fix: read anthropic 1m flag from live model cache state

src/hooks/context-window-monitor.test.ts

@@ -249,7 +249,9 @@ describe("context-window-monitor", () => {
 
   it("should use 1M limit when model cache flag is enabled", async () => {
     //#given
-    const hook = createContextWindowMonitorHook(ctx as never, true)
+    const hook = createContextWindowMonitorHook(ctx as never, {
+      anthropicContext1MEnabled: true,
+    })
     const sessionID = "ses_1m_flag"
 
     await hook.event({
@@ -286,7 +288,9 @@ describe("context-window-monitor", () => {
   it("should keep env var fallback when model cache flag is disabled", async () => {
     //#given
     process.env[ANTHROPIC_CONTEXT_ENV_KEY] = "true"
-    const hook = createContextWindowMonitorHook(ctx as never, false)
+    const hook = createContextWindowMonitorHook(ctx as never, {
+      anthropicContext1MEnabled: false,
+    })
     const sessionID = "ses_env_fallback"
 
     await hook.event({

src/hooks/context-window-monitor.ts

@@ -5,8 +5,12 @@ const ANTHROPIC_DISPLAY_LIMIT = 1_000_000
 const DEFAULT_ANTHROPIC_ACTUAL_LIMIT = 200_000
 const CONTEXT_WARNING_THRESHOLD = 0.70
 
-function getAnthropicActualLimit(anthropicContext1MEnabled: boolean): number {
-  return anthropicContext1MEnabled ||
+type ModelCacheStateLike = {
+  anthropicContext1MEnabled: boolean
+}
+
+function getAnthropicActualLimit(modelCacheState?: ModelCacheStateLike): number {
+  return (modelCacheState?.anthropicContext1MEnabled ?? false) ||
     process.env.ANTHROPIC_1M_CONTEXT === "true" ||
     process.env.VERTEX_ANTHROPIC_1M_CONTEXT === "true"
     ? 1_000_000
@@ -37,7 +41,7 @@ function isAnthropicProvider(providerID: string): boolean {
 
 export function createContextWindowMonitorHook(
   _ctx: PluginInput,
-  anthropicContext1MEnabled = false,
+  modelCacheState?: ModelCacheStateLike,
 ) {
   const remindedSessions = new Set<string>()
   const tokenCache = new Map<string, CachedTokenState>()
@@ -59,7 +63,7 @@ export function createContextWindowMonitorHook(
     const totalInputTokens = (lastTokens?.input ?? 0) + (lastTokens?.cache?.read ?? 0)
 
     const actualUsagePercentage =
-      totalInputTokens / getAnthropicActualLimit(anthropicContext1MEnabled)
+      totalInputTokens / getAnthropicActualLimit(modelCacheState)
 
     if (actualUsagePercentage < CONTEXT_WARNING_THRESHOLD) return
 

src/hooks/directory-agents-injector/hook.ts

@@ -29,10 +29,10 @@ interface EventInput {
 
 export function createDirectoryAgentsInjectorHook(
   ctx: PluginInput,
-  anthropicContext1MEnabled?: boolean,
+  modelCacheState?: { anthropicContext1MEnabled: boolean },
 ) {
   const sessionCaches = new Map<string, Set<string>>();
-  const truncator = createDynamicTruncator(ctx, anthropicContext1MEnabled);
+  const truncator = createDynamicTruncator(ctx, modelCacheState);
 
   const toolExecuteAfter = async (input: ToolExecuteInput, output: ToolExecuteOutput) => {
     const toolName = input.tool.toLowerCase();

src/hooks/directory-readme-injector/hook.ts

@@ -29,10 +29,10 @@ interface EventInput {
 
 export function createDirectoryReadmeInjectorHook(
   ctx: PluginInput,
-  anthropicContext1MEnabled?: boolean,
+  modelCacheState?: { anthropicContext1MEnabled: boolean },
 ) {
   const sessionCaches = new Map<string, Set<string>>();
-  const truncator = createDynamicTruncator(ctx, anthropicContext1MEnabled);
+  const truncator = createDynamicTruncator(ctx, modelCacheState);
 
   const toolExecuteAfter = async (input: ToolExecuteInput, output: ToolExecuteOutput) => {
     const toolName = input.tool.toLowerCase();

src/hooks/preemptive-compaction.test.ts

@@ -269,7 +269,9 @@ describe("preemptive-compaction", () => {
 
   it("should use 1M limit when model cache flag is enabled", async () => {
     //#given
-    const hook = createPreemptiveCompactionHook(ctx as never, true)
+    const hook = createPreemptiveCompactionHook(ctx as never, {
+      anthropicContext1MEnabled: true,
+    })
     const sessionID = "ses_1m_flag"
 
     await hook.event({
@@ -306,7 +308,9 @@ describe("preemptive-compaction", () => {
   it("should keep env var fallback when model cache flag is disabled", async () => {
     //#given
     process.env[ANTHROPIC_CONTEXT_ENV_KEY] = "true"
-    const hook = createPreemptiveCompactionHook(ctx as never, false)
+    const hook = createPreemptiveCompactionHook(ctx as never, {
+      anthropicContext1MEnabled: false,
+    })
     const sessionID = "ses_env_fallback"
 
     await hook.event({

src/hooks/preemptive-compaction.ts

@@ -2,8 +2,12 @@ import { log } from "../shared/logger"
 
 const DEFAULT_ACTUAL_LIMIT = 200_000
 
-function getAnthropicActualLimit(anthropicContext1MEnabled: boolean): number {
-  return anthropicContext1MEnabled ||
+type ModelCacheStateLike = {
+  anthropicContext1MEnabled: boolean
+}
+
+function getAnthropicActualLimit(modelCacheState?: ModelCacheStateLike): number {
+  return (modelCacheState?.anthropicContext1MEnabled ?? false) ||
     process.env.ANTHROPIC_1M_CONTEXT === "true" ||
     process.env.VERTEX_ANTHROPIC_1M_CONTEXT === "true"
     ? 1_000_000
@@ -47,7 +51,7 @@ type PluginInput = {
 
 export function createPreemptiveCompactionHook(
   ctx: PluginInput,
-  anthropicContext1MEnabled = false,
+  modelCacheState?: ModelCacheStateLike,
 ) {
   const compactionInProgress = new Set<string>()
   const compactedSessions = new Set<string>()
@@ -65,7 +69,7 @@ export function createPreemptiveCompactionHook(
 
     const actualLimit =
       isAnthropicProvider(cached.providerID)
-        ? getAnthropicActualLimit(anthropicContext1MEnabled)
+        ? getAnthropicActualLimit(modelCacheState)
         : DEFAULT_ACTUAL_LIMIT
 
     const lastTokens = cached.tokens

src/hooks/rules-injector/hook.ts

@@ -31,9 +31,9 @@ const TRACKED_TOOLS = ["read", "write", "edit", "multiedit"];
 
 export function createRulesInjectorHook(
   ctx: PluginInput,
-  anthropicContext1MEnabled?: boolean,
+  modelCacheState?: { anthropicContext1MEnabled: boolean },
 ) {
-  const truncator = createDynamicTruncator(ctx, anthropicContext1MEnabled);
+  const truncator = createDynamicTruncator(ctx, modelCacheState);
   const { getSessionCache, clearSessionCache } = createSessionCacheStore();
   const { processFilePathForInjection } = createRuleInjectionProcessor({
     workspaceDirectory: ctx.directory,

src/hooks/tool-output-truncator.ts

@@ -27,12 +27,12 @@ const TOOL_SPECIFIC_MAX_TOKENS: Record<string, number> = {
 }
 
 interface ToolOutputTruncatorOptions {
-  anthropicContext1MEnabled?: boolean
+  modelCacheState?: { anthropicContext1MEnabled: boolean }
   experimental?: ExperimentalConfig
 }
 
 export function createToolOutputTruncatorHook(ctx: PluginInput, options?: ToolOutputTruncatorOptions) {
-  const truncator = createDynamicTruncator(ctx, options?.anthropicContext1MEnabled)
+  const truncator = createDynamicTruncator(ctx, options?.modelCacheState)
   const truncateAll = options?.experimental?.truncate_all_tool_outputs ?? false
 
   const toolExecuteAfter = async (