billion-context-pi 0.1.64 → 0.1.65-pr.358.149

package/dist/index.js

@@ -10,7 +10,7 @@ import { readFileSync as readFileSync2 } from "fs";
 import { homedir as homedir8 } from "os";
 import { join as join12 } from "path";
 
-// node_modules/acp-kernel/dist/chunk-JGNVAFFE.js
+// node_modules/acp-kernel/dist/chunk-6TAK7DSI.js
 import { createRequire } from "module";
 var require2 = createRequire(import.meta.url);
 function defaultCountTokens(text) {
@@ -19,6 +19,12 @@ function defaultCountTokens(text) {
   const cjkCount = cjk?.length ?? 0;
   return cjkCount + Math.ceil((text.length - cjkCount) / 4);
 }
+function thinkingTokenValue(thinking) {
+  return typeof thinking === "number" && Number.isFinite(thinking) && thinking > 0 ? thinking : 0;
+}
+function countMessageTokens(message, countTokens = defaultCountTokens) {
+  return countTokens(message.text ?? "") + thinkingTokenValue(message.thinkingTokens);
+}
 function estimateTokensFast(text) {
   if (!text) return 0;
   return Math.ceil(text.length / 4);
@@ -1564,7 +1570,8 @@ function renderMessage(message, map, countTokens, strategy, snapshot = null) {
     "^" + escapeRegex(TAG_OPEN) + "[^>]*" + GT + escapeRegex(ref) + escapeRegex(TAG_CLOSE) + "\\n?"
   );
   const cleanText = (message.text || "").replace(ownTagRe, "");
-  const tokens = snapshot ? snapshot[ref] ?? (snapshot[ref] = countTokens(cleanText)) : countTokens(cleanText);
+  const textTokens = snapshot ? snapshot[ref] ?? (snapshot[ref] = countTokens(cleanText)) : countTokens(cleanText);
+  const tokens = textTokens + thinkingTokenValue(message.thinkingTokens);
   const type = classifyType(message);
   const prefix = acpTag(ref, tokens, type) + "\n";
   if (!cleanText) return { ...message, text: prefix };
@@ -1683,7 +1690,7 @@ function computeProtectedRefs(messages, state, config, countTokens = estimateTex
     if (isNeverPreserveRecent(msg)) continue;
     const ref = state.messageRefs.byRaw[msg.id];
     if (!ref || ref === "BLOCKED") continue;
-    visible.push({ ref, tokens: countTokens(msg.text ?? "") });
+    visible.push({ ref, tokens: countMessageTokens(msg, countTokens) });
   }
   if (preserveN > 0) {
     for (const m2 of visible.slice(-preserveN)) {
@@ -1726,7 +1733,7 @@ function buildCompressibleRanges(messages, state, config, protectedZoneRefs, cou
       protectedMsgs.push({
         ref,
         gapBefore: skipSinceProtected,
-        tokens: countTokens(msg.text ?? ""),
+        tokens: countMessageTokens(msg, countTokens),
         tools: msg.toolName ? [msg.toolName] : []
       });
       skipSinceProtected = false;
@@ -1741,7 +1748,7 @@ function buildCompressibleRanges(messages, state, config, protectedZoneRefs, cou
     compressibleMsgs.push({
       ref,
       gapBefore: skipSinceCompressible,
-      tokens: countTokens(msg.text ?? ""),
+      tokens: countMessageTokens(msg, countTokens),
       chars: (msg.text ?? "").length,
       isTool: isToolMessage(msg),
       isUser: msg.role === "user"
@@ -2387,7 +2394,7 @@ function applySingleRange(input) {
   let compressedTokens = 0;
   for (const id of filteredIds) {
     const message = input.messages.find((entry) => entry.id === id);
-    compressedTokens += input.countTokens(message?.text ?? "");
+    compressedTokens += message ? countMessageTokens(message, input.countTokens) : 0;
   }
   for (const consumedId of consumedBlockIds) {
     const consumed = blockById(input.state, consumedId);
@@ -2594,6 +2601,9 @@ function decideNudge(input) {
   const growthReady = firstSightMassReady || growthSinceReference >= growthFloor;
   const t2Count = tiers[2]?.targetBlocks.length ?? 0;
   const t3Count = tiers[3]?.targetBlocks.length ?? 0;
+  const tierCountUsageFloor = config.nudge.minContextLimitPct;
+  const t2CountReady = t2Count >= config.tiers.tier2Trigger && usage >= tierCountUsageFloor;
+  const t3CountReady = t3Count >= config.tiers.tier3Trigger && usage >= tierCountUsageFloor;
   if (pressure) {
     const candidates = [1];
     if (config.tiers.enabled) {
@@ -2616,19 +2626,19 @@ function decideNudge(input) {
     if (t1Eff >= nudgeGrowthTokens) {
       injectedTier = 1;
       injectedReason = `T1 effective ${t1Eff} >= ${nudgeGrowthTokens}, growth ${growthSinceReference}, usage ${Math.round(usage * 100)}%`;
-    } else if (config.tiers.enabled && (t2Count >= config.tiers.tier2Trigger || t2Pen >= tier2Threshold && t2Pen > t1Eff)) {
+    } else if (config.tiers.enabled && (t2CountReady || t2Pen >= tier2Threshold && t2Pen > t1Eff)) {
       const lastShown = state.nudge.lastShownByTier[2] ?? 0;
       const cadenceMet = lastShown === 0 || tokenCount - lastShown >= growthFloor;
       if (cadenceMet) {
         injectedTier = 2;
-        injectedReason = t2Count >= config.tiers.tier2Trigger ? `T2 distill ready: ${t2Count} tier-1 blocks >= tier2Trigger ${config.tiers.tier2Trigger} (${t2Pen} tokens), usage ${Math.round(usage * 100)}%` : `T2 distill ready: ${tiers[2].targetBlocks.length} tier-1 blocks (${t2Pen} tokens) >= ${tier2Threshold} (1.5x) and > T1 effective ${t1Eff}, usage ${Math.round(usage * 100)}%`;
+        injectedReason = t2CountReady ? `T2 distill ready: ${t2Count} tier-1 blocks >= tier2Trigger ${config.tiers.tier2Trigger} (${t2Pen} tokens), usage ${Math.round(usage * 100)}%` : `T2 distill ready: ${tiers[2].targetBlocks.length} tier-1 blocks (${t2Pen} tokens) >= ${tier2Threshold} (1.5x) and > T1 effective ${t1Eff}, usage ${Math.round(usage * 100)}%`;
       }
-    } else if (config.tiers.enabled && (t3Count >= config.tiers.tier3Trigger || t3Pen >= tier2Threshold && t3Pen > t2Pen && t3Pen > t1Eff)) {
+    } else if (config.tiers.enabled && (t3CountReady || t3Pen >= tier2Threshold && t3Pen > t2Pen && t3Pen > t1Eff)) {
       const lastShown = state.nudge.lastShownByTier[3] ?? 0;
       const cadenceMet = lastShown === 0 || tokenCount - lastShown >= growthFloor;
       if (cadenceMet) {
         injectedTier = 3;
-        injectedReason = t3Count >= config.tiers.tier3Trigger ? `T3 condense ready: ${t3Count} tier-2 blocks >= tier3Trigger ${config.tiers.tier3Trigger} (${t3Pen} tokens), usage ${Math.round(usage * 100)}%` : `T3 condense ready: ${tiers[3].targetBlocks.length} tier-2 blocks (${t3Pen} tokens) >= ${tier2Threshold} (1.5x) and > T2 ${t2Pen} and > T1 effective ${t1Eff}, usage ${Math.round(usage * 100)}%`;
+        injectedReason = t3CountReady ? `T3 condense ready: ${t3Count} tier-2 blocks >= tier3Trigger ${config.tiers.tier3Trigger} (${t3Pen} tokens), usage ${Math.round(usage * 100)}%` : `T3 condense ready: ${tiers[3].targetBlocks.length} tier-2 blocks (${t3Pen} tokens) >= ${tier2Threshold} (1.5x) and > T2 ${t2Pen} and > T1 effective ${t1Eff}, usage ${Math.round(usage * 100)}%`;
       }
     }
   }
@@ -2645,9 +2655,14 @@ function decideNudge(input) {
   } else {
     const tiersList = [1, 2, 3];
     const eligible = tiersList.filter((t) => config.tiers.enabled || t === 1);
-    const countReady = (t) => t === 2 ? t2Count >= config.tiers.tier2Trigger : t === 3 ? t3Count >= config.tiers.tier3Trigger : false;
+    const countReadyUngated = (t) => t === 2 ? t2Count >= config.tiers.tier2Trigger : t === 3 ? t3Count >= config.tiers.tier3Trigger : false;
+    const countReady = (t) => countReadyUngated(t) && usage >= tierCountUsageFloor;
     const ready = eligible.filter((t) => (tiers[t]?.pending ?? 0) >= nudgeGrowthTokens).map((t) => `T${t} ${tiers[t].pending}`);
-    const readyCount = eligible.filter((t) => (tiers[t]?.pending ?? 0) < nudgeGrowthTokens && countReady(t)).map((t) => `T${t} ${t === 2 ? t2Count : t3Count} blocks (count)`);
+    const readyCount = eligible.filter(
+      (t) => (tiers[t]?.pending ?? 0) < nudgeGrowthTokens && countReadyUngated(t)
+    ).map(
+      (t) => `T${t} ${t === 2 ? t2Count : t3Count} blocks (count${usage >= tierCountUsageFloor ? "" : ", usage-gated"})`
+    );
     const readyAll = [...ready, ...readyCount];
     const readyHint = readyAll.length > 0 ? `, ready: ${readyAll.join(", ")}` : "";
     const blocked = eligible.filter(
@@ -2709,7 +2724,7 @@ function computeContextBreakdown(messages, total, growth, countTokens) {
   const count = countTokens ?? ((t) => Math.ceil(t.length / 4));
   let system = 0, tool = 0, summaries = 0, code = 0, text = 0;
   for (const msg of messages) {
-    const tokens = count(msg.text ?? "");
+    const tokens = countMessageTokens(msg, count);
     if (msg.text?.startsWith("[Compressed conversation section]")) {
       summaries += tokens;
     } else if (msg.contentType === "tool-call" || msg.contentType === "tool-result") {
@@ -2871,7 +2886,7 @@ function collectVisible(messages, state, countTokens) {
     if (coveredIds.has(message.id)) return;
     const ref = refForRaw(state.messageRefs, message.id);
     if (!ref) return;
-    const tokens = countTokens(message.text ?? "");
+    const tokens = countMessageTokens(message, countTokens);
     const tool = isToolMessage(message) ? message.toolName ?? (message.toolCallId ? toolCallNames.get(message.toolCallId) : void 0) ?? "tool" : "text";
     if (tokens > 0) visible.push({ ref, tokens, tool, index });
   });
@@ -4088,6 +4103,8 @@ function projectMessage(message, id) {
     }];
   }
   if (role === "assistant") {
+    const thinking = thinkingTokenCount(msg.content);
+    const thinkingField = thinking > 0 ? { thinkingTokens: thinking } : {};
     const calls = allToolCalls(msg.content);
     if (calls.length > 0) {
       const textParts = extractText(msg.content);
@@ -4096,9 +4113,9 @@ function projectMessage(message, id) {
         const argStr = stringifyArgs(call.arguments);
         const text2 = argStr && textParts ? `${textParts}
 ${argStr}` : argStr || textParts;
-        return [{ id, role: "assistant", contentType: "tool-call", toolName: call.name, toolCallId: call.id, text: text2 }];
+        return [{ id, role: "assistant", contentType: "tool-call", toolName: call.name, toolCallId: call.id, text: text2, ...thinkingField }];
       }
-      return calls.map((call) => {
+      return calls.map((call, i) => {
         const argStr = stringifyArgs(call.arguments);
         return {
           id: `${id}#${call.id}`,
@@ -4106,13 +4123,14 @@ ${argStr}` : argStr || textParts;
           contentType: "tool-call",
           toolName: call.name,
           toolCallId: call.id,
-          text: argStr || textParts
+          text: argStr || textParts,
+          ...i === 0 ? thinkingField : {}
         };
       });
     }
     const text = extractText(msg.content);
     if (!text.trim()) return [];
-    return [{ id, role: "assistant", contentType: "text", text }];
+    return [{ id, role: "assistant", contentType: "text", text, ...thinkingField }];
   }
   const customText = extractText(msg.content) || fallbackText(msg);
   return customText.length > 0 ? [{ id, role: "user", contentType: "text", text: customText }] : [];
@@ -4140,6 +4158,15 @@ function extractText(content) {
   }
   return parts.join("\n");
 }
+function thinkingTokenCount(content) {
+  if (!Array.isArray(content)) return 0;
+  const parts = [];
+  for (const block of content) {
+    const b2 = block;
+    if (b2.type === "thinking" && typeof b2.thinking === "string") parts.push(b2.thinking);
+  }
+  return parts.length > 0 ? defaultCountTokens(parts.join("\n")) : 0;
+}
 function stripRefTag(text) {
   return text.replace(REF_TAG, "").replace(TRAILING_REF_TAG, "");
 }
@@ -9769,6 +9796,7 @@ function estimateTokens(messages, coveredIds, imageTokensById) {
     if (m2.toolName === "compress") continue;
     if (coveredIds?.has(m2.id)) continue;
     tokens += defaultCountTokens(m2.text ?? "");
+    tokens += m2.thinkingTokens ?? 0;
     const img = imageTokensById?.get(m2.id);
     if (img) tokens += img;
   }
@@ -16833,12 +16861,13 @@ async function statusReport(runtime, ctx) {
   const systemPromptTokens = systemPromptText ? defaultCountTokens(systemPromptText) : 0;
   const imageTokens = collectImageTokens(entries, modelSupportsImages(ctx.model));
   const imageTokensTotal = [...imageTokens.values()].reduce((a, b2) => a + b2, 0);
-  const sessionTokens = !anchorStale && realUsage?.tokens && realUsage.tokens > 0 ? realUsage.tokens : defaultCountTokens(coreMessages.map((m2) => m2.text ?? "").join("\n")) + imageTokensTotal;
+  const thinkingTokensTotal = coreMessages.reduce((sum, m2) => sum + (m2.thinkingTokens ?? 0), 0);
+  const sessionTokens = !anchorStale && realUsage?.tokens && realUsage.tokens > 0 ? realUsage.tokens : defaultCountTokens(coreMessages.map((m2) => m2.text ?? "").join("\n")) + imageTokensTotal + thinkingTokensTotal;
   const coveredIds = collectCoveredMessageIds(state);
   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 = "0.1.64" ? `billion-context-pi@${"0.1.64"}` : void 0;
+  const versionStr = "0.1.65" ? `billion-context-pi@${"0.1.65"}` : void 0;
   let text = buildStatusPanel({
     version: versionStr,
     tokenCount: sessionTokens,
@@ -16846,7 +16875,7 @@ async function statusReport(runtime, ctx) {
     state: turn.state,
     nudge: turn.nudge,
     modelContextLimit: config.modelContextLimit,
-    unprunedTokens: coreMessages.reduce((sum, m2) => sum + defaultCountTokens(m2.text ?? "") + (imageTokens.get(m2.id) ?? 0), 0),
+    unprunedTokens: coreMessages.reduce((sum, m2) => sum + defaultCountTokens(m2.text ?? "") + (m2.thinkingTokens ?? 0) + (imageTokens.get(m2.id) ?? 0), 0),
     cacheUsages: cacheUsageSamples(entries ?? [])
   });
   const delegateUsage = getDelegateUsage();
@@ -17536,7 +17565,7 @@ async function autoInstallLatest(latest, extDirOverride) {
       "--no-fund",
       "--no-save"
     ];
-    const prevVersion = (await readPackageJson(join11(extDir, "package.json")))?.version ?? "0.1.64";
+    const prevVersion = (await readPackageJson(join11(extDir, "package.json")))?.version ?? "0.1.65";
     const { code, stderr } = await runNpmImpl(installArgs(latest), { cwd: npmDir, timeout: 6e4 });
     if (code !== 0) {
       logWarn("update", {
@@ -17554,7 +17583,7 @@ async function autoInstallLatest(latest, extDirOverride) {
     }
     const verify = await verifyInstall(npmDir, latest);
     if (!verify.ok) {
-      const rollbackTo = SEMVER_RE.test(prevVersion) ? prevVersion : "0.1.64";
+      const rollbackTo = SEMVER_RE.test(prevVersion) ? prevVersion : "0.1.65";
       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 });
@@ -17629,7 +17658,7 @@ async function checkForUpdate(autoUpdate, notify) {
     }
     const latest = await fetchLatestVersion(tag);
     if (!latest) return;
-    const current = runtimeVersion ?? "0.1.64";
+    const current = runtimeVersion ?? "0.1.65";
     const hasUpdate = isVersionNewer(latest, current);
     debug.event("update-check", {
       current,
@@ -17807,7 +17836,7 @@ 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 ? "0.1.64" : 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 ? "0.1.65" : null, model: modelInfo?.id ?? null, modelApi: modelInfo?.api ?? null, contextWindow: modelInfo?.contextWindow ?? null });
     try {
       await runtime.reloadConfig(ctx.cwd);
       const delegateCfg = resolveDelegate(runtime.adapter);