open-agents-ai 0.184.56 → 0.184.58

package/dist/index.js

@@ -62035,6 +62035,38 @@ function getVersion3() {
   }
   return "0.0.0";
 }
+function getEndpointUsage(label) {
+  let u = endpointUsage.get(label);
+  const today = (/* @__PURE__ */ new Date()).toISOString().slice(0, 10);
+  if (!u) {
+    u = { requestTimestamps: [], tokensToday: 0, tokenDate: today };
+    endpointUsage.set(label, u);
+  }
+  if (u.tokenDate !== today) {
+    u.tokensToday = 0;
+    u.tokenDate = today;
+  }
+  return u;
+}
+function checkEndpointRateLimit(ep, estimatedTokens) {
+  const rpm = ep.limits?.maxRequestsPerMinute ?? 60;
+  const tpd = ep.limits?.maxTokensPerDay ?? 1e5;
+  const usage = getEndpointUsage(ep.label);
+  const now = Date.now();
+  usage.requestTimestamps = usage.requestTimestamps.filter((t) => now - t < 6e4);
+  if (usage.requestTimestamps.length >= rpm) {
+    return `Rate limit exceeded for ${ep.label}: ${rpm} requests/minute. Try again in ${Math.ceil((usage.requestTimestamps[0] + 6e4 - now) / 1e3)}s.`;
+  }
+  if (estimatedTokens && usage.tokensToday + estimatedTokens > tpd) {
+    return `Daily token limit exceeded for ${ep.label}: ${usage.tokensToday}/${tpd} tokens today.`;
+  }
+  return null;
+}
+function recordEndpointUsage(label, tokens) {
+  const usage = getEndpointUsage(label);
+  usage.requestTimestamps.push(Date.now());
+  usage.tokensToday += tokens;
+}
 async function refreshEndpointRegistry() {
   endpointRegistry.length = 0;
   modelRouteMap.clear();
@@ -62068,6 +62100,10 @@ async function refreshEndpointRegistry() {
           label,
           url: sp.tunnelUrl,
           type: "vllm",
+          limits: {
+            maxRequestsPerMinute: sp.limits?.maxRequestsPerMinute ?? 60,
+            maxTokensPerDay: sp.limits?.maxTokensPerDay ?? 1e5
+          },
           authKey: sp.authKey
         });
       }
@@ -62415,6 +62451,13 @@ async function handleV1ChatCompletions(req, res, ollamaUrl) {
   const targetUrl = route?.endpoint.url ?? ollamaUrl;
   const targetType = route?.endpoint.type ?? loadConfig().backendType ?? "ollama";
   const originalModel = route?.originalId ?? model;
+  if (route?.endpoint) {
+    const limitErr = checkEndpointRateLimit(route.endpoint);
+    if (limitErr) {
+      jsonResponse(res, 429, { error: limitErr });
+      return;
+    }
+  }
   const routedBody = { ...requestBody, model: originalModel };
   if (targetType === "vllm" || targetType === "openai") {
     const payload = JSON.stringify(routedBody);
@@ -62449,8 +62492,11 @@ async function handleV1ChatCompletions(req, res, ollamaUrl) {
         }
         const parsed = JSON.parse(result.body);
         if (parsed.usage) {
+          const totalTok = (parsed.usage.prompt_tokens ?? 0) + (parsed.usage.completion_tokens ?? 0);
           metrics.totalTokensIn += parsed.usage.prompt_tokens ?? 0;
           metrics.totalTokensOut += parsed.usage.completion_tokens ?? 0;
+          if (route?.endpoint)
+            recordEndpointUsage(route.endpoint.label, totalTok);
         }
         jsonResponse(res, 200, parsed);
       } catch (err) {
@@ -63104,6 +63150,19 @@ async function handleRequest(req, res, ollamaUrl, verbose) {
       await handlePutConfigEndpoint(req, res);
       return;
     }
+    if (pathname === "/v1/usage" && method === "GET") {
+      const usageData = {};
+      for (const ep of endpointRegistry) {
+        const u = getEndpointUsage(ep.label);
+        usageData[ep.label] = {
+          requestsLastMinute: u.requestTimestamps.filter((t) => Date.now() - t < 6e4).length,
+          tokensToday: u.tokensToday,
+          limits: ep.limits ?? { maxRequestsPerMinute: "unlimited", maxTokensPerDay: "unlimited" }
+        };
+      }
+      jsonResponse(res, 200, { usage: usageData, totalTokensIn: metrics.totalTokensIn, totalTokensOut: metrics.totalTokensOut });
+      return;
+    }
     if (pathname === "/v1/commands" && method === "GET") {
       handleGetCommands(res);
       return;
@@ -63229,43 +63288,93 @@ function startApiServer(options = {}) {
       }
     });
   });
+  let retried = false;
   server.on("error", (err) => {
-    if (err.code === "EADDRINUSE") {
-      process.stderr.write(`  Port ${port} already in use \u2014 API server not started.
+    if (err.code === "EADDRINUSE" && !retried) {
+      retried = true;
+      process.stderr.write(`  Port ${port} in use \u2014 reclaiming...
+`);
+      try {
+        const { execSync: es } = __require("node:child_process");
+        const pids = es(`lsof -ti :${port} 2>/dev/null || fuser ${port}/tcp 2>/dev/null || true`, { encoding: "utf8" }).trim().split(/[\n\s]+/).filter(Boolean);
+        for (const pid of pids) {
+          const numPid = parseInt(pid, 10);
+          if (numPid && numPid !== process.pid) {
+            try {
+              process.kill(numPid, "SIGTERM");
+            } catch {
+            }
+            setTimeout(() => {
+              try {
+                process.kill(numPid, "SIGKILL");
+              } catch {
+              }
+            }, 500);
+          }
+        }
+      } catch {
+      }
+      setTimeout(() => {
+        server.listen(port, host);
+      }, 2e3);
+    } else if (err.code === "EADDRINUSE" && retried) {
+      process.stderr.write(`  Port ${port} still in use after reclaim attempt \u2014 API server not started.
 `);
     } else {
       process.stderr.write(`  API server error: ${err.message}
 `);
     }
   });
-  refreshEndpointRegistry().catch(() => {
-  }).then(() => {
-    server.listen(port, host, () => {
-      const version = getVersion3();
-      process.stderr.write(`
+  endpointRegistry.push({
+    label: (() => {
+      const u = config.backendUrl;
+      if (u.includes("127.0.0.1") || u.includes("localhost"))
+        return "local";
+      try {
+        return new URL(u).hostname.split(".")[0];
+      } catch {
+        return "primary";
+      }
+    })(),
+    url: config.backendUrl,
+    type: config.backendType || "ollama",
+    authKey: config.apiKey
+  });
+  server.listen(port, host, () => {
+    const version = getVersion3();
+    process.stderr.write(`
   open-agents API server v${version}
 `);
-      process.stderr.write(`  Listening on http://${host}:${port}
+    process.stderr.write(`  Listening on http://${host}:${port}
 `);
-      process.stderr.write(`  Endpoints: ${endpointRegistry.length} registered
+    process.stderr.write(`  Primary: ${config.backendUrl} (${config.backendType || "ollama"})
 `);
-      for (const ep of endpointRegistry) {
-        process.stderr.write(`    ${ep.label}: ${ep.url} (${ep.type})${ep.authKey ? " [auth]" : ""}
+    if (process.env["OA_API_KEYS"]) {
+      const keyCount = process.env["OA_API_KEYS"].split(",").length;
+      process.stderr.write(`  Auth: ${keyCount} scoped key(s) (read/run/admin)
 `);
-      }
-      if (process.env["OA_API_KEYS"]) {
-        const keyCount = process.env["OA_API_KEYS"].split(",").length;
-        process.stderr.write(`  Auth: ${keyCount} scoped key(s) (read/run/admin)
+    } else if (process.env["OA_API_KEY"]) {
+      process.stderr.write(`  Auth: single Bearer token (admin scope)
 `);
-      } else if (process.env["OA_API_KEY"]) {
-        process.stderr.write(`  Auth: single Bearer token (admin scope)
+    } else {
+      process.stderr.write(`  Auth: disabled (set OA_API_KEY or OA_API_KEYS to enable)
 `);
-      } else {
-        process.stderr.write(`  Auth: disabled (set OA_API_KEY or OA_API_KEYS to enable)
+    }
+    process.stderr.write(`  Discovering sponsors in background...
+
 `);
-      }
-      process.stderr.write(`
+    refreshEndpointRegistry().then(() => {
+      if (endpointRegistry.length > 1) {
+        process.stderr.write(`  Sponsors: ${endpointRegistry.length - 1} endpoint(s) discovered
+`);
+        for (const ep of endpointRegistry.slice(1)) {
+          process.stderr.write(`    ${ep.label}: ${ep.url} (${ep.type})${ep.authKey ? " [auth]" : ""}
+`);
+        }
+        process.stderr.write(`
 `);
+      }
+    }).catch(() => {
     });
   });
   const shutdown = () => {
@@ -63296,7 +63405,7 @@ async function apiServeCommand(opts, config) {
     server.on("close", resolve36);
   });
 }
-var endpointRegistry, modelRouteMap, metrics, startedAt, runningProcesses;
+var endpointRegistry, modelRouteMap, endpointUsage, metrics, startedAt, runningProcesses;
 var init_serve = __esm({
   "packages/cli/dist/api/serve.js"() {
     "use strict";
@@ -63306,6 +63415,7 @@ var init_serve = __esm({
     init_profiles();
     endpointRegistry = [];
     modelRouteMap = /* @__PURE__ */ new Map();
+    endpointUsage = /* @__PURE__ */ new Map();
     metrics = {
       requests: /* @__PURE__ */ new Map(),
       totalTokensIn: 0,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "open-agents-ai",
-  "version": "0.184.56",
+  "version": "0.184.58",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",