billion-context-pi 0.1.37 → 0.1.39

package/dist/index.js

@@ -76,6 +76,7 @@ function createInitialState() {
   return {
     blocks: [],
     messageRefs: { byRaw: {}, byRef: {} },
+    tokenSnapshot: {},
     nudge: {
       lastPerMessageNudgeTokens: 0,
       lastNudgeShownTokens: 0,
@@ -245,11 +246,23 @@ function syncBlocks(messages, state) {
       byRaw: { ...state.messageRefs.byRaw },
       byRef: { ...state.messageRefs.byRef }
     },
+    // Snapshot is keyed by ref with primitive values — shallow copy suffices.
+    tokenSnapshot: { ...state.tokenSnapshot ?? {} },
     nudge: { ...state.nudge, anchors: { ...state.nudge.anchors } },
     stats: { ...state.stats },
     nextBlockId: state.nextBlockId,
     nextRunId: state.nextRunId
   };
+  const liveRefs = new Set(
+    messages.map((m) => result.messageRefs.byRaw[m.id]).filter((r) => typeof r === "string")
+  );
+  if (Object.keys(result.tokenSnapshot).length !== liveRefs.size) {
+    const pruned = {};
+    for (const [ref, n] of Object.entries(result.tokenSnapshot)) {
+      if (liveRefs.has(ref)) pruned[ref] = n;
+    }
+    result.tokenSnapshot = pruned;
+  }
   const consumedBlockIds = /* @__PURE__ */ new Set();
   for (const block of result.blocks) {
     for (const consumedId of block.directBlockIds) {
@@ -293,7 +306,8 @@ function defaultConfig(modelContextLimit, overrides = {}) {
       growthCap: 5e4,
       minGrowthFloor: 2e4,
       minGrowthRatio: 0.45,
-      emergencyThresholdPct: 0.95
+      emergencyThresholdPct: 0.95,
+      tier2GrowthMultiplier: 1.5
     },
     promotionThreshold: 5,
     truncate: { threshold: 0.95 },
@@ -723,7 +737,7 @@ var TAG_CLOSE = LT + "/acp" + GT;
 function acpTag(ref, tokens, type) {
   return TAG_OPEN + 'tokens="' + formatTokens(tokens) + '" type="' + type + '"' + GT + ref + TAG_CLOSE;
 }
-function renderMessage(message, map, countTokens, strategy) {
+function renderMessage(message, map, countTokens, strategy, snapshot = null) {
   const ref = refForRaw(map, message.id);
   if (!ref || ref === BLOCKED_REF) return message;
   if (strategy === "none") return message;
@@ -734,26 +748,33 @@ function renderMessage(message, map, countTokens, strategy) {
     "^" + escapeRegex(TAG_OPEN) + "[^>]*" + GT + escapeRegex(ref) + escapeRegex(TAG_CLOSE) + "\\n?"
   );
   const cleanText = (message.text || "").replace(ownTagRe, "");
-  const tokens = countTokens(cleanText);
+  const tokens = snapshot ? snapshot[ref] ?? (snapshot[ref] = countTokens(cleanText)) : countTokens(cleanText);
   const type = classifyType(message);
   const prefix = acpTag(ref, tokens, type) + "\n";
   if (!cleanText) return { ...message, text: prefix };
   return { ...message, text: prefix + cleanText };
 }
-function renderVisibleRefs(messages, state, countTokens = (text) => Math.ceil(text.length / 4), strategy = "all") {
+function renderWithSnapshot(messages, state, countTokens = (text) => Math.ceil(text.length / 4), strategy = "all") {
   const map = state.messageRefs;
-  return messages.map(
-    (message) => renderMessage(message, map, countTokens, strategy)
+  const snapshot = { ...state.tokenSnapshot ?? {} };
+  const rendered = messages.map(
+    (message) => renderMessage(message, map, countTokens, strategy, snapshot)
   );
+  return { messages: rendered, tokenSnapshot: snapshot };
 }
 function createRenderRefsNode(strategy) {
   return {
     name: "render-refs",
     run(io, ctx) {
-      return {
-        ...io,
-        messages: renderVisibleRefs(io.messages, io.state, ctx.countTokens, strategy)
-      };
+      const { messages, tokenSnapshot } = renderWithSnapshot(
+        io.messages,
+        io.state,
+        ctx.countTokens,
+        strategy
+      );
+      const prev = io.state.tokenSnapshot;
+      const changed = !prev || Object.keys(tokenSnapshot).length !== Object.keys(prev).length;
+      return changed ? { ...io, messages, state: { ...io.state, tokenSnapshot } } : { ...io, messages };
     }
   };
 }
@@ -949,6 +970,7 @@ function buildCompressibleRanges(messages, state, config, protectedZoneRefs, cou
       ref,
       refNum: rn,
       tokens: countTokens(msg.text ?? ""),
+      chars: (msg.text ?? "").length,
       isTool: isToolMessage(msg),
       isUser: msg.role === "user"
     });
@@ -969,6 +991,7 @@ function buildCompressibleRanges(messages, state, config, protectedZoneRefs, cou
         endRef: info.ref,
         count: 1,
         tokens: info.tokens,
+        chars: info.chars,
         toolPct: info.isTool ? 100 : 0,
         textPct: info.isTool ? 0 : 100
       };
@@ -976,6 +999,7 @@ function buildCompressibleRanges(messages, state, config, protectedZoneRefs, cou
       cur.endRef = info.ref;
       cur.count++;
       cur.tokens += info.tokens;
+      cur.chars = (cur.chars ?? 0) + info.chars;
       if (info.isTool) {
         cur.toolPct = Math.round((cur.toolPct * (cur.count - 1) + 100) / cur.count);
       } else {
@@ -1018,6 +1042,51 @@ function buildCompressibleRanges(messages, state, config, protectedZoneRefs, cou
     protected: protectedRanges
   };
 }
+function mergeBatch(batch) {
+  const first = batch[0];
+  const last = batch[batch.length - 1];
+  const count = batch.reduce((s, r) => s + r.count, 0);
+  const tokens = batch.reduce((s, r) => s + r.tokens, 0);
+  const chars = batch.reduce((s, r) => s + rangeChars(r), 0);
+  const toolPct = Math.round(
+    batch.reduce((s, r) => s + r.toolPct * r.count, 0) / count
+  );
+  const merged = {
+    startRef: first.startRef,
+    endRef: last.endRef,
+    count,
+    tokens,
+    chars,
+    toolPct,
+    textPct: 100 - toolPct
+  };
+  if (batch.some((r) => r.dangerous === true)) {
+    merged.dangerous = true;
+  }
+  return merged;
+}
+function rangeChars(r) {
+  return r.chars ?? r.tokens * 4;
+}
+function mergeRangesToThreshold(ranges, minChars) {
+  if (minChars <= 0 || ranges.length === 0) return ranges;
+  const result = [];
+  let batch = [];
+  let batchChars = 0;
+  for (const r of ranges) {
+    batch.push(r);
+    batchChars += rangeChars(r);
+    if (batchChars >= minChars) {
+      result.push(mergeBatch(batch));
+      batch = [];
+      batchChars = 0;
+    }
+  }
+  if (batch.length > 0) {
+    result.push(mergeBatch(batch));
+  }
+  return result;
+}
 function runPipeline(nodes, initial, ctx) {
   let io = initial;
   for (const node of nodes) {
@@ -1297,7 +1366,10 @@ var recommendNode = {
     const nothingToCompress = contextRanges.compressible.length === 0;
     const recommendation = {
       contextRanges,
-      recommendedRanges: contextRanges.compressible,
+      recommendedRanges: mergeRangesToThreshold(
+        contextRanges.compressible,
+        ctx.config.compress.minCompressRange
+      ),
       nothingToCompress
     };
     return { ...io, effects: { ...io.effects, recommendation } };
@@ -1323,6 +1395,7 @@ var nudgeNode = {
     if (baseline > 0 && ctx.tokenCount < baseline - nudgeGrowthTokens) {
       stamped.lastPerMessageNudgeTokens = ctx.tokenCount;
       stamped.lastNudgeShownTokens = 0;
+      stamped.lastShownByTier = {};
     }
     if (stamped.lastPerMessageNudgeTokens === 0) {
       stamped.lastPerMessageNudgeTokens = ctx.tokenCount;
@@ -1591,10 +1664,11 @@ function resolveAdaptiveGrowth(modelContextLimit, nudge) {
     )
   );
 }
-function pendingByTier(state, recommendation, countTokens) {
+function pendingByTier(state, recommendation, countTokens, minCompressRange) {
   const out = {};
-  const compressible = recommendation?.contextRanges.compressible ?? [];
-  out[1] = { pending: compressible.reduce((s, r) => s + r.tokens, 0), targetBlocks: [] };
+  const merged = recommendation?.recommendedRanges ?? [];
+  const effective = minCompressRange > 0 ? merged.filter((r) => (r.chars ?? r.tokens * 4) >= minCompressRange) : merged;
+  out[1] = { pending: effective.reduce((s, r) => s + r.tokens, 0), targetBlocks: [] };
   const active = activeBlocks(state);
   const t1 = active.filter((b) => b.tier === 1);
   const t2 = active.filter((b) => b.tier === 2);
@@ -1609,6 +1683,7 @@ function decideNudge(input) {
   const nudgeGrowthTokens = resolveAdaptiveGrowth(limit, config.nudge);
   const overLimit = usage >= config.nudge.maxContextLimitPct;
   const emergencyOverride = usage >= config.nudge.emergencyThresholdPct;
+  const pressure = overLimit || emergencyOverride;
   const baseline = state.nudge.lastPerMessageNudgeTokens;
   const hadPendingNudge = state.nudge.lastNudgeShownTokens > 0;
   const hasPendingNudge = hadPendingNudge;
@@ -1620,44 +1695,67 @@ function decideNudge(input) {
   );
   const growthSinceReference = tokenCount - growthReference;
   const rec = recommendation;
-  const tiers = pendingByTier(state, rec, countTokens);
+  const tiers = pendingByTier(
+    state,
+    rec,
+    countTokens,
+    config.compress.minCompressRange
+  );
+  const tier2Threshold = Math.round(
+    nudgeGrowthTokens * (config.nudge.tier2GrowthMultiplier ?? 1.5)
+  );
   let injectedTier = null;
   let injectedReason = "";
   const growthReady = growthSinceReference >= growthFloor;
-  if (!overLimit && growthReady) {
-    for (const tier of [1, 2, 3]) {
-      if (!config.tiers.enabled && tier > 1) break;
-      const info = tiers[tier];
-      if (!info || info.pending < nudgeGrowthTokens) continue;
-      const lastShown = state.nudge.lastShownByTier[tier] ?? 0;
-      const cadenceMet = lastShown === 0 || tokenCount - lastShown >= growthFloor;
-      if (!cadenceMet) continue;
-      injectedTier = tier;
-      injectedReason = tier === 1 ? `T1 compressible ${info.pending} >= ${nudgeGrowthTokens}, growth ${growthSinceReference}, usage ${Math.round(usage * 100)}%` : `T${tier} distill ready: ${info.targetBlocks.length} tier-${tier - 1} blocks (${info.pending} tokens) >= ${nudgeGrowthTokens}, usage ${Math.round(usage * 100)}%`;
-      break;
+  const t1Eff = tiers[1]?.pending ?? 0;
+  const t2Pen = tiers[2]?.pending ?? 0;
+  const t3Pen = tiers[3]?.pending ?? 0;
+  if (pressure) {
+    const candidates = [1];
+    if (config.tiers.enabled) {
+      candidates.push(2, 3);
+    }
+    let best = null;
+    let bestPending = 0;
+    for (const t of candidates) {
+      const p = tiers[t]?.pending ?? 0;
+      if (p > bestPending) {
+        bestPending = p;
+        best = t;
+      }
     }
-  } else if (overLimit) {
-    for (const tier of [1, 2, 3]) {
-      if (!config.tiers.enabled && tier > 1) break;
-      const info = tiers[tier];
-      if (!info || info.pending < config.compress.minCompressRange) continue;
-      injectedTier = tier;
-      injectedReason = emergencyOverride ? `EMERGENCY: usage ${Math.round(usage * 100)}% >= ${Math.round(config.nudge.emergencyThresholdPct * 100)}%, T${tier} pending ${info.pending}` : `OVER-LIMIT: usage ${Math.round(usage * 100)}% >= ${Math.round(config.nudge.maxContextLimitPct * 100)}%, T${tier} pending ${info.pending}`;
-      break;
+    if (best !== null && bestPending > 0) {
+      injectedTier = best;
+      const label = emergencyOverride ? "EMERGENCY" : "OVER-LIMIT";
+      injectedReason = best === 1 ? `${label} T1: max effective pending ${bestPending}, usage ${Math.round(usage * 100)}%` : `${label} T${best} distill: max pending ${bestPending} (T1 effective ${t1Eff}, T2 ${t2Pen}, T3 ${t3Pen}), usage ${Math.round(usage * 100)}%`;
+    }
+  } else if (growthReady) {
+    if (t1Eff >= nudgeGrowthTokens) {
+      injectedTier = 1;
+      injectedReason = `T1 effective ${t1Eff} >= ${nudgeGrowthTokens}, growth ${growthSinceReference}, usage ${Math.round(usage * 100)}%`;
+    } else if (config.tiers.enabled && t2Pen >= tier2Threshold && t2Pen > t1Eff) {
+      const lastShown = state.nudge.lastShownByTier[2] ?? 0;
+      const cadenceMet = lastShown === 0 || tokenCount - lastShown >= growthFloor;
+      if (cadenceMet) {
+        injectedTier = 2;
+        injectedReason = `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 && 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 = `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)}%`;
+      }
     }
   }
-  const shouldInject = injectedTier !== null || overLimit && (rec?.recommendedRanges?.length ?? 0) > 0;
+  const shouldInject = injectedTier !== null;
   let reason;
-  if (emergencyOverride && injectedTier !== null) {
-    reason = injectedReason;
-  } else if (emergencyOverride) {
-    reason = `EMERGENCY: usage ${Math.round(usage * 100)}% >= ${Math.round(config.nudge.emergencyThresholdPct * 100)}% (no compressible content)`;
-  } else if (overLimit && injectedTier !== null) {
-    reason = injectedReason;
-  } else if (overLimit) {
-    reason = `OVER-LIMIT: usage ${Math.round(usage * 100)}% >= ${Math.round(config.nudge.maxContextLimitPct * 100)}% (no compressible content)`;
-  } else if (injectedTier !== null) {
+  if (injectedTier !== null) {
     reason = injectedReason;
+  } else if (pressure) {
+    const label = emergencyOverride ? "EMERGENCY" : "OVER-LIMIT";
+    reason = `${label}: usage ${Math.round(usage * 100)}% but no tier has effective compressible content (T1 effective ${t1Eff}, T2 ${t2Pen}, T3 ${t3Pen}) \u2014 nudge suppressed to avoid offering ranges below minCompressRange`;
   } else {
     const tiersList = [1, 2, 3];
     const eligible = tiersList.filter((t) => config.tiers.enabled || t === 1);
@@ -1731,6 +1829,7 @@ function cloneState(state) {
       byRaw: { ...state.messageRefs.byRaw },
       byRef: { ...state.messageRefs.byRef }
     },
+    tokenSnapshot: { ...state.tokenSnapshot ?? {} },
     nudge: { ...state.nudge, anchors: { ...state.nudge.anchors } },
     stats: { ...state.stats },
     nextBlockId: state.nextBlockId,
@@ -2006,20 +2105,23 @@ ${lines.join("\n")}`;
 function renderNudgeText(decision, prompts = defaultPrompts) {
   const breakdownStr = formatBreakdown(decision.contextBreakdown);
   const rangesStr = formatRanges(decision.compressibleRanges, decision.protectedRanges ?? []);
+  const isEmergency = !!decision.breakdown?.emergencyOverride || !!decision.breakdown?.overLimit;
   if (decision.tier !== null && decision.tier >= 2) {
     const isT2 = decision.tier === 2;
     const targets = decision.tierTargetBlocks ?? [];
     const blockList = formatTierTargetBlocks(targets);
     const startId = targets[0]?.blockId ?? "b1";
     const endId = targets[targets.length - 1]?.blockId ?? "b5";
+    const voice = isEmergency ? "emergency" : "gentle";
+    const triggerLine = isEmergency ? `[EMERGENCY \u2014 TIER ${decision.tier} ${isT2 ? "DISTILLATION" : "CONDENSATION"}] Context limit reached \u2014 distill NOW into a denser summary to reclaim tokens.` : `[TIER ${decision.tier} ${isT2 ? "DISTILLATION" : "CONDENSATION"} TRIGGER]`;
     return {
-      voice: "gentle",
+      voice,
       text: [
         efficiencyNote(prompts),
         "",
         breakdownStr,
         "",
-        `[TIER ${decision.tier} ${isT2 ? "DISTILLATION" : "CONDENSATION"} TRIGGER]`,
+        triggerLine,
         isT2 ? `Your tier-1 compression summaries have accumulated. Distill them into a single denser tier-2 summary. Use block IDs as boundaries (startId and endId as bN). Any raw (uncompressed) messages sitting between the boundary blocks are absorbed into the tier-2 block as well \u2014 apply HOW TO COMPRESS to those raw messages and the TIER 2 distillation rules to the existing summaries, so the whole span is covered and nothing is lost.` : `Your tier-2 compression summaries have accumulated. Condense them further into a tier-3 ultra-condensed summary. Use block IDs as boundaries (startId and endId as bN). Any raw (uncompressed) messages sitting between the boundary blocks are absorbed into the tier-3 block as well \u2014 apply HOW TO COMPRESS to those raw messages and the TIER 3 condensation rules to the existing summaries, so the whole span is covered and nothing is lost.`,
         blockList,
         `Example: compress({ content: [{ startId: "${startId}", endId: "${endId}", summary: "..." }] })`,
@@ -2030,7 +2132,6 @@ function renderNudgeText(decision, prompts = defaultPrompts) {
       ].join("\n")
     };
   }
-  const isEmergency = !!decision.breakdown?.emergencyOverride || !!decision.breakdown?.overLimit;
   if (isEmergency) {
     return {
       voice: "emergency",
@@ -2601,7 +2702,20 @@ function resolveDelegate(adapter) {
     displayUsage: adapter.displayUsage ?? "separate"
   };
 }
-function resolveConfig(adapter, liveContextLimit) {
+function mergeCompress(global, provider, model) {
+  return {
+    maxContextLimit: model?.maxContextLimit ?? provider?.maxContextLimit ?? global?.maxContextLimit,
+    emergencyThresholdPercent: model?.emergencyThresholdPercent ?? provider?.emergencyThresholdPercent ?? global?.emergencyThresholdPercent,
+    nudgeGrowthTokens: model?.nudgeGrowthTokens ?? provider?.nudgeGrowthTokens ?? global?.nudgeGrowthTokens
+  };
+}
+function resolveCompress(compress, provider, modelId) {
+  if (!compress) return {};
+  const prov = provider ? compress.providers?.[provider] : void 0;
+  const model = prov && modelId ? prov.models?.[modelId] : void 0;
+  return mergeCompress(compress, prov, model);
+}
+function resolveConfig(adapter, liveContextLimit, provider, modelId) {
   const envLimit = process.env.ACP_MODEL_CONTEXT_LIMIT;
   const envLimitNum = envLimit ? Number(envLimit) : NaN;
   const FALLBACK_LIMIT = 15e4;
@@ -2611,14 +2725,14 @@ function resolveConfig(adapter, liveContextLimit) {
     preserveRecentMessages: adapter.preserveRecentMessages ?? 5,
     ...adapter.coreOverrides
   });
-  const c = adapter.compress;
-  if (c?.maxContextLimit !== void 0) config.nudge.maxContextLimitPct = parsePercent(c.maxContextLimit);
-  if (c?.emergencyThresholdPercent !== void 0) {
+  const c = resolveCompress(adapter.compress, provider, modelId);
+  if (c.maxContextLimit !== void 0) config.nudge.maxContextLimitPct = parsePercent(c.maxContextLimit);
+  if (c.emergencyThresholdPercent !== void 0) {
     const pct2 = parsePercent(c.emergencyThresholdPercent);
     config.nudge.emergencyThresholdPct = pct2;
     config.truncate.threshold = pct2;
   }
-  if (c?.nudgeGrowthTokens !== void 0) {
+  if (c.nudgeGrowthTokens !== void 0) {
     config.nudge.growthFloor = c.nudgeGrowthTokens;
     config.nudge.growthCap = c.nudgeGrowthTokens;
   }
@@ -2631,6 +2745,89 @@ function parsePercent(v) {
   return Number(s);
 }
 
+// src/density.ts
+var DENSITY_MIN = 0.5;
+var DENSITY_MAX = 2.5;
+var MIN_DELTA_EST = 50;
+var CONFIRM_RATIO = 0.2;
+var INITIAL_DENSITY = 1;
+var DensityEstimator = class {
+  models = /* @__PURE__ */ new Map();
+  /** 重置指定模型（模型切换/会话开始时调用）。 */
+  resetModel(modelId) {
+    this.models.delete(modelId);
+  }
+  /** 返回当前密度系数（未知模型返回初始 1）。 */
+  densityFor(modelId) {
+    return this.models.get(modelId)?.density ?? INITIAL_DENSITY;
+  }
+  /**
+   * 每轮 context 事件调用。realTotal 为 provider 真实 usage（可空），
+   * estTotal 为本地估算总 token。postCompression 为压缩刚发生标志。
+   */
+  update(modelId, realTotal, estTotal, postCompression = false) {
+    if (realTotal === null) return;
+    let est = this.models.get(modelId);
+    if (!est) {
+      est = {
+        density: INITIAL_DENSITY,
+        anchorReal: null,
+        anchorEst: null,
+        pendingDensity: null,
+        confirmCount: 0,
+        postCompressionSkip: false
+      };
+      this.models.set(modelId, est);
+    }
+    if (postCompression) {
+      est.postCompressionSkip = true;
+      return;
+    }
+    if (est.postCompressionSkip) {
+      est.postCompressionSkip = false;
+      est.anchorReal = realTotal;
+      est.anchorEst = estTotal;
+      est.pendingDensity = null;
+      est.confirmCount = 0;
+      return;
+    }
+    if (est.anchorReal === null || est.anchorEst === null) {
+      est.anchorReal = realTotal;
+      est.anchorEst = estTotal;
+      return;
+    }
+    const dReal = realTotal - est.anchorReal;
+    const dEst = estTotal - est.anchorEst;
+    if (dEst < MIN_DELTA_EST) return;
+    est.anchorReal = realTotal;
+    est.anchorEst = estTotal;
+    const instant = clamp(dReal / dEst, DENSITY_MIN, DENSITY_MAX);
+    if (est.pendingDensity === null) {
+      est.pendingDensity = instant;
+      est.confirmCount = 1;
+    } else if (Math.abs(instant - est.pendingDensity) / est.pendingDensity <= CONFIRM_RATIO) {
+      est.confirmCount += 1;
+    } else {
+      est.pendingDensity = instant;
+      est.confirmCount = 1;
+    }
+    if (est.confirmCount >= 2) {
+      est.density = est.pendingDensity;
+      est.confirmCount = 0;
+      est.pendingDensity = null;
+    }
+  }
+  /** 注入器：估算文本 token = defaultCountTokens × density。 */
+  estimateWithDensity(modelId, text) {
+    const d = this.densityFor(modelId);
+    if (d === 1) return defaultCountTokens(text);
+    return Math.round(defaultCountTokens(text) * d);
+  }
+};
+function clamp(v, lo, hi) {
+  return v < lo ? lo : v > hi ? hi : v;
+}
+
 // src/messages.ts
 var REF_TAG_SOURCE = "(?:<acp\\s[^>]*>m\\d{5}</acp>|\\[m\\d{1,5}\\])";
 var REF_TAG = new RegExp(`^${REF_TAG_SOURCE}\\s?\\n?`);
@@ -3136,6 +3333,7 @@ function mergeInitialState(parsed) {
   return {
     blocks: parsed.blocks ?? fresh.blocks,
     messageRefs: parsed.messageRefs ?? fresh.messageRefs,
+    tokenSnapshot: parsed.tokenSnapshot ?? fresh.tokenSnapshot,
     nudge: { ...fresh.nudge, ...parsed.nudge ?? {} },
     stats: { ...fresh.stats, ...parsed.stats ?? {} },
     nextBlockId: parsed.nextBlockId ?? fresh.nextBlockId,
@@ -3358,8 +3556,14 @@ function pruneOrphanRefs(state, messages) {
   }
 }
 function createRuntime(adapter) {
-  const core = createCore({ countTokens: defaultCountTokens });
+  const density = new DensityEstimator();
+  let countModelId = "default";
+  const core = createCore({
+    // 密度校准版 countTokens（Phase 2）：默认回落 defaultCountTokens（density=1）
+    countTokens: (text) => density.estimateWithDensity(countModelId, text)
+  });
   const store = new SessionStateStore();
+  const lastActiveBlockIds = /* @__PURE__ */ new Map();
   const locks = /* @__PURE__ */ new Map();
   let adapterRef = adapter;
   let promptsRef = defaultPrompts;
@@ -3384,7 +3588,8 @@ function createRuntime(adapter) {
     return m?.contextWindow ?? 0;
   }
   function configFor(ctx) {
-    return resolveConfig(adapterRef, liveContextLimit(ctx));
+    const m = ctx.model;
+    return resolveConfig(adapterRef, liveContextLimit(ctx), m?.provider, m?.id);
   }
   async function stateFor(ctx, liveMessages) {
     const sm = ctx.sessionManager;
@@ -3407,7 +3612,19 @@ function createRuntime(adapter) {
     const sm = ctx.sessionManager;
     await store.save(state, sm.getSessionFile() ?? void 0, sm.getSessionId());
   }
-  return { core, store, get adapter() {
+  function noteActiveBlocks(sid, activeBlockIds) {
+    const current = new Set(activeBlockIds);
+    const prev = lastActiveBlockIds.get(sid);
+    const isNew = prev !== void 0 && activeBlockIds.some((id) => !prev.has(id));
+    lastActiveBlockIds.set(sid, current);
+    return isNew;
+  }
+  function clearSessionTracking(sid) {
+    lastActiveBlockIds.delete(sid);
+  }
+  return { core, store, density, setCountModel: (m) => {
+    countModelId = m;
+  }, noteActiveBlocks, clearSessionTracking, get adapter() {
     return adapterRef;
   }, setAdapter: (a) => {
     adapterRef = a;
@@ -7834,6 +8051,9 @@ function estimateTokens(messages, coveredIds) {
   }
   return tokens;
 }
+function calibrateTokens(estimate, density) {
+  return density === 1 ? estimate : Math.round(estimate * density);
+}
 function lastUserMessageId(entries) {
   for (let i = entries.length - 1; i >= 0; i--) {
     const e = entries[i];
@@ -7842,6 +8062,23 @@ function lastUserMessageId(entries) {
   return void 0;
 }
 
+// src/compat.ts
+function normalizeSystemPrompt(input) {
+  if (input === void 0) return "";
+  if (Array.isArray(input)) return input.join("\n");
+  return input;
+}
+function formatSystemPromptForEvent(base, append) {
+  const normalized = normalizeSystemPrompt(base);
+  return `${normalized}
+
+${append}`;
+}
+function getSystemPromptText(ctx) {
+  const result = ctx.getSystemPrompt?.();
+  return normalizeSystemPrompt(result);
+}
+
 // src/compress-tool.ts
 function formatK2(n) {
   return n >= 1e3 ? `${(n / 1e3).toFixed(1)}K` : String(n);
@@ -7887,21 +8124,26 @@ async function handleCompress(args, runtime, ctx, toolCallId) {
   if (ranges.length === 0) return "No ranges provided.";
   const { state: initialState, coreMessages } = await runtime.stateFor(ctx);
   const config = runtime.configFor(ctx);
-  const estimatedTokens = estimateTokens(coreMessages, collectCoveredMessageIds(initialState));
-  const realUsage = ctx.getContextUsage?.();
+  const modelId = ctx.model?.id ?? "default";
+  const systemPromptText = getSystemPromptText(ctx);
+  const systemPromptTokens = systemPromptText ? defaultCountTokens(systemPromptText) : 0;
+  const sentTokens = estimateTokens(coreMessages, collectCoveredMessageIds(initialState)) + systemPromptTokens;
   const turn = runtime.core.processTurn({
     messages: coreMessages,
     state: initialState,
     config,
-    tokenCount: realUsage?.tokens && realUsage.tokens > 0 ? realUsage.tokens : estimatedTokens
+    tokenCount: calibrateTokens(sentTokens, runtime.density.densityFor(modelId))
   });
   const state = turn.state;
   const messages = turn.messages;
-  const beforeTokens = estimateTokens(messages, collectCoveredMessageIds(state));
+  const density = runtime.density.densityFor(modelId);
+  const beforeTokens = calibrateTokens(estimateTokens(messages, collectCoveredMessageIds(state)), density);
   const summaryMaxChars = args.summaryMaxChars;
   const topLevelTopic = args.topic;
   debug.event("compress-in", {
     sid: ctx.sessionManager.getSessionId(),
+    modelId,
+    density,
     ranges: ranges.length,
     spans: ranges.map((r) => ({ span: `${r.startId}..${r.endId}`, summaryLen: r.summary.length, summary: r.summary, topic: r.topic ?? topLevelTopic ?? null })),
     blocksBefore: state.blocks.length,
@@ -7958,7 +8200,8 @@ async function handleCompress(args, runtime, ctx, toolCallId) {
 
 // src/decompress-tool.ts
 import { writeFile, mkdir } from "fs/promises";
-import { resolve, relative, isAbsolute, join as join3 } from "path";
+import { existsSync as existsSync2, lstatSync, readlinkSync, realpathSync } from "fs";
+import { resolve, relative, isAbsolute, join as join3, basename as basename2, dirname as dirname3 } from "path";
 import { tmpdir, homedir as homedir2 } from "os";
 var AUTO_DIR = join3(homedir2() || tmpdir(), ".cache", "pi", "acp-decompress");
 var PREVIEW_CHARS = 600;
@@ -8002,14 +8245,39 @@ var ALLOWED_DIRS = [
 function resolveToFilePath(targetPath) {
   const expanded = targetPath.startsWith("~/") ? join3(homedir2(), targetPath.slice(2)) : targetPath;
   const resolved = resolve(expanded);
-  const isAllowed = ALLOWED_DIRS.some((dir) => {
-    const rel = relative(dir, resolved);
+  let probe = resolved;
+  const suffix = [];
+  while (!existsSync2(probe) && probe !== dirname3(probe)) {
+    suffix.unshift(basename2(probe));
+    probe = dirname3(probe);
+  }
+  const real = existsSync2(probe) ? realpathSync(probe) : probe;
+  let checked = real;
+  for (const part of suffix) {
+    checked = join3(checked, part);
+    try {
+      if (lstatSync(checked).isSymbolicLink()) {
+        const target = readlinkSync(checked);
+        checked = isAbsolute(target) ? resolve(target) : resolve(dirname3(checked), target);
+      }
+    } catch {
+    }
+  }
+  const allowed = ALLOWED_DIRS.map((d) => {
+    try {
+      return realpathSync(d);
+    } catch {
+      return d;
+    }
+  });
+  const isAllowed = allowed.some((dir) => {
+    const rel = relative(dir, checked);
     return rel === "" || !rel.startsWith("..") && !isAbsolute(rel);
   });
   if (!isAllowed) {
     return { error: `Error: toFile path must be under ${tmpdir()}, ~/.cache/opencode, or ~/.cache/pi. Got: ${targetPath}` };
   }
-  return resolved;
+  return checked;
 }
 function autoFilePath(blockId) {
   return join3(AUTO_DIR, `${blockId}-${Date.now()}.txt`);
@@ -8075,7 +8343,7 @@ ${text}`;
 }
 async function handleDecompress(args, runtime, ctx) {
   const { state, coreMessages } = await runtime.stateFor(ctx);
-  const arg = args.blockId.trim();
+  const arg = (args.blockId ?? "").trim();
   const owner = state.blocks.find((b) => b.effectiveMessageIds.includes(arg));
   if (owner) {
     return handleMessageRef(arg, owner.blockId, args, ctx);
@@ -8247,20 +8515,125 @@ function formatSize(tokens) {
   return `${(tokens / 1e6).toFixed(1)}M`;
 }
 
+// node_modules/billion-context-kit/dist/index.js
+var VIABLE_RANGE_MIN_TOKENS = 200;
+function viableRanges(ranges) {
+  return ranges.filter((r) => r.tokens >= VIABLE_RANGE_MIN_TOKENS);
+}
+function topicFallback(summary) {
+  const first = summary.split(/[.\n]/)[0] ?? "";
+  const t = first.trim().replace(/^["'`]+/, "").trim();
+  return t.length <= 30 ? t : `${t.slice(0, 30).trimEnd()}\u2026`;
+}
+function formatCompactTokens(count) {
+  if (count < 1e3) return count.toString();
+  if (count < 1e4) return `${(count / 1e3).toFixed(1)}k`;
+  if (count < 1e6) return `${Math.round(count / 1e3)}k`;
+  if (count < 1e7) return `${(count / 1e6).toFixed(1)}M`;
+  return `${Math.round(count / 1e6)}M`;
+}
+function bar(value, total, width = 20) {
+  if (total === 0) return "";
+  const filled = Math.max(0, Math.min(width, Math.round(value / total * width)));
+  return "\u2588".repeat(filled) + "\u2591".repeat(width - filled);
+}
+function buildStatusPanel(input) {
+  const { tokenCount, state, nudge, modelContextLimit } = input;
+  const fmt2 = input.fmtTokens ?? formatCompactTokens;
+  const bd = nudge?.contextBreakdown;
+  const limit = modelContextLimit;
+  const classified = bd ? bd.system + bd.tool + bd.summaries + bd.code + bd.text : 0;
+  const systemPromptTokens = input.systemPromptTokens;
+  const sentTotal = classified + systemPromptTokens;
+  const sessionOnly = input.unprunedTokens !== void 0 ? Math.max(0, input.unprunedTokens - sentTotal) : 0;
+  const displayTotal = tokenCount;
+  const displayPct = limit > 0 ? Math.round(displayTotal / limit * 100) : 0;
+  const sentPct = limit > 0 ? Math.round(sentTotal / limit * 100) : 0;
+  const activeBlocksList = state.blocks.filter((b) => b.active);
+  const totalBlocksList = state.blocks;
+  const lines = [];
+  lines.push("\u256D\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u256E");
+  lines.push("\u2502           ACP Context Analysis              \u2502");
+  lines.push("\u2570\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u256F");
+  if (input.version) lines.push(input.version);
+  lines.push("");
+  lines.push(`Context (session accounting, host footer scale): ${displayPct}% (${fmt2(displayTotal)} / ${fmt2(limit)}) \u2014 never shrinks; includes compressed originals`);
+  if (nudge && bd) {
+    const growth = bd.growth;
+    if (growth > 0 && displayTotal > 0) {
+      lines.push(`Growth: +${fmt2(growth)} since last nudge`);
+    }
+    lines.push("");
+    lines.push(`Sent to LLM (after compression, est.): ${fmt2(sentTotal)}${limit > 0 ? ` (${sentPct}% of limit)` : ""}`);
+    if (input.unprunedTokens !== void 0 && sessionOnly > 0) {
+      lines.push(`Session-only (compressed originals, est.): ${fmt2(sessionOnly)} \u2014 pruned from every request; the footer/nudge still count them`);
+    }
+    lines.push("");
+    lines.push("Token Breakdown (sent view):");
+    const categories = [
+      { label: "Tool", value: bd.tool },
+      { label: "SysPrompt", value: systemPromptTokens },
+      { label: "Text", value: bd.text },
+      { label: "Code", value: bd.code },
+      { label: "Summaries", value: bd.summaries }
+    ];
+    for (const cat of categories) {
+      if (cat.value <= 0) continue;
+      const pct2 = sentTotal > 0 ? Math.round(cat.value / sentTotal * 100) : 0;
+      const b = bar(cat.value, sentTotal);
+      lines.push(`  ${cat.label.padEnd(10)} ${b} ${String(pct2).padStart(3)}%  ${fmt2(cat.value)}`);
+    }
+  }
+  lines.push("");
+  if (nudge) {
+    if (nudge.shouldInject) {
+      const tierInfo = nudge.tier ? ` [T${nudge.tier} distillation]` : "";
+      lines.push(`Nudge: ACTIVE${tierInfo} \u2014 ${nudge.reason}`);
+    } else {
+      lines.push(`Nudge: idle \u2014 ${nudge.reason}`);
+    }
+  }
+  const ranges = viableRanges(nudge?.compressibleRanges ?? []);
+  const protectedRanges = nudge?.protectedRanges ?? [];
+  if (ranges.length > 0 || protectedRanges.length > 0) {
+    lines.push("");
+    lines.push(formatRanges(ranges, protectedRanges));
+  }
+  if (activeBlocksList.length > 0) {
+    lines.push("");
+    lines.push(`Blocks: ${activeBlocksList.length} active / ${totalBlocksList.length} total (${fmt2(state.stats.tokensCompressed)} tokens compressed)`);
+    for (const b of activeBlocksList) {
+      const topic = b.topic ? `: ${b.topic}` : `: ${topicFallback(b.summary || "")}`;
+      const summaryTok = defaultCountTokens(b.summary || "");
+      const origTok = b.compressedTokens > 0 ? b.compressedTokens : summaryTok;
+      lines.push(`  [${b.blockId}] T${b.tier} ${fmt2(origTok)}\u2192${fmt2(summaryTok)}${topic}`);
+    }
+  } else if (totalBlocksList.length > 0) {
+    lines.push("");
+    lines.push(`Blocks: 0 active / ${totalBlocksList.length} total (${fmt2(state.stats.tokensCompressed)} tokens compressed)`);
+  } else {
+    lines.push("");
+    lines.push("Blocks: none (nothing compressed yet)");
+  }
+  lines.push("");
+  lines.push("Tag visibility: tags injected to LLM only (deep copy), not persisted in session, not shown in terminal.");
+  return lines.join("\n");
+}
+
 // src/delegate-tool.ts
 import {
   spawn
 } from "child_process";
-import { createWriteStream, existsSync as existsSync2 } from "fs";
+import { createWriteStream, existsSync as existsSync3 } from "fs";
 import { mkdir as mkdir2, mkdtemp, writeFile as writeFile2, rm, appendFile } from "fs/promises";
 import { tmpdir as tmpdir2 } from "os";
-import { dirname as dirname3, join as join4, resolve as resolvePath } from "path";
+import { dirname as dirname4, join as join4, resolve as resolvePath } from "path";
 
 // src/footer-status.ts
 var FOOTER_STATUS_KEY = "billion-context-pi";
 var ui;
 var lastFooterText = "";
-function formatCompactTokens(count) {
+function formatCompactTokens2(count) {
   if (count < 1e3) return count.toString();
   if (count < 1e4) return `${(count / 1e3).toFixed(1)}k`;
   if (count < 1e6) return `${Math.round(count / 1e3)}k`;
@@ -8277,7 +8650,7 @@ function updateFooterStatus() {
   let text;
   if (usage && usage.totalTokens > 0) {
     const costStr = usage.cost.total > 0 ? ` ($${usage.cost.total.toFixed(4)})` : "";
-    text = `sub-agents \u2191${formatCompactTokens(usage.input)} \u2193${formatCompactTokens(usage.output)}${costStr}`;
+    text = `sub-agents \u2191${formatCompactTokens2(usage.input)} \u2193${formatCompactTokens2(usage.output)}${costStr}`;
   }
   if ((text ?? "") === lastFooterText) return;
   lastFooterText = text ?? "";
@@ -8670,11 +9043,11 @@ function delegateSpawnOptions(cwd, env) {
 var PI_CLI_ENTRY_RE = /[\\/]pi-coding-agent[\\/]dist[\\/]cli\.js$/;
 var PI_PACKAGE_REL = join4("@earendil-works", "pi-coding-agent", "dist", "cli.js");
 function probeUpFromArgv(argv1) {
-  let dir = resolvePath(dirname3(argv1) || process.cwd());
+  let dir = resolvePath(dirname4(argv1) || process.cwd());
   for (; ; ) {
     const candidate = join4(dir, "node_modules", PI_PACKAGE_REL);
-    if (existsSync2(candidate)) return candidate;
-    const parent = dirname3(dir);
+    if (existsSync3(candidate)) return candidate;
+    const parent = dirname4(dir);
     if (parent === dir) return null;
     dir = parent;
   }
@@ -8699,7 +9072,7 @@ function resolvePiCliEntry(argv1, env = process.env, piHost = true) {
     const probed = probeUpFromArgv(argv1);
     if (probed) return probed;
     for (const candidate of piCliGlobalCandidates(env)) {
-      if (existsSync2(candidate)) return candidate;
+      if (existsSync3(candidate)) return candidate;
     }
     logWarn("delegate", { event: "cli-entry-unresolved", argv1, fallback: "argv[1]" });
   }
@@ -9499,13 +9872,16 @@ function makeStatusTool(runtime) {
 async function handleStatus(args, runtime, ctx) {
   const { state, coreMessages } = await runtime.stateFor(ctx);
   const config = runtime.configFor(ctx);
-  const tokenCount = estimateTokens(coreMessages, collectCoveredMessageIds(state));
-  const realUsage = ctx.getContextUsage?.();
+  const coveredIds = collectCoveredMessageIds(state);
+  const modelId = ctx.model?.id ?? "default";
+  const systemPromptText = getSystemPromptText(ctx);
+  const systemPromptTokens = systemPromptText ? defaultCountTokens(systemPromptText) : 0;
+  const sentTokens = estimateTokens(coreMessages, coveredIds) + systemPromptTokens;
   const turn = runtime.core.processTurn({
     messages: coreMessages,
     state,
     config,
-    tokenCount: realUsage?.tokens && realUsage.tokens > 0 ? realUsage.tokens : tokenCount
+    tokenCount: calibrateTokens(sentTokens, runtime.density.densityFor(modelId))
   });
   const processed = turn.messages;
   const base = buildStatusReport(turn.state, processed, defaultCountTokens, {
@@ -9517,7 +9893,7 @@ async function handleStatus(args, runtime, ctx) {
   });
   if (args.scope) return base;
   const nudge = turn.nudge;
-  const ranges = nudge?.compressibleRanges ?? [];
+  const ranges = viableRanges(nudge?.compressibleRanges ?? []);
   const protectedRanges = nudge?.protectedRanges ?? [];
   const extra = [];
   if (nudge) {
@@ -9548,23 +9924,6 @@ async function handleStatus(args, runtime, ctx) {
 ${extra.join("\n")}` : base;
 }
 
-// src/compat.ts
-function normalizeSystemPrompt(input) {
-  if (input === void 0) return "";
-  if (Array.isArray(input)) return input.join("\n");
-  return input;
-}
-function formatSystemPromptForEvent(base, append) {
-  const normalized = normalizeSystemPrompt(base);
-  return `${normalized}
-
-${append}`;
-}
-function getSystemPromptText(ctx) {
-  const result = ctx.getSystemPrompt?.();
-  return normalizeSystemPrompt(result);
-}
-
 // src/commands.ts
 function makeCommands(runtime) {
   return [
@@ -9632,106 +9991,35 @@ ${text}`);
     }
   ];
 }
-function fmtTokens(n) {
-  return formatCompactTokens(n);
-}
-function bar(value, total, width = 20) {
-  if (total === 0) return "";
-  const filled = Math.max(0, Math.min(width, Math.round(value / total * width)));
-  return "\u2588".repeat(filled) + "\u2591".repeat(width - filled);
-}
 async function statusReport(runtime, ctx) {
   const { state, coreMessages } = await runtime.stateFor(ctx);
   const config = runtime.configFor(ctx);
   const realUsage = ctx.getContextUsage?.();
-  const tokenCount = realUsage?.tokens && realUsage.tokens > 0 ? realUsage.tokens : defaultCountTokens(coreMessages.map((m) => m.text ?? "").join("\n"));
-  const turn = runtime.core.processTurn({ messages: coreMessages, state, config, tokenCount });
-  const nudge = turn.nudge;
-  const bd = nudge?.contextBreakdown;
-  const limit = config.modelContextLimit;
-  const classified = bd ? bd.system + bd.tool + bd.summaries + bd.code + bd.text : 0;
   const systemPromptText = getSystemPromptText(ctx);
   const systemPromptTokens = systemPromptText ? defaultCountTokens(systemPromptText) : 0;
-  const framework = bd ? Math.max(0, tokenCount - classified - systemPromptTokens) : 0;
-  const displayTotal = tokenCount;
-  const displayPct = limit > 0 ? Math.round(displayTotal / limit * 100) : 0;
-  const activeBlocksList = state.blocks.filter((b) => b.active);
-  const totalBlocksList = state.blocks;
-  const lines = [];
-  const versionStr = "0.1.37" ? `billion-context-pi@${"0.1.37"}` : "";
-  lines.push("\u256D\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u256E");
-  lines.push("\u2502           ACP Context Analysis              \u2502");
-  lines.push("\u2570\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u256F");
-  if (versionStr) lines.push(versionStr);
-  lines.push("");
-  lines.push(`Context: ${displayPct}% (${fmtTokens(displayTotal)} / ${fmtTokens(limit)})`);
-  if (nudge && bd) {
-    const growth = bd.growth;
-    if (growth > 0 && displayTotal > 0) {
-      lines.push(`Growth: +${fmtTokens(growth)} since last nudge`);
-    }
-    if (displayTotal > 0) {
-      lines.push("");
-      lines.push("Token Breakdown:");
-      const categories = [
-        { label: "Tool", value: bd.tool },
-        { label: "SysPrompt", value: systemPromptTokens },
-        { label: "Framework", value: framework },
-        { label: "Text", value: bd.text },
-        { label: "Code", value: bd.code },
-        { label: "Summaries", value: bd.summaries }
-      ];
-      for (const cat of categories) {
-        if (cat.value <= 0) continue;
-        const pct2 = displayTotal > 0 ? Math.round(cat.value / displayTotal * 100) : 0;
-        const b = bar(cat.value, displayTotal);
-        lines.push(`  ${cat.label.padEnd(10)} ${b} ${String(pct2).padStart(3)}%  ${fmtTokens(cat.value)}`);
-      }
-    }
-  }
-  lines.push("");
-  if (nudge) {
-    if (nudge.shouldInject) {
-      const tierInfo = nudge.tier ? ` [T${nudge.tier} distillation]` : "";
-      lines.push(`Nudge: ACTIVE${tierInfo} \u2014 ${nudge.reason}`);
-    } else {
-      lines.push(`Nudge: idle \u2014 ${nudge.reason}`);
-    }
-  }
-  const ranges = nudge?.compressibleRanges ?? [];
-  const protectedRanges = nudge?.protectedRanges ?? [];
-  if (ranges.length > 0 || protectedRanges.length > 0) {
-    lines.push("");
-    lines.push(formatRanges(ranges, protectedRanges));
-  }
-  if (activeBlocksList.length > 0) {
-    lines.push("");
-    lines.push(`Blocks: ${activeBlocksList.length} active / ${totalBlocksList.length} total (${fmtTokens(state.stats.tokensCompressed)} tokens compressed)`);
-    for (const b of activeBlocksList) {
-      const topic = b.topic ? `: ${b.topic}` : "";
-      const summaryTok = defaultCountTokens(b.summary || "");
-      const origTok = b.compressedTokens > 0 ? b.compressedTokens : summaryTok;
-      lines.push(`  [${b.blockId}] T${b.tier} ${fmtTokens(origTok)}\u2192${fmtTokens(summaryTok)}${topic}`);
-    }
-  } else if (totalBlocksList.length > 0) {
-    lines.push("");
-    lines.push(`Blocks: 0 active / ${totalBlocksList.length} total (${fmtTokens(state.stats.tokensCompressed)} tokens compressed)`);
-  } else {
-    lines.push("");
-    lines.push("Blocks: none (nothing compressed yet)");
-  }
-  lines.push("");
+  const sessionTokens = realUsage?.tokens && realUsage.tokens > 0 ? realUsage.tokens : defaultCountTokens(coreMessages.map((m) => m.text ?? "").join("\n"));
+  const coveredIds = collectCoveredMessageIds(state);
+  const modelId = ctx.model?.id ?? "default";
+  const sentTokens = estimateTokens(coreMessages, coveredIds) + systemPromptTokens;
+  const turn = runtime.core.processTurn({ messages: coreMessages, state, config, tokenCount: calibrateTokens(sentTokens, runtime.density.densityFor(modelId)) });
+  const versionStr = "0.1.39" ? `billion-context-pi@${"0.1.39"}` : void 0;
+  let text = buildStatusPanel({
+    version: versionStr,
+    tokenCount: sessionTokens,
+    systemPromptTokens,
+    state: turn.state,
+    nudge: turn.nudge,
+    modelContextLimit: config.modelContextLimit,
+    unprunedTokens: coreMessages.reduce((sum, m) => sum + defaultCountTokens(m.text ?? ""), 0)
+  });
   const delegateUsage = getDelegateUsage();
   if (delegateUsage && delegateUsage.totalTokens > 0) {
-    lines.push("");
     const cost = delegateUsage.cost.total;
     const costStr = cost > 0 ? ` ($${cost.toFixed(4)})` : "";
-    lines.push("\u2500\u2500 Session delegate usage (excluded from main totals) \u2500\u2500");
-    lines.push(`Tokens: ${delegateUsage.input.toLocaleString()} in, ${delegateUsage.output.toLocaleString()} out (${delegateUsage.totalTokens.toLocaleString()} total)${costStr}`);
+    text += "\n\n\u2500\u2500 Session delegate usage (excluded from main totals) \u2500\u2500\n";
+    text += `Tokens: ${delegateUsage.input.toLocaleString()} in, ${delegateUsage.output.toLocaleString()} out (${delegateUsage.totalTokens.toLocaleString()} total)${costStr}`;
   }
-  lines.push("");
-  lines.push("Tag visibility: tags injected to LLM only (deep copy), not persisted in session, not shown in terminal.");
-  return lines.join("\n");
+  return text;
 }
 
 // src/system-prompt.ts
@@ -9927,7 +10215,7 @@ function wireToolGuardrails(pi, runtime) {
 
 // src/update.ts
 import { readFile, writeFile as writeFile3, mkdir as mkdir3 } from "fs/promises";
-import { join as join5, dirname as dirname4 } from "path";
+import { join as join5, dirname as dirname5 } from "path";
 import { fileURLToPath } from "url";
 import { execFile } from "child_process";
 import { homedir as homedir3 } from "os";
@@ -9960,7 +10248,7 @@ async function readLastCheck() {
 }
 async function writeLastCheck(timestamp) {
   try {
-    await mkdir3(dirname4(THROTTLE_FILE), { recursive: true });
+    await mkdir3(dirname5(THROTTLE_FILE), { recursive: true });
     await writeFile3(THROTTLE_FILE, String(timestamp), "utf-8");
   } catch {
   }
@@ -9974,20 +10262,20 @@ async function readPackageJson(path4) {
   }
 }
 function findNpmRoot(extDir) {
-  let dir = dirname4(extDir);
+  let dir = dirname5(extDir);
   for (; ; ) {
-    if (dir.endsWith("node_modules")) return dirname4(dir);
-    const parent = dirname4(dir);
+    if (dir.endsWith("node_modules")) return dirname5(dir);
+    const parent = dirname5(dir);
     if (parent === dir) return void 0;
     dir = parent;
   }
 }
 async function findExtensionDir() {
-  let dir = dirname4(fileURLToPath(import.meta.url));
+  let dir = dirname5(fileURLToPath(import.meta.url));
   for (; ; ) {
     const pkg = await readPackageJson(join5(dir, "package.json"));
     if (pkg?.name === PACKAGE_NAME) return dir;
-    const parent = dirname4(dir);
+    const parent = dirname5(dir);
     if (parent === dir) return void 0;
     dir = parent;
   }
@@ -10036,7 +10324,7 @@ async function checkForUpdate(autoUpdate, notify) {
     const data = await res.json();
     const latest = data.version;
     if (!latest) return;
-    const current = runtimeVersion ?? "0.1.37";
+    const current = runtimeVersion ?? "0.1.39";
     const hasUpdate = isNewer(latest, current);
     debug.event("update-check", {
       current,
@@ -10072,7 +10360,7 @@ async function getRuntimeVersion() {
 
 // src/setup-subagent-tools.ts
 import { readFile as readFile2, writeFile as writeFile4, stat, copyFile, rename } from "fs/promises";
-import { existsSync as existsSync3 } from "fs";
+import { existsSync as existsSync4 } from "fs";
 import { homedir as homedir4 } from "os";
 import { join as join6 } from "path";
 import { CONFIG_DIR_NAME as CONFIG_DIR_NAME3 } from "@earendil-works/pi-coding-agent";
@@ -10137,7 +10425,7 @@ async function ensureSubagentAcpTools(settingsPath) {
     return { path: path4, action: "skipped", reason: "all builtin agents already have ACP tools" };
   }
   const backupPath = `${path4}.acp-bak`;
-  if (!existsSync3(backupPath)) {
+  if (!existsSync4(backupPath)) {
     try {
       await copyFile(path4, backupPath);
     } catch {
@@ -10284,10 +10572,14 @@ function wireSessionLifecycle(pi, runtime) {
   pi.on("session_start", async (_event, ctx) => {
     runtime.store.invalidate();
     runtime.clearNudgeTracking();
+    const modelId = ctx.model?.id ?? "default";
+    runtime.density.resetModel(modelId);
     resetDelegateUsage();
     setDelegateDisplayUsage("separate");
     const sid = ctx.sessionManager.getSessionId();
-    logInfo("session", { event: "start", sid, cwd: ctx.cwd, debug: runtime.adapter.debug ?? null, version: true ? "0.1.37" : null });
+    runtime.clearSessionTracking(sid);
+    const modelInfo = ctx.model;
+    logInfo("session", { event: "start", sid, cwd: ctx.cwd, debug: runtime.adapter.debug ?? null, version: true ? "0.1.39" : null, model: modelInfo?.id ?? null, modelApi: modelInfo?.api ?? null, contextWindow: modelInfo?.contextWindow ?? null });
     try {
       const user = await loadUserConfig(ctx.cwd);
       runtime.setAdapter(applyUserConfig(runtime.adapter, user));
@@ -10323,29 +10615,39 @@ function wireContextTransform(pi, runtime) {
     const sid = ctx.sessionManager.getSessionId();
     const release = await runtime.acquireLock(sid);
     try {
+      const modelId = ctx.model?.id ?? "default";
+      runtime.setCountModel(modelId);
       const { state, coreMessages, entries } = await runtime.stateFor(ctx, event.messages);
       const config = runtime.configFor(ctx);
       const coveredIds = collectCoveredMessageIds(state);
       const realUsage = ctx.getContextUsage?.();
-      const estimated = estimateTokens(coreMessages, coveredIds);
-      const tokenCount = realUsage?.tokens && realUsage.tokens > 0 ? realUsage.tokens : estimated;
+      const systemPromptText = getSystemPromptText(ctx);
+      const systemPromptTokens = systemPromptText ? defaultCountTokens(systemPromptText) : 0;
+      const sentTokens = estimateTokens(coreMessages, coveredIds) + systemPromptTokens;
+      const tokenCount = calibrateTokens(sentTokens, runtime.density.densityFor(modelId));
+      const postCompression = runtime.noteActiveBlocks(
+        sid,
+        state.blocks.filter((b) => b.active).map((b) => b.blockId)
+      );
       debug.event("context-in", {
         sid,
+        modelId,
+        density: runtime.density.densityFor(modelId),
         eventMsgs: event.messages?.length ?? 0,
         entries: entries.length,
         coreMsgs: coreMessages.length,
         tokenCount,
-        estimatedTokens: estimated,
-        realTokens: realUsage?.tokens ?? null,
-        realPercent: realUsage?.percent ?? null,
+        sessionTokens: realUsage?.tokens ?? null,
         limit: config.modelContextLimit,
         blocksBefore: state.blocks.length,
         activeBefore: state.blocks.filter((b) => b.active).length
       });
       const turn = runtime.core.processTurn({ messages: coreMessages, state, config, tokenCount });
       await runtime.save(turn.state, ctx);
+      runtime.density.update(modelId, realUsage?.tokens ?? null, sentTokens, postCompression);
       logInfo("turn", {
         sid,
+        model: ctx.model?.id ?? null,
         inMsgs: coreMessages.length,
         outMsgs: turn.messages.length,
         tokens: tokenCount,
@@ -10357,6 +10659,8 @@ function wireContextTransform(pi, runtime) {
         activeBlocks: turn.state.blocks.filter((b) => b.active).length
       });
       debug.event("processTurn", {
+        modelId,
+        density: runtime.density.densityFor(modelId),
         outMsgs: turn.messages.length,
         summaryMsgs: turn.messages.filter((m) => m.id.startsWith("acp_summary")).length,
         prunedMsgs: coreMessages.length - turn.messages.length + turn.messages.filter((m) => m.id.startsWith("acp_summary")).length,
@@ -10376,6 +10680,7 @@ function wireContextTransform(pi, runtime) {
       const debugOn2 = debug.enabled;
       if (turn.nudge?.shouldInject) {
         const emergency = turn.nudge.breakdown?.emergencyOverride === 1;
+        turn.nudge.compressibleRanges = viableRanges(turn.nudge.compressibleRanges);
         const turnKey = lastUserMessageId(entries) ?? sid;
         const alreadyShown = !emergency && runtime.nudgeShownFor(turnKey);
         if (!alreadyShown) {