@askalf/dario 5.2.2 → 5.2.3

package/dist/analytics.d.ts

@@ -63,6 +63,30 @@ export declare function billingBucketFromClaim(claim: string | null | undefined)
  * non-subscription billing classification or the `unknown` sentinel below.
  */
 export declare const SUBSCRIPTION_CLAIMS: ReadonlySet<string>;
+/**
+ * One-line per-request usage summary for verbose (-v / -vv) logs.
+ *
+ * dario already parses `input_tokens` / `cache_read_input_tokens` /
+ * `cache_creation_input_tokens` off every response into analytics and the
+ * `--log-file`, but never printed them to the console — so anyone debugging
+ * subscription burn (dario#678) only ever saw the request body and the
+ * billing *bucket*, never the cache accounting that actually governs cost.
+ * This surfaces it next to the existing `billing:` line so a plain `-vv`
+ * capture is self-diagnosing.
+ *
+ * `cachedPct` = cache_read / (input + cache_read + cache_create): the share of
+ * *prompt* tokens served from cache rather than freshly billed. On a repeated
+ * prompt a LOW value means the prefix is being re-created (a >5-minute gap
+ * expired the 5m TTL, or the cached prefix changed) — the exact signal the
+ * cache-TTL discussion turns on. Output tokens are excluded from the ratio
+ * (they are never cacheable). Pure + total-zero-safe for unit testing.
+ */
+export declare function formatUsageLogLine(requestCount: number, u: {
+    inputTokens?: number;
+    outputTokens?: number;
+    cacheReadTokens?: number;
+    cacheCreateTokens?: number;
+}): string;
 /**
  * The sentinel `claim` dario assigns when a response carried no rate-limit
  * header at all (non-200s, stream aborts, early rejects — see `pool.ts`

package/dist/analytics.js

@@ -62,6 +62,33 @@ export const SUBSCRIPTION_CLAIMS = new Set([
     'five_hour_overage_included',
     'seven_day_overage_included',
 ]);
+/**
+ * One-line per-request usage summary for verbose (-v / -vv) logs.
+ *
+ * dario already parses `input_tokens` / `cache_read_input_tokens` /
+ * `cache_creation_input_tokens` off every response into analytics and the
+ * `--log-file`, but never printed them to the console — so anyone debugging
+ * subscription burn (dario#678) only ever saw the request body and the
+ * billing *bucket*, never the cache accounting that actually governs cost.
+ * This surfaces it next to the existing `billing:` line so a plain `-vv`
+ * capture is self-diagnosing.
+ *
+ * `cachedPct` = cache_read / (input + cache_read + cache_create): the share of
+ * *prompt* tokens served from cache rather than freshly billed. On a repeated
+ * prompt a LOW value means the prefix is being re-created (a >5-minute gap
+ * expired the 5m TTL, or the cached prefix changed) — the exact signal the
+ * cache-TTL discussion turns on. Output tokens are excluded from the ratio
+ * (they are never cacheable). Pure + total-zero-safe for unit testing.
+ */
+export function formatUsageLogLine(requestCount, u) {
+    const inp = u.inputTokens ?? 0;
+    const out = u.outputTokens ?? 0;
+    const cr = u.cacheReadTokens ?? 0;
+    const cc = u.cacheCreateTokens ?? 0;
+    const promptTotal = inp + cr + cc;
+    const pct = promptTotal > 0 ? Math.round((cr / promptTotal) * 100) : 0;
+    return `[dario] #${requestCount} usage: in=${inp} out=${out} cache_read=${cr} cache_create=${cc} (${pct}% of prompt from cache)`;
+}
 /**
  * The sentinel `claim` dario assigns when a response carried no rate-limit
  * header at all (non-200s, stream aborts, early rejects — see `pool.ts`

package/dist/proxy.js

@@ -13,7 +13,7 @@ import { buildCCRequest, applyCcPromptCaching, parseEffortSuffix, reverseMapResp
 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';
-import { Analytics, billingBucketFromClaim, SUBSCRIPTION_CLAIMS } from './analytics.js';
+import { Analytics, billingBucketFromClaim, formatUsageLogLine, SUBSCRIPTION_CLAIMS } from './analytics.js';
 import { OverageGuard, buildHaltErrorBody } from './overage-guard.js';
 import { notify as osNotify } from './notify.js';
 import { loadAllAccounts, loadAccount, refreshAccountToken, resyncLoginFromCredentialsIfStale, ensureLoginCredentialsInPool } from './accounts.js';
@@ -3217,6 +3217,11 @@ export async function startProxy(opts = {}) {
                     preserve_tools: preserveToolsEffective,
                     stream: true,
                 });
+                if (verbose)
+                    console.log(formatUsageLogLine(requestCount, {
+                        inputTokens: streamInputTokens, outputTokens: streamOutputTokens,
+                        cacheReadTokens: streamCacheReadTokens, cacheCreateTokens: streamCacheCreateTokens,
+                    }));
             }
             else {
                 // Buffer and forward
@@ -3280,6 +3285,8 @@ export async function startProxy(opts = {}) {
                     preserve_tools: preserveToolsEffective,
                     stream: false,
                 });
+                if (verbose && bufferedUsage)
+                    console.log(formatUsageLogLine(requestCount, bufferedUsage));
                 if (verbose)
                     console.log(`[dario] #${requestCount} ${upstream.status}`);
             }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@askalf/dario",
-  "version": "5.2.2",
+  "version": "5.2.3",
   "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": {