ratio : void 0;
 };
-var emptyPlan = (startIndex) => ({
+var emptyPlan = (startIndex, diagnostics) => ({
   shouldCompress: false,
   compressCount: 0,
   msgsToCompress: [],
   msgsToKeep: [],
   newSummarizedBeforeId: void 0,
-  startIndex
+  startIndex,
+  ...diagnostics ? { diagnostics } : {}
 });
 function findPendingMessages(allMsgs, summarizedBeforeId) {
   let startIndex = 0;
@@ -464,15 +466,42 @@ function calculateCompressCount(pendingMsgs, rawMessageBudget, opts) {
   }
   return pendingMsgs.length - keepCount;
 }
+function resolveEffectiveCompactionBudgets(args) {
+  const {
+    contextWindow,
+    triggerRatio,
+    historyBudget,
+    rawMessageBudget,
+    summaryTokens,
+    cushionRatio = COMPACTION_RETENTION_CUSHION_RATIO
+  } = args;
+  const maxRetainedRatio = Math.max(
+    0.2,
+    Number((triggerRatio - cushionRatio).toFixed(4))
+  );
+  const maxRetainedHistoryBudget = Math.floor(contextWindow * maxRetainedRatio);
+  const effectiveHistoryBudget = Math.min(historyBudget, maxRetainedHistoryBudget);
+  const effectiveRawMessageBudget = Math.min(
+    rawMessageBudget,
+    Math.max(0, effectiveHistoryBudget - summaryTokens)
+  );
+  return {
+    effectiveHistoryBudget,
+    effectiveRawMessageBudget,
+    maxRetainedRatio
+  };
+}
 function guardToolChainBoundary(pendingMsgs, compressCount) {
   let count = compressCount;
-  while (count > 0 && count < pendingMsgs.length && pendingMsgs[count].role === "tool") {
-    count--;
+  while (count < pendingMsgs.length && pendingMsgs[count].role === "tool") {
+    count++;
   }
-  if (count > 0) {
-    const lastCompressed = pendingMsgs[count - 1];
-    if (hasOpenEndedToolCall(lastCompressed)) {
+  while (count > 0 && hasOpenEndedToolCall(pendingMsgs[count - 1])) {
+    if (count < pendingMsgs.length && pendingMsgs[count].role === "tool") {
+      count++;
+    } else {
       count--;
+      break;
     }
   }
   return count;
@@ -489,7 +518,20 @@ function planCompression(input) {
     coldResume
   } = input;
   const { pendingMsgs, startIndex } = findPendingMessages(allMsgs, summarizedBeforeId);
-  if (pendingMsgs.length === 0) return emptyPlan(startIndex);
+  if (pendingMsgs.length === 0) {
+    return emptyPlan(startIndex, {
+      pendingMsgCount: 0,
+      totalUsed: estimateTokenCount(summary || ""),
+      historyBudget: 0,
+      triggerRatio: resolveCompressionTriggerRatio(contextWindow),
+      realUsageRatio: normalizeContextUsageRatio(realContextUsagePercent),
+      triggeredByRealUsage: false,
+      triggeredByEstimate: false,
+      triggeredByColdResume: false,
+      shouldRunActiveSummary: false,
+      emptyReason: "no-pending"
+    });
+  }
   const summaryTokens = estimateTokenCount(summary || "");
   const pendingTokens = pendingMsgs.reduce((sum, msg) => sum + getMessageTokenCount(msg), 0);
   const totalUsed = summaryTokens + pendingTokens;
@@ -517,14 +559,47 @@ function planCompression(input) {
     coldResume,
     lastMsg: pendingMsgs[pendingMsgs.length - 1]
   });
-  if (!trigger) return emptyPlan(startIndex);
-  let compressCount = calculateCompressCount(pendingMsgs, rawMessageBudget, {
-    keepTailCount: shouldRunActiveSummary || triggeredByRealUsage || triggeredByColdResume,
+  const diagnostics = {
+    pendingMsgCount: pendingMsgs.length,
     totalUsed,
-    historyBudget
+    historyBudget,
+    triggerRatio,
+    realUsageRatio: normalizeContextUsageRatio(realContextUsagePercent),
+    triggeredByRealUsage,
+    triggeredByEstimate: normalizeContextUsageRatio(realContextUsagePercent) === void 0 && totalUsed >= historyBudget,
+    triggeredByColdResume,
+    shouldRunActiveSummary
+  };
+  if (!trigger) {
+    return emptyPlan(startIndex, {
+      ...diagnostics,
+      emptyReason: "not-triggered"
+    });
+  }
+  const { effectiveHistoryBudget, effectiveRawMessageBudget } = resolveEffectiveCompactionBudgets({
+    contextWindow,
+    triggerRatio,
+    historyBudget,
+    rawMessageBudget,
+    summaryTokens
   });
+  let compressCount = calculateCompressCount(
+    pendingMsgs,
+    effectiveRawMessageBudget,
+    {
+      keepTailCount: shouldRunActiveSummary || triggeredByRealUsage || triggeredByColdResume,
+      totalUsed,
+      historyBudget: effectiveHistoryBudget
+    }
+  );
   compressCount = guardToolChainBoundary(pendingMsgs, compressCount);
-  if (compressCount < MIN_COMPRESS_COUNT) return emptyPlan(startIndex);
+  if (compressCount < MIN_COMPRESS_COUNT) {
+    return emptyPlan(startIndex, {
+      ...diagnostics,
+      compressCount,
+      emptyReason: "below-min-compress-count"
+    });
+  }
   const msgsToCompress = pendingMsgs.slice(0, compressCount);
   const msgsToKeep = pendingMsgs.slice(compressCount);
   const newSummarizedBeforeId = msgsToCompress[msgsToCompress.length - 1].id;
@@ -534,7 +609,8 @@ function planCompression(input) {
     msgsToCompress,
     msgsToKeep,
     newSummarizedBeforeId,
-    startIndex
+    startIndex,
+    diagnostics: { ...diagnostics, compressCount }
   };
 }
 

package/{chunk-EXMHJN7C.js → chunk-EYAQT56M.js}

@@ -6,10 +6,10 @@ import {
 } from "./chunk-AGGIDPCZ.js";
 import {
   getModelConfig
-} from "./chunk-UJ63UXGE.js";
+} from "./chunk-OIVSQH2O.js";
 import {
   isPlatformHostedImageModel
-} from "./chunk-AJHEFRGJ.js";
+} from "./chunk-Q67DOFOO.js";
 import {
   asOptionalFiniteNumber
 } from "./chunk-WWC7I6TA.js";
@@ -31,9 +31,10 @@ var getModelPricing = (provider, modelName, _requestedServiceTier) => {
 };
 var getModelPricingForModel = (provider, modelName, model) => {
   if (!model.price) return null;
+  const rawCacheHit = typeof model.price.inputCacheHit === "number" ? model.price.inputCacheHit : model.price.cachingRead;
   return {
     inputPrice: model.price.input,
-    inputCacheHitPrice: typeof model.price.inputCacheHit === "number" ? model.price.inputCacheHit : 0,
+    inputCacheHitPrice: typeof rawCacheHit === "number" ? rawCacheHit : 0,
     outputPrice: model.price.output
   };
 };

package/{chunk-CJCJF5XB.js → chunk-FDTR4QJ6.js}

@@ -46,6 +46,7 @@ function extractMessage(error) {
 }
 function parseResetsInMs(text) {
   if (!text) return void 0;
+  const sanitized = text.replace(/\btrace[\s_-]?id\b[^)\n]*/gi, " ").replace(/\b[0-9a-f]{16,}\b/gi, " ");
   const unitMs = (n, unit) => {
     if (/^d/.test(unit)) return n * 24 * 3600 * 1e3;
     if (/^h/.test(unit)) return n * 3600 * 1e3;
@@ -54,12 +55,13 @@ function parseResetsInMs(text) {
     return NaN;
   };
   const re = /(\d+)\s*(d(?:ay|ays)?|h(?:r|rs|our|ours)?|m(?:in|ins|inute|inutes)?|s(?:ec|ecs|econd|econds)?)/gi;
+  const MAX_SINGLE_MATCH_MS = 7 * 24 * 3600 * 1e3;
   let totalMs = 0;
   let matched = false;
   let m;
-  while ((m = re.exec(text)) !== null) {
+  while ((m = re.exec(sanitized)) !== null) {
     const ms = unitMs(Number(m[1]), m[2]);
-    if (!Number.isFinite(ms)) continue;
+    if (!Number.isFinite(ms) || ms > MAX_SINGLE_MATCH_MS) continue;
     totalMs += ms;
     matched = true;
   }