openzoo 0.50.7 → 0.50.9

package/lib/models.js

@@ -366,9 +366,41 @@ export function augmentModelList(payload, { aliases: withAliases = true } = {})
     }
   }
 
+  // FAMILY SHORTCUTS — `/model grok`, `/model deepseek`, `/model gemini`.
+  //
+  // PROVEN NECESSARY on a live VM: the proxy's fuzzy /v1/models/<id> probe
+  // answered `grok -> x-ai/grok-4.6` with a 200, and Claude Code v2.1.246
+  // REFUSED THE ID ANYWAY — this client validates /model purely against the
+  // list it fetched at startup and never probes per-id. So the only fuzzy a
+  // list-validating harness can have is fuzzy that is already IN the list.
+  //
+  // One row per vendor family, resolved through the same scorer as the POST
+  // path, so `grok` lands on the family flagship rather than a hardcoded pick.
+  const FAMILY_TOKENS = ['gpt', 'claude', 'gemini', 'grok', 'deepseek', 'qwen',
+    'mistral', 'llama', 'glm', 'kimi', 'minimax', 'command', 'nova', 'sonar'];
+  const family = [];
+  if (withAliases) {
+    const ids = data.map((m) => m.id);
+    const taken2 = new Set([...have, ...aliases.map((a) => a.id), ...bare.map((b) => b.id)]);
+    for (const tok of FAMILY_TOKENS) {
+      if (taken2.has(tok)) continue;
+      const target = data.find((m) => m.id === resolveModel(tok, ids));
+      if (!target) continue;
+      taken2.add(tok);
+      family.push({
+        id: tok,
+        object: 'model',
+        owned_by: 'openzoo-alias',
+        ...(target.context_length ? { context_length: target.context_length, context_window: target.context_window ?? target.context_length } : {}),
+        ...(target.pricing ? { pricing: target.pricing } : {}),
+        served_by: target.id,
+      });
+    }
+  }
+
   // Auto is no longer published: routing lives on the backend, and a shim
   // that advertises openzoo/auto would have to route it.
-  return { ...payload, object: payload?.object || 'list', data: [...data, ...aliases, ...bare] };
+  return { ...payload, object: payload?.object || 'list', data: [...data, ...aliases, ...bare, ...family] };
 }
 
 /**

package/lib/proxy.js

@@ -11,7 +11,23 @@ import { PayClient, QuoteTooHighError, UnderfundedError } from './pay.js';
 import { tokenBalance } from './x402.js';
 import { evmTokenBalance } from './evm.js';
 import { autoContext } from './autobind.js';
-import { modelsListForRequest, isHarnessAliasId } from './models.js';
+import { modelsListForRequest, isHarnessAliasId, resolveModel, quoteableRows } from './models.js';
+
+/**
+ * Quoteable catalog ids, cached 5 minutes, for the fuzzy /v1/models/<id> probe.
+ * Free read, but harnesses probe on every /model keystroke confirmation —
+ * without a cache each probe is an upstream round trip.
+ */
+let catalogIdsMemo = { at: 0, ids: [] };
+async function catalogIdsCached(modelsUrl, headers) {
+  if (Date.now() - catalogIdsMemo.at < 5 * 60_000 && catalogIdsMemo.ids.length) return catalogIdsMemo.ids;
+  const r = await fetch(modelsUrl, { headers, signal: AbortSignal.timeout(8000) });
+  if (!r.ok) throw new Error(`catalog ${r.status}`);
+  const payload = await r.json();
+  const ids = quoteableRows(payload?.data).map((m) => m.id);
+  if (ids.length) catalogIdsMemo = { at: Date.now(), ids };
+  return ids;
+}
 import { withNamespace } from './namespace.js';
 import { loadSessionSpend, saveSessionSpend } from './session.js';
 import { creditBalance, quotedPrices } from './info.js';
@@ -618,10 +634,35 @@ export async function startProxy({ silent = false, requireToken = null, sessionM
       return;
     }
     const probe = req.method === 'GET' && /^\/v1\/models\/(.+)$/.exec(path);
-    if (probe && isHarnessAliasId(decodeURIComponent(probe[1]))) {
-      res.writeHead(200, { 'content-type': 'application/json' });
-      res.end(JSON.stringify({ id: decodeURIComponent(probe[1]), object: 'model', owned_by: 'openzoo-alias' }));
-      return;
+    if (probe) {
+      const wanted = decodeURIComponent(probe[1]);
+      if (isHarnessAliasId(wanted)) {
+        res.writeHead(200, { 'content-type': 'application/json' });
+        res.end(JSON.stringify({ id: wanted, object: 'model', owned_by: 'openzoo-alias' }));
+        return;
+      }
+      // FUZZY, LIKE THE REST OF THE PIPELINE. The gateway already
+      // similarity-rewrites whatever model id lands in a POST body — but a
+      // harness never gets that far, because it validates the id CLIENT-SIDE
+      // with this exact probe first. Answering 200 only for a hand-typed
+      // alias list meant `/model deepseek pro` and `/model
+      // deepseek-ai/deepseek-v4-pro` were both refused at the door while the
+      // backend would have served either happily. The door must be exactly as
+      // forgiving as the room behind it.
+      try {
+        const ids = await catalogIdsCached(url.replace(/\/v1\/models\/.*$/, '/v1/models'), init.headers);
+        if (ids.includes(wanted)) {
+          res.writeHead(200, { 'content-type': 'application/json' });
+          res.end(JSON.stringify({ id: wanted, object: 'model' }));
+          return;
+        }
+        const resolved = resolveModel(wanted, ids);
+        if (resolved) {
+          res.writeHead(200, { 'content-type': 'application/json' });
+          res.end(JSON.stringify({ id: wanted, object: 'model', owned_by: 'openzoo-alias', served_by: resolved }));
+          return;
+        }
+      } catch { /* catalog unreachable — fall through to the paid path below */ }
     }
 
     try {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "openzoo",
-  "version": "0.50.7",
+  "version": "0.50.9",
   "description": "Local x402-paying proxy + MCP server for openzoo.fun — point any OpenAI-compatible harness (Cursor, Claude Code, aider, SDKs) at localhost and it pays per call from a local burner wallet. Solana and Base rails live; Robinhood experimental.",
   "license": "MIT",
   "type": "module",