proxitor 0.9.0-beta.3 → 0.9.0-beta.4

package/README.md

@@ -314,16 +314,18 @@ curl http://localhost:8828/health
 Proxitor automatically logs cache token usage from upstream responses — both non-streaming JSON and streaming SSE. No configuration needed.
 
 ```
-[abc123] Cache read: 50000, write: 25000 tokens
-[def456] Cache: no cached tokens
+[abc123] Cache read: 50000, write: 25000 tokens (99.6% hit)
+[def456] Cache read: 1088 tokens (90.0% hit)
+[ghi789] Cache: no cached tokens
 ```
 
-Supports both provider formats:
+Supports all three provider formats:
 
 | Provider format | Fields |
-|---|---|
+| --- | --- |
 | Anthropic | `usage.cache_read_input_tokens` / `usage.cache_creation_input_tokens` |
 | OpenAI / OpenRouter | `usage.prompt_tokens_details.cached_tokens` / `cache_write_tokens` |
+| Responses API | `usage.input_tokens_details.cached_tokens` / `cache_write_tokens` |
 
 When both formats are present (e.g., OpenRouter relaying an Anthropic response), Anthropic fields take priority.
 

package/dist/cli.mjs

@@ -8,7 +8,7 @@ import { formatWithOptions, styleText } from "node:util";
 import { existsSync, mkdirSync, readFileSync, writeFileSync } from "node:fs";
 import { dirname, join, resolve, sep } from "node:path";
 import * as l$1 from "node:readline";
-import l__default from "node:readline";
+import f from "node:readline";
 import { createHash } from "node:crypto";
 import { createServer } from "node:net";
 import { STATUS_CODES, createServer as createServer$1 } from "node:http";
@@ -11463,7 +11463,7 @@ var V = class {
 					this.state = "cancel", this.close();
 				}, { once: true });
 			}
-			this.rl = l__default.createInterface({
+			this.rl = f.createInterface({
 				input: this.input,
 				tabSize: 2,
 				prompt: "",
@@ -19607,7 +19607,7 @@ const r = Object.create(null), i = (e) => globalThis.process?.env || import.meta
 		const e = i(true);
 		return Object.keys(e);
 	}
-}), t = typeof process < "u" && process.env && process.env.NODE_ENV || "", f = [
+}), t = typeof process < "u" && process.env && process.env.NODE_ENV || "", f$1 = [
 	["APPVEYOR"],
 	[
 		"AWS_AMPLIFY",
@@ -19699,7 +19699,7 @@ const r = Object.create(null), i = (e) => globalThis.process?.env || import.meta
 	]
 ];
 function b() {
-	if (globalThis.process?.env) for (const e of f) {
+	if (globalThis.process?.env) for (const e of f$1) {
 		const s = e[1] || e[0];
 		if (globalThis.process?.env[s]) return {
 			name: e[0].toLowerCase(),
@@ -21544,7 +21544,7 @@ async function runConfigMenu(client) {
 }
 //#endregion
 //#region src/version.ts
-const version = "0.9.0-beta.3";
+const version = "0.9.0-beta.4";
 //#endregion
 //#region src/commands/doctor.ts
 const DEFAULT_TIMEOUT_MS = 3e3;
@@ -24571,6 +24571,37 @@ const buildUpstreamReq = createMiddleware(async (c, next) => {
 });
 //#endregion
 //#region src/proxy/cache-logging.ts
+function applyOpenAIDetails(details, result) {
+	let found = false;
+	if (typeof details.cached_tokens === "number" && details.cached_tokens > 0) {
+		result.cacheRead = details.cached_tokens;
+		found = true;
+	}
+	if (typeof details.cache_write_tokens === "number" && details.cache_write_tokens > 0) {
+		result.cacheCreate = details.cache_write_tokens;
+		found = true;
+	}
+	return found;
+}
+function applyAnthropicUsage(usage, result) {
+	if (typeof usage.cache_read_input_tokens === "number" && usage.cache_read_input_tokens > 0) result.cacheRead = usage.cache_read_input_tokens;
+	if (typeof usage.cache_creation_input_tokens === "number" && usage.cache_creation_input_tokens > 0) result.cacheCreate = usage.cache_creation_input_tokens;
+	if (typeof usage.input_tokens === "number" && usage.input_tokens > 0) result.inputTokens = usage.input_tokens + result.cacheRead + result.cacheCreate;
+}
+function applyOpenAIUsage(usage, result) {
+	const promptDetails = usage.prompt_tokens_details;
+	if (typeof promptDetails === "object" && promptDetails !== null) applyOpenAIDetails(promptDetails, result);
+	if (result.cacheRead === 0 && result.cacheCreate === 0) {
+		const inputDetails = usage.input_tokens_details;
+		if (typeof inputDetails === "object" && inputDetails !== null) applyOpenAIDetails(inputDetails, result);
+	}
+	if (typeof usage.prompt_tokens === "number" && usage.prompt_tokens > 0) result.inputTokens = usage.prompt_tokens;
+	else if (typeof usage.input_tokens === "number" && usage.input_tokens > 0) result.inputTokens = usage.input_tokens;
+}
+function extractFromUsage(usage, result) {
+	if (typeof usage.cache_read_input_tokens === "number" || typeof usage.cache_creation_input_tokens === "number") applyAnthropicUsage(usage, result);
+	else applyOpenAIUsage(usage, result);
+}
 function extractCacheUsage(bodyText) {
 	try {
 		const parsed = JSON.parse(bodyText);
@@ -24579,59 +24610,29 @@ function extractCacheUsage(bodyText) {
 		if (typeof usage !== "object" || usage === null) return void 0;
 		const result = {
 			cacheRead: 0,
-			cacheCreate: 0
+			cacheCreate: 0,
+			inputTokens: 0
 		};
-		if (typeof usage.cache_read_input_tokens === "number") result.cacheRead = usage.cache_read_input_tokens;
-		if (typeof usage.cache_creation_input_tokens === "number") result.cacheCreate = usage.cache_creation_input_tokens;
-		const details = usage.prompt_tokens_details;
-		if (typeof details === "object" && details !== null) {
-			if (typeof details.cached_tokens === "number" && details.cached_tokens > 0 && result.cacheRead === 0) result.cacheRead = details.cached_tokens;
-			if (typeof details.cache_write_tokens === "number" && details.cache_write_tokens > 0 && result.cacheCreate === 0) result.cacheCreate = details.cache_write_tokens;
-		}
+		extractFromUsage(usage, result);
 		return result;
 	} catch {
 		return;
 	}
 }
-function applyAnthropicFields(u, result) {
-	let found = false;
-	if (typeof u.cache_read_input_tokens === "number" && u.cache_read_input_tokens > 0) {
-		result.cacheRead = u.cache_read_input_tokens;
-		found = true;
-	}
-	if (typeof u.cache_creation_input_tokens === "number" && u.cache_creation_input_tokens > 0) {
-		result.cacheCreate = u.cache_creation_input_tokens;
-		found = true;
-	}
-	return found;
-}
-function applyOpenAIFields(details, result) {
-	let found = false;
-	if (typeof details.cached_tokens === "number" && details.cached_tokens > 0) {
-		result.cacheRead = details.cached_tokens;
-		found = true;
-	}
-	if (typeof details.cache_write_tokens === "number" && details.cache_write_tokens > 0) {
-		result.cacheCreate = details.cache_write_tokens;
-		found = true;
-	}
-	return found;
-}
 function extractFromEvent(parsed, result) {
 	if (typeof parsed !== "object" || parsed === null) return false;
-	const usage = (parsed.message ?? parsed).usage;
+	const record = parsed;
+	const usage = (record.message ?? record.response ?? parsed).usage;
 	if (typeof usage !== "object" || usage === null) return false;
-	const u = usage;
-	let found = false;
-	found = applyAnthropicFields(u, result) || found;
-	const details = u.prompt_tokens_details;
-	if (typeof details === "object" && details !== null) found = applyOpenAIFields(details, result) || found;
-	return found;
+	const before = result.cacheRead + result.cacheCreate;
+	extractFromUsage(usage, result);
+	return result.cacheRead + result.cacheCreate > before;
 }
 function extractCacheUsageFromSSE(fullText) {
 	const result = {
 		cacheRead: 0,
-		cacheCreate: 0
+		cacheCreate: 0,
+		inputTokens: 0
 	};
 	let found = false;
 	for (const line of fullText.split("\n")) {
@@ -24648,7 +24649,8 @@ function formatCacheUsage(usage, reqId) {
 	const parts = [];
 	if (usage.cacheRead > 0) parts.push(`read: ${usage.cacheRead}`);
 	if (usage.cacheCreate > 0) parts.push(`write: ${usage.cacheCreate}`);
-	logger.info(withReq(reqId, parts.length > 0 ? `Cache ${parts.join(", ")} tokens` : "Cache: no cached tokens"));
+	const pct = usage.inputTokens > 0 && usage.cacheRead > 0 ? ` (${(usage.cacheRead / usage.inputTokens * 100).toFixed(1)}% hit)` : "";
+	logger.info(withReq(reqId, parts.length > 0 ? `Cache ${parts.join(", ")} tokens${pct}` : "Cache: no cached tokens"));
 }
 function createLoggingStream(contentType, reqId) {
 	const chunks = [];
@@ -24811,10 +24813,6 @@ function shouldInjectCacheControl(mode, modelName, path) {
 	if (mode === "always") return true;
 	return isAnthropicEndpoint(modelName, path);
 }
-const TTL_SECONDS = {
-	"5m": 300,
-	"1h": 3600
-};
 /**
 * Build cache_control value for injection.
 * Merges existing cache_control with configured TTL.
@@ -24823,7 +24821,7 @@ const TTL_SECONDS = {
 function buildCacheControl(existing, ttl, isAnthropic) {
 	const result = existing !== null && typeof existing === "object" && !Array.isArray(existing) ? { ...existing } : { type: "ephemeral" };
 	if (!("type" in result)) result.type = "ephemeral";
-	if (ttl && isAnthropic) result.ttl = TTL_SECONDS[ttl];
+	if (ttl && isAnthropic) result.ttl = ttl;
 	return result;
 }
 //#endregion