@relayplane/proxy 1.8.9 → 1.8.11

package/dist/standalone-proxy.js

@@ -84,6 +84,13 @@ const downgrade_js_1 = require("./downgrade.js");
 const agent_tracker_js_1 = require("./agent-tracker.js");
 const version_status_js_1 = require("./utils/version-status.js");
 const signup_nudge_js_1 = require("./signup-nudge.js");
+const star_nudge_js_1 = require("./star-nudge.js");
+const estimate_js_1 = require("./estimate.js");
+// Per-IP rate limit state for /v1/estimate (60 req/min per IP)
+const estimateRateMap = new Map();
+// Fix A: Purge expired rate-limit entries every 5 minutes to prevent memory leak.
+// Without this, IPs that make one request and disappear stay in the map forever.
+setInterval(() => (0, estimate_js_1.purgeExpiredRateLimitEntries)(estimateRateMap, Date.now()), 5 * 60 * 1000);
 const PROXY_VERSION = (() => {
     try {
         const pkgPath = path.join(__dirname, '..', 'package.json');
@@ -323,6 +330,8 @@ function sendCloudTelemetry(taskType, model, tokensIn, tokensOut, latencyMs, suc
         // Uses setImmediate to guarantee zero added latency on the response path —
         // the nudge prints to stderr only after the current I/O cycle completes.
         setImmediate(() => (0, signup_nudge_js_1.checkAndShowNudge)());
+        // Star nudge fires at 50 requests (separate from signup nudge at 100)
+        setImmediate(() => (0, star_nudge_js_1.checkAndShowStarNudge)());
     }
     catch {
         // Telemetry should never break the proxy
@@ -2484,8 +2493,9 @@ async function startProxy(config = {}) {
     // Load persistent history from disk
     loadHistoryFromDisk();
     (0, agent_tracker_js_1.loadAgentRegistry)();
-    // Check once at startup whether the signup nudge has already been shown
+    // Check once at startup whether the nudges have already been shown
     (0, signup_nudge_js_1.initNudge)();
+    (0, star_nudge_js_1.initStarNudge)();
     // Flush history on shutdown
     const handleShutdown = () => {
         (0, agent_tracker_js_1.flushAgentRegistry)();
@@ -3895,6 +3905,39 @@ async function startProxy(config = {}) {
             }
             return;
         }
+        // === Pre-flight cost estimation endpoint (Pro-tier) ===
+        if (req.method === 'POST' && (url === '/v1/estimate' || url.endsWith('/v1/estimate'))) {
+            log('Pre-flight estimate request');
+            // --- Per-IP rate limit: 60 requests/minute ---
+            // Fix B: Use only the raw socket address — never x-forwarded-for.
+            // x-forwarded-for is a client-controlled header and is trivially spoofed;
+            // any attacker can send "X-Forwarded-For: 1.2.3.4" to bypass per-IP limits.
+            // The socket remoteAddress reflects the actual TCP connection and cannot be faked.
+            const clientIp = req.socket?.remoteAddress ?? 'unknown';
+            const now = Date.now();
+            // Fix C: Delegate rate limit logic to the testable checkEstimateRateLimit() function
+            // (extracted in estimate.ts so it can be unit-tested in isolation).
+            const rateLimitResult = (0, estimate_js_1.checkEstimateRateLimit)(estimateRateMap, clientIp, now);
+            if (!rateLimitResult.allowed) {
+                res.writeHead(429, { 'Content-Type': 'application/json', 'Retry-After': '60' });
+                res.end(JSON.stringify({ error: 'rate_limit_exceeded', message: 'Too many estimate requests. Limit: 60/minute.' }));
+                return;
+            }
+            // --- Read body with size limit (uses existing MAX_BODY_SIZE helper) ---
+            let body;
+            try {
+                body = await readRequestBody(req);
+            }
+            catch (err) {
+                res.writeHead(413, { 'Content-Type': 'application/json' });
+                res.end(JSON.stringify({ error: 'payload_too_large', message: 'Request body too large (max 10MB)' }));
+                return;
+            }
+            const result = (0, estimate_js_1.handleEstimateRequest)(body);
+            res.writeHead(result.status, { 'Content-Type': 'application/json' });
+            res.end(JSON.stringify(result.body));
+            return;
+        }
         // === Token counting endpoint ===
         if (req.method === 'POST' && url.includes('/v1/messages/count_tokens')) {
             log('Token count request');
@@ -3942,7 +3985,7 @@ async function startProxy(config = {}) {
         // === OpenAI-compatible /v1/chat/completions endpoint ===
         if (req.method !== 'POST' || !url.includes('/chat/completions')) {
             res.writeHead(404, { 'Content-Type': 'application/json' });
-            res.end(JSON.stringify({ error: 'Not found. Supported: POST /v1/messages, POST /v1/chat/completions, GET /v1/models' }));
+            res.end(JSON.stringify({ error: 'Not found. Supported: POST /v1/messages, POST /v1/chat/completions, POST /v1/estimate, GET /v1/models' }));
             return;
         }
         // Parse request body