@askalf/dario 5.2.4 → 5.2.5

package/dist/cc-template.d.ts

@@ -475,8 +475,27 @@ export declare const CC_CACHE_CONTROL: CacheControl;
  * without the enabling beta is not a shape real CC produces, and forwarding
  * half of it risks an upstream 400. DARIO_CACHE_TTL_5M=1 restores the
  * pre-fix behavior (always bare 5m) as the operator escape hatch.
+ *
+ * DARIO_CACHE_TTL_1H=1 is the opposite override: force `ttl:'1h'` on every
+ * breakpoint regardless of what the client sent, for a client that can't
+ * emit the 1h stamp itself (an SDK/agent harness that only stamps bare 5m —
+ * dario#678). The proxy adds the enabling `extended-cache-ttl-` beta to the
+ * outbound set so the 1h is honored. Deliberate override of the mirror
+ * guardrail: 1h cache *writes* bill ~2× the 5m rate, so it only wins when
+ * idle gaps routinely exceed the 5-minute window; on rapid back-to-back
+ * turns it costs more. 5M takes precedence if both are set.
  */
 export declare function effectiveCacheControl(clientBody: Record<string, unknown>, clientBeta?: string): CacheControl;
+/** The anthropic-beta flag that enables the 1-hour prompt-cache TTL. */
+export declare const EXTENDED_CACHE_TTL_BETA = "extended-cache-ttl-2025-04-11";
+/**
+ * When DARIO_CACHE_TTL_1H forces the 1h stamp, the outbound beta set must also
+ * carry `extended-cache-ttl-` or Anthropic ignores the ttl. Add it (idempotent)
+ * unless DARIO_CACHE_TTL_5M overrides (5M wins, matching effectiveCacheControl).
+ * Pure — `env` is injectable for tests. Returns `beta` unchanged when the flag
+ * is off or the beta is already present.
+ */
+export declare function withForced1hBeta(beta: string, env?: Record<string, string | undefined>): string;
 /**
  * Place CC-style prompt-cache breakpoints on the conversation. The system
  * prompt is already cached at build time (2 system breakpoints); this adds a

package/dist/cc-template.js

@@ -1260,10 +1260,21 @@ export const CC_CACHE_CONTROL = { type: 'ephemeral' };
  * without the enabling beta is not a shape real CC produces, and forwarding
  * half of it risks an upstream 400. DARIO_CACHE_TTL_5M=1 restores the
  * pre-fix behavior (always bare 5m) as the operator escape hatch.
+ *
+ * DARIO_CACHE_TTL_1H=1 is the opposite override: force `ttl:'1h'` on every
+ * breakpoint regardless of what the client sent, for a client that can't
+ * emit the 1h stamp itself (an SDK/agent harness that only stamps bare 5m —
+ * dario#678). The proxy adds the enabling `extended-cache-ttl-` beta to the
+ * outbound set so the 1h is honored. Deliberate override of the mirror
+ * guardrail: 1h cache *writes* bill ~2× the 5m rate, so it only wins when
+ * idle gaps routinely exceed the 5-minute window; on rapid back-to-back
+ * turns it costs more. 5M takes precedence if both are set.
  */
 export function effectiveCacheControl(clientBody, clientBeta) {
     if (process.env['DARIO_CACHE_TTL_5M'] === '1')
         return CC_CACHE_CONTROL;
+    if (process.env['DARIO_CACHE_TTL_1H'] === '1')
+        return { type: 'ephemeral', ttl: '1h' };
     if (!clientBeta || !clientBeta.includes('extended-cache-ttl-'))
         return CC_CACHE_CONTROL;
     const scan = (blocks) => {
@@ -1289,6 +1300,22 @@ export function effectiveCacheControl(clientBody, clientBeta) {
     }
     return CC_CACHE_CONTROL;
 }
+/** The anthropic-beta flag that enables the 1-hour prompt-cache TTL. */
+export const EXTENDED_CACHE_TTL_BETA = 'extended-cache-ttl-2025-04-11';
+/**
+ * When DARIO_CACHE_TTL_1H forces the 1h stamp, the outbound beta set must also
+ * carry `extended-cache-ttl-` or Anthropic ignores the ttl. Add it (idempotent)
+ * unless DARIO_CACHE_TTL_5M overrides (5M wins, matching effectiveCacheControl).
+ * Pure — `env` is injectable for tests. Returns `beta` unchanged when the flag
+ * is off or the beta is already present.
+ */
+export function withForced1hBeta(beta, env = process.env) {
+    if (env['DARIO_CACHE_TTL_1H'] !== '1' || env['DARIO_CACHE_TTL_5M'] === '1')
+        return beta;
+    if (beta.split(',').includes(EXTENDED_CACHE_TTL_BETA))
+        return beta;
+    return beta.length > 0 ? beta + ',' + EXTENDED_CACHE_TTL_BETA : EXTENDED_CACHE_TTL_BETA;
+}
 /**
  * Place CC-style prompt-cache breakpoints on the conversation. The system
  * prompt is already cached at build time (2 system breakpoints); this adds a

package/dist/proxy.js

@@ -9,7 +9,7 @@ import { arch, platform } from 'node:process';
 import { getAccessToken, getStatus } from './oauth.js';
 import { buildHealthResponse, derivePoolStatus, shouldDiscloseHealthInternals } from './health-response.js';
 import { darioVersion } from './version.js';
-import { buildCCRequest, applyCcPromptCaching, parseEffortSuffix, reverseMapResponse, createStreamingReverseMapper, orderHeadersForOutbound, isMcpToolName, CC_TEMPLATE, effectiveCacheControl } from './cc-template.js';
+import { buildCCRequest, applyCcPromptCaching, parseEffortSuffix, reverseMapResponse, createStreamingReverseMapper, orderHeadersForOutbound, isMcpToolName, CC_TEMPLATE, effectiveCacheControl, withForced1hBeta } from './cc-template.js';
 import { stampCch, hasCchSeed } from './cch.js';
 import { describeTemplate, detectDrift, checkCCCompat } from './live-fingerprint.js';
 import { AccountPool, computeStickyKey, parseRateLimits, modelFamily, isInAuthCooldown, authCooldownMs, reconcilePoolAccounts } from './pool.js';
@@ -2399,6 +2399,12 @@ export async function startProxy(opts = {}) {
                     if (toAdd.length > 0)
                         beta += ',' + toAdd.join(',');
                 }
+                // Forced 1h cache (DARIO_CACHE_TTL_1H): effectiveCacheControl stamps
+                // ttl:'1h', but the 1h is only honored WITH the extended-cache-ttl
+                // beta — add it here (no-op unless the flag is set). If the upstream
+                // 400s the flag on a non-sub account, the rejected-set strip below
+                // drops it on the retry.
+                beta = withForced1hBeta(beta);
                 // Strip any beta flags the upstream has previously rejected on this
                 // account so we don't re-pay the 400 round-trip (dario#42 afk-mode
                 // fallout: captured templates carry tier-gated flags whose availability

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@askalf/dario",
-  "version": "5.2.4",
+  "version": "5.2.5",
   "description": "Use your Claude Pro/Max subscription in any tool — Cursor, Cline, Aider, the Agent SDK, your scripts — at subscription pricing, not per-token API bills. One local Anthropic + OpenAI-compatible endpoint.",
   "type": "module",
   "bin": {