open-agents-ai 0.184.54 → 0.184.56

package/dist/index.js

@@ -62035,6 +62035,110 @@ function getVersion3() {
   }
   return "0.0.0";
 }
+async function refreshEndpointRegistry() {
+  endpointRegistry.length = 0;
+  modelRouteMap.clear();
+  const config = loadConfig();
+  const primaryLabel = (() => {
+    if (config.backendUrl.includes("127.0.0.1") || config.backendUrl.includes("localhost"))
+      return "local";
+    try {
+      return new URL(config.backendUrl).hostname.split(".")[0];
+    } catch {
+      return "primary";
+    }
+  })();
+  endpointRegistry.push({
+    label: primaryLabel,
+    url: config.backendUrl,
+    type: config.backendType || "ollama",
+    authKey: config.apiKey
+  });
+  try {
+    const resp = await fetch("https://openagents.nexus/api/v1/sponsors", {
+      signal: AbortSignal.timeout(5e3)
+    });
+    if (resp.ok) {
+      const data = await resp.json();
+      for (const sp of (data.sponsors || []).filter((s) => s.status === "active" && s.tunnelUrl)) {
+        if (sp.tunnelUrl === config.backendUrl)
+          continue;
+        const label = sp.name.replace(/[^a-zA-Z0-9]/g, "").slice(0, 12).toLowerCase() || "sponsor";
+        endpointRegistry.push({
+          label,
+          url: sp.tunnelUrl,
+          type: "vllm",
+          authKey: sp.authKey
+        });
+      }
+    }
+  } catch {
+  }
+}
+async function fetchAggregatedModels() {
+  const allModels = [];
+  modelRouteMap.clear();
+  const fetches = endpointRegistry.map(async (ep) => {
+    try {
+      const isOllama = ep.type === "ollama";
+      const modelsPath = isOllama ? "/api/tags" : "/v1/models";
+      const result = await ollamaRequest(ep.url, modelsPath, "GET");
+      if (result.status !== 200)
+        return;
+      const body = JSON.parse(result.body);
+      if (isOllama) {
+        const models = body.models ?? [];
+        for (const m of models) {
+          const prefixed = `${ep.label}/${m.name}`;
+          modelRouteMap.set(prefixed, { endpoint: ep, originalId: m.name });
+          allModels.push({
+            id: prefixed,
+            object: "model",
+            created: m.modified_at ? Math.floor(new Date(m.modified_at).getTime() / 1e3) : 0,
+            owned_by: ep.label
+          });
+        }
+      } else {
+        const models = body.data ?? [];
+        for (const m of models) {
+          const originalId = m.id || m.name;
+          const prefixed = `${ep.label}/${originalId}`;
+          modelRouteMap.set(prefixed, { endpoint: ep, originalId });
+          allModels.push({
+            id: prefixed,
+            object: "model",
+            created: m.created || 0,
+            owned_by: ep.label
+          });
+        }
+      }
+    } catch {
+    }
+  });
+  await Promise.allSettled(fetches);
+  return allModels.sort((a, b) => a.id.localeCompare(b.id));
+}
+function resolveModelEndpoint(modelId) {
+  const exact = modelRouteMap.get(modelId);
+  if (exact)
+    return exact;
+  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 };
+  }
+  return null;
+}
 function recordMetric(method, path, status) {
   const key = `${method}|${path}|${status}`;
   const existing = metrics.requests.get(key);
@@ -62104,10 +62208,10 @@ async function parseJsonBody(req) {
     return null;
   }
 }
-function backendAuthHeaders() {
-  const config = loadConfig();
-  if (config.apiKey)
-    return { Authorization: `Bearer ${config.apiKey}` };
+function backendAuthHeaders(endpoint) {
+  const key = endpoint?.authKey ?? loadConfig().apiKey;
+  if (key)
+    return { Authorization: `Bearer ${key}` };
   return {};
 }
 function ollamaRequest(ollamaUrl, path, method, body) {
@@ -62285,33 +62389,15 @@ function handleMetrics(res) {
 }
 async function handleV1Models(res, ollamaUrl) {
   try {
-    const config = loadConfig();
-    const isVllm = config.backendType === "vllm";
-    if (isVllm) {
-      const result2 = await ollamaRequest(ollamaUrl, "/v1/models", "GET");
-      if (result2.status !== 200) {
-        jsonResponse(res, result2.status, { error: "Failed to fetch models from backend" });
-        return;
-      }
-      jsonResponse(res, 200, JSON.parse(result2.body));
-      return;
-    }
-    const result = await ollamaRequest(ollamaUrl, "/api/tags", "GET");
-    if (result.status !== 200) {
-      jsonResponse(res, result.status, { error: "Failed to fetch models from backend" });
-      return;
-    }
-    const ollamaBody = JSON.parse(result.body);
-    const models = (ollamaBody.models ?? []).map((m) => ({
-      id: m.name,
-      object: "model",
-      created: m.modified_at ? Math.floor(new Date(m.modified_at).getTime() / 1e3) : 0,
-      owned_by: "local"
-    }));
-    jsonResponse(res, 200, { object: "list", data: models });
+    const models = await fetchAggregatedModels();
+    jsonResponse(res, 200, {
+      object: "list",
+      data: models,
+      _endpoints: endpointRegistry.map((e) => ({ label: e.label, url: e.url, type: e.type }))
+    });
   } catch (err) {
     jsonResponse(res, 502, {
-      error: "Failed to proxy to backend",
+      error: "Failed to fetch models",
       message: err instanceof Error ? err.message : String(err)
     });
   }
@@ -62325,10 +62411,13 @@ async function handleV1ChatCompletions(req, res, ollamaUrl) {
   const requestBody = body;
   const stream = requestBody["stream"] === true;
   const model = requestBody["model"] || "unknown";
-  const config = loadConfig();
-  const isVllm = config.backendType === "vllm";
-  if (isVllm) {
-    const payload = JSON.stringify(requestBody);
+  const route = resolveModelEndpoint(model);
+  const targetUrl = route?.endpoint.url ?? ollamaUrl;
+  const targetType = route?.endpoint.type ?? loadConfig().backendType ?? "ollama";
+  const originalModel = route?.originalId ?? model;
+  const routedBody = { ...requestBody, model: originalModel };
+  if (targetType === "vllm" || targetType === "openai") {
+    const payload = JSON.stringify(routedBody);
     if (stream) {
       res.writeHead(200, {
         "Content-Type": "text/event-stream",
@@ -62338,7 +62427,7 @@ async function handleV1ChatCompletions(req, res, ollamaUrl) {
         "Access-Control-Allow-Methods": "GET, POST, PUT, PATCH, DELETE, OPTIONS",
         "Access-Control-Allow-Headers": "Content-Type, Authorization"
       });
-      ollamaStream(ollamaUrl, "/v1/chat/completions", "POST", payload, (chunk) => {
+      ollamaStream(targetUrl, "/v1/chat/completions", "POST", payload, (chunk) => {
         res.write(chunk);
       }, () => {
         res.end();
@@ -62353,7 +62442,7 @@ async function handleV1ChatCompletions(req, res, ollamaUrl) {
       });
     } else {
       try {
-        const result = await ollamaRequest(ollamaUrl, "/v1/chat/completions", "POST", payload);
+        const result = await ollamaRequest(targetUrl, "/v1/chat/completions", "POST", payload);
         if (result.status !== 200) {
           jsonResponse(res, result.status, { error: "Backend request failed", details: result.body });
           return;
@@ -62371,7 +62460,7 @@ async function handleV1ChatCompletions(req, res, ollamaUrl) {
     return;
   }
   const ollamaPayload = JSON.stringify({
-    ...requestBody,
+    ...routedBody,
     stream,
     think: false
   });
@@ -62386,7 +62475,7 @@ async function handleV1ChatCompletions(req, res, ollamaUrl) {
     });
     const chatId = `chatcmpl-${randomBytes16(12).toString("hex")}`;
     let buffer = "";
-    ollamaStream(ollamaUrl, "/api/chat", "POST", ollamaPayload, (chunk) => {
+    ollamaStream(targetUrl, "/api/chat", "POST", ollamaPayload, (chunk) => {
       buffer += chunk;
       const lines = buffer.split("\n");
       buffer = lines.pop() ?? "";
@@ -62476,7 +62565,7 @@ async function handleV1ChatCompletions(req, res, ollamaUrl) {
     });
   } else {
     try {
-      const result = await ollamaRequest(ollamaUrl, "/api/chat", "POST", ollamaPayload);
+      const result = await ollamaRequest(targetUrl, "/api/chat", "POST", ollamaPayload);
       if (result.status !== 200) {
         jsonResponse(res, result.status, {
           error: "Ollama request failed",
@@ -63149,29 +63238,35 @@ function startApiServer(options = {}) {
 `);
     }
   });
-  server.listen(port, host, () => {
-    const version = getVersion3();
-    process.stderr.write(`
+  refreshEndpointRegistry().catch(() => {
+  }).then(() => {
+    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}:${port}
 `);
-    const beType = config.backendType || "ollama";
-    process.stderr.write(`  Backend (${beType}): ${ollamaUrl}
+      process.stderr.write(`  Endpoints: ${endpointRegistry.length} registered
 `);
-    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)
+      for (const ep of endpointRegistry) {
+        process.stderr.write(`    ${ep.label}: ${ep.url} (${ep.type})${ep.authKey ? " [auth]" : ""}
 `);
-    } else if (process.env["OA_API_KEY"]) {
-      process.stderr.write(`  Auth: single Bearer token (admin scope)
+      }
+      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 {
-      process.stderr.write(`  Auth: disabled (set OA_API_KEY or OA_API_KEYS to enable)
+      } else if (process.env["OA_API_KEY"]) {
+        process.stderr.write(`  Auth: single Bearer token (admin scope)
 `);
-    }
-    process.stderr.write(`
+      } else {
+        process.stderr.write(`  Auth: disabled (set OA_API_KEY or OA_API_KEYS to enable)
+`);
+      }
+      process.stderr.write(`
 `);
+    });
   });
   const shutdown = () => {
     process.stderr.write("\n  Shutting down API server...\n");
@@ -63201,7 +63296,7 @@ async function apiServeCommand(opts, config) {
     server.on("close", resolve36);
   });
 }
-var metrics, startedAt, runningProcesses;
+var endpointRegistry, modelRouteMap, metrics, startedAt, runningProcesses;
 var init_serve = __esm({
   "packages/cli/dist/api/serve.js"() {
     "use strict";
@@ -63209,6 +63304,8 @@ var init_serve = __esm({
     init_oa_directory();
     init_render();
     init_profiles();
+    endpointRegistry = [];
+    modelRouteMap = /* @__PURE__ */ new Map();
     metrics = {
       requests: /* @__PURE__ */ new Map(),
       totalTokensIn: 0,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "open-agents-ai",
-  "version": "0.184.54",
+  "version": "0.184.56",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",