@relayplane/proxy 1.8.10 → 1.8.12

package/dist/standalone-proxy.js

@@ -85,6 +85,12 @@ 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');
@@ -987,12 +993,29 @@ function getAuthForModel(model, authConfig, envApiKey) {
     }
     return { apiKey: envApiKey, isMax: false };
 }
+/**
+ * Set Anthropic auth header for a token.
+ * OAT tokens (sk-ant-oat*) require Authorization: Bearer + oauth beta header.
+ * Standard API keys use x-api-key.
+ */
+function setAnthropicAuth(headers, token) {
+    if (token.startsWith('sk-ant-oat')) {
+        headers['Authorization'] = `Bearer ${token}`;
+        const existing = headers['anthropic-beta'];
+        const oauthBeta = 'oauth-2025-04-20';
+        if (!existing) {
+            headers['anthropic-beta'] = oauthBeta;
+        }
+        else if (!existing.includes(oauthBeta)) {
+            headers['anthropic-beta'] = `${existing},${oauthBeta}`;
+        }
+    }
+    else {
+        headers['x-api-key'] = token;
+    }
+}
 /**
  * Build Anthropic headers with hybrid auth support
- *
- * All Anthropic token types (sk-ant-api*, sk-ant-oat*) work via x-api-key header.
- * The Bearer header approach was incorrectly assumed for oat tokens but Anthropic
- * rejects it with "OAuth authentication is currently not supported."
  */
 function buildAnthropicHeadersWithAuth(ctx, apiKey, isMaxToken, isRerouted) {
     const headers = {
@@ -1000,23 +1023,25 @@ function buildAnthropicHeadersWithAuth(ctx, apiKey, isMaxToken, isRerouted) {
         'anthropic-version': ctx.versionHeader || '2023-06-01',
     };
     // Auth priority: incoming auth (passthrough) > configured API key > env key
-    // All Anthropic tokens use x-api-key header (including sk-ant-oat* MAX tokens)
     if (ctx.authHeader) {
-        // Extract token from "Bearer <token>" and send as x-api-key
         const token = ctx.authHeader.replace(/^Bearer\s+/i, '');
-        headers['x-api-key'] = token;
+        setAnthropicAuth(headers, token);
     }
     else if (ctx.apiKeyHeader) {
-        // Incoming x-api-key header — pass through as-is
-        headers['x-api-key'] = ctx.apiKeyHeader;
+        setAnthropicAuth(headers, ctx.apiKeyHeader);
     }
     else if (apiKey) {
-        // Fallback to configured API key
-        headers['x-api-key'] = apiKey;
+        setAnthropicAuth(headers, apiKey);
     }
     // Pass through beta headers
     if (ctx.betaHeaders) {
-        headers['anthropic-beta'] = ctx.betaHeaders;
+        const existing = headers['anthropic-beta'];
+        if (!existing) {
+            headers['anthropic-beta'] = ctx.betaHeaders;
+        }
+        else if (!existing.includes(ctx.betaHeaders)) {
+            headers['anthropic-beta'] = `${existing},${ctx.betaHeaders}`;
+        }
     }
     // Pass through OAuth identity headers (required by Anthropic for OAuth token validation)
     if (ctx.userAgent) {
@@ -1029,35 +1054,32 @@ function buildAnthropicHeadersWithAuth(ctx, apiKey, isMaxToken, isRerouted) {
 }
 /**
  * Build Anthropic headers with auth passthrough support
- *
- * Auth priority:
- * 1. Incoming Authorization header (Bearer token from Claude Code OAuth)
- * 2. Incoming x-api-key header
- * 3. ANTHROPIC_API_KEY env var (or MAX token for Opus models)
  */
 function buildAnthropicHeaders(ctx, envApiKey) {
     const headers = {
         'Content-Type': 'application/json',
         'anthropic-version': ctx.versionHeader || '2023-06-01',
     };
-    // Auth: prefer incoming auth for passthrough, fallback to env
-    // All Anthropic tokens use x-api-key (including sk-ant-oat* MAX tokens)
+    // Auth priority: incoming auth > x-api-key header > env key
     if (ctx.authHeader) {
-        // Extract token from "Bearer <token>" and send as x-api-key
         const token = ctx.authHeader.replace(/^Bearer\s+/i, '');
-        headers['x-api-key'] = token;
+        setAnthropicAuth(headers, token);
     }
     else if (ctx.apiKeyHeader) {
-        // Direct x-api-key header
-        headers['x-api-key'] = ctx.apiKeyHeader;
+        setAnthropicAuth(headers, ctx.apiKeyHeader);
     }
     else if (envApiKey) {
-        // Fallback to env var
-        headers['x-api-key'] = envApiKey;
+        setAnthropicAuth(headers, envApiKey);
     }
-    // Pass through beta headers (prompt caching, extended thinking, etc.)
+    // Pass through beta headers
     if (ctx.betaHeaders) {
-        headers['anthropic-beta'] = ctx.betaHeaders;
+        const existing = headers['anthropic-beta'];
+        if (!existing) {
+            headers['anthropic-beta'] = ctx.betaHeaders;
+        }
+        else if (!existing.includes(ctx.betaHeaders)) {
+            headers['anthropic-beta'] = `${existing},${ctx.betaHeaders}`;
+        }
     }
     // Pass through OAuth identity headers (required by Anthropic for OAuth token validation)
     if (ctx.userAgent) {
@@ -3899,6 +3921,39 @@ 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');
@@ -3946,7 +4001,7 @@ 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