@askalf/dario 3.30.0 → 3.30.1

package/dist/cc-template.d.ts

@@ -166,9 +166,8 @@ export interface RequestContext {
  * Replaces the entire request structure — tools, fields, ordering — with
  * what real CC sends. Only the conversation content is preserved.
  */
-export declare function buildCCRequest(clientBody: Record<string, unknown>, billingTag: string, cache1h: {
+export declare function buildCCRequest(clientBody: Record<string, unknown>, billingTag: string, cacheControl: {
     type: 'ephemeral';
-    ttl: '1h';
 }, identity: {
     deviceId: string;
     accountUuid: string;

package/dist/cc-template.js

@@ -688,7 +688,7 @@ const TOOL_MAP = {
  * Replaces the entire request structure — tools, fields, ordering — with
  * what real CC sends. Only the conversation content is preserved.
  */
-export function buildCCRequest(clientBody, billingTag, cache1h, identity, opts = {}) {
+export function buildCCRequest(clientBody, billingTag, cacheControl, identity, opts = {}) {
     const model = clientBody.model || 'claude-sonnet-4-6';
     const isHaiku = model.toLowerCase().includes('haiku');
     const messages = clientBody.messages || [];
@@ -931,8 +931,8 @@ export function buildCCRequest(clientBody, billingTag, cache1h, identity, opts =
         messages,
         system: [
             { type: 'text', text: billingTag },
-            { type: 'text', text: CC_AGENT_IDENTITY, cache_control: cache1h },
-            { type: 'text', text: fullSystemPrompt, cache_control: cache1h },
+            { type: 'text', text: CC_AGENT_IDENTITY, cache_control: cacheControl },
+            { type: 'text', text: fullSystemPrompt, cache_control: cacheControl },
         ],
     };
     // Tools come before metadata in CC's key order.
@@ -949,12 +949,12 @@ export function buildCCRequest(clientBody, billingTag, cache1h, identity, opts =
             session_id: identity.sessionId,
         }),
     };
-    ccRequest.max_tokens = 64000;
+    ccRequest.max_tokens = 32000;
     // Model-specific fields — order: thinking, context_management, output_config
     if (!isHaiku) {
         ccRequest.thinking = { type: 'adaptive' };
         ccRequest.context_management = { edits: [{ type: 'clear_thinking_20251015', keep: 'all' }] };
-        ccRequest.output_config = { effort: 'medium' };
+        ccRequest.output_config = { effort: 'high' };
     }
     ccRequest.stream = stream;
     // Replay the captured top-level key order. The hardcoded build order above

package/dist/proxy.js

@@ -878,8 +878,8 @@ export async function startProxy(opts = {}) {
                         const buildTag = computeBuildTag(userMsg, cliVersion);
                         const cch = computeCch();
                         const fullVersion = `${cliVersion}.${buildTag}`;
-                        const billingTag = `x-anthropic-billing-header: cc_version=${fullVersion}; cc_entrypoint=cli; cch=${cch};`;
-                        const CACHE_1H = { type: 'ephemeral', ttl: '1h' };
+                        const billingTag = `x-anthropic-billing-header: cc_version=${fullVersion}; cc_entrypoint=sdk-cli; cch=${cch};`;
+                        const CACHE_EPHEMERAL = { type: 'ephemeral' };
                         // Session stickiness: rebind the pre-selected pool account to
                         // whatever the sticky-key resolver picks. If this is a new
                         // conversation the key binds to the current best account
@@ -920,7 +920,7 @@ export async function startProxy(opts = {}) {
                         const bodyIdentity = poolAccount
                             ? poolAccount.identity
                             : { deviceId: identity.deviceId, accountUuid: identity.accountUuid, sessionId: preBodySessionId };
-                        const { body: ccBody, toolMap, detectedClient } = buildCCRequest(r, billingTag, CACHE_1H, bodyIdentity, {
+                        const { body: ccBody, toolMap, detectedClient } = buildCCRequest(r, billingTag, CACHE_EPHEMERAL, bodyIdentity, {
                             preserveTools: opts.preserveTools ?? false,
                             hybridTools: opts.hybridTools ?? false,
                             noAutoDetect: opts.noAutoDetect ?? false,

package/dist/shim/runtime.cjs

@@ -145,8 +145,8 @@ function rewriteBody(bodyText, tmpl) {
   const billingTag = body.system[0];
   body.system = [
     billingTag,
-    { type: 'text', text: tmpl.agent_identity, cache_control: { type: 'ephemeral', ttl: '1h' } },
-    { type: 'text', text: tmpl.system_prompt,  cache_control: { type: 'ephemeral', ttl: '1h' } },
+    { type: 'text', text: tmpl.agent_identity, cache_control: { type: 'ephemeral' } },
+    { type: 'text', text: tmpl.system_prompt,  cache_control: { type: 'ephemeral' } },
   ];
   body.tools = tmpl.tools;
   return JSON.stringify(body);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@askalf/dario",
-  "version": "3.30.0",
+  "version": "3.30.1",
   "description": "A local LLM router. One endpoint, every provider — Claude subscriptions, OpenAI, OpenRouter, Groq, local LiteLLM, any OpenAI-compat endpoint — your tools don't need to change.",
   "type": "module",
   "bin": {