open-agents-ai 0.184.55 → 0.184.57

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
         });
       }
@@ -62122,10 +62158,18 @@ function resolveModelEndpoint(modelId) {
   const exact = modelRouteMap.get(modelId);
   if (exact)
     return exact;
-  for (const [prefixed, route] of modelRouteMap) {
+  for (const [, route] of modelRouteMap) {
     if (route.originalId === modelId)
       return route;
   }
+  const slashIdx = modelId.indexOf("/");
+  if (slashIdx > 0) {
+    const prefix = modelId.slice(0, slashIdx);
+    const rest = modelId.slice(slashIdx + 1);
+    const ep = endpointRegistry.find((e) => e.label === prefix);
+    if (ep)
+      return { endpoint: ep, originalId: rest };
+  }
   if (endpointRegistry.length > 0) {
     return { endpoint: endpointRegistry[0], originalId: modelId };
   }
@@ -62407,6 +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);
@@ -62441,8 +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) {
@@ -63096,6 +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;
@@ -63288,7 +63355,7 @@ 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";
@@ -63298,6 +63365,7 @@ 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": "0.184.55",
+  "version": "0.184.57",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",