llm-switcher 1.2.10 → 1.2.11

package/CHANGELOG.md

@@ -1,5 +1,9 @@
 # Changelog — LLM Switcher
 
+## Release 1.2.11
+
+- **Test connection on a Bifrost model:** The test sent a plain `ping` request, and Anthropic answers that with a fake 429 for a Claude Code account. Now, when intact names `bifrost_ua` for the model, the test reports that the key works and the model is served, and it sends no message.
+
 ## Release 1.2.10
 
 - **Bifrost in the log:** The log line and the request inspector showed `anthropic -> openai-chat` for a Bifrost request, because the gateway logged the profile's `outFormat` before it checked Bifrost. Now both show `bifrost`.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "llm-switcher",
-  "version": "1.2.10",
+  "version": "1.2.11",
   "description": "Zero-dependency multi-protocol edge gateway & provider switcher for Claude Code and Codex, with OpenAI-compatible, Anthropic and Vertex upstreams",
   "keywords": [
     "llm",

package/proxy.mjs

@@ -1123,6 +1123,14 @@ async function testUpstream(body, cfg) {
   const apiKey = resolveApiKey(cfg, body.key, body.apiKey, baseURL);
   const model = body.model || 'default';
   const profile = { baseURL, apiKey, mode: body.mode, outFormat: body.outFormat || undefined };
+  // A Bifrost model answers only its own client: a plain test request would draw a fake 429.
+  // The model lookup itself proves the key and the model, so it is the test.
+  const lookupStart = Date.now();
+  const bifrostUA = await bifrostUAFor(profile, model);
+  if (bifrostUA) {
+    return { status: 200, json: { ok: true, latency: Date.now() - lookupStart, outFormat: 'bifrost',
+      sample: `Bifrost: the key works and ${model} is served. Only the real client (${bifrostUA}) can call it, so no test message was sent.` } };
+  }
   const outFormat = resolveOutFormat(profile, model);
   const { url, headers } = upstreamEndpoint(profile, outFormat, model, false, null);
   const ir = { model, system: '', messages: [{ role: 'user', content: 'ping' }], tools: [], toolChoice: null, params: { maxTokens: 16, temperature: null, topP: null, topK: null, stop: [] }, thinking: { type: 'disabled' }, stream: false };

package/tests/gateway.e2e.test.mjs

@@ -1251,6 +1251,17 @@ test('Codex WS: conversation.item.create is echoed and response.cancel stops the
   ws.socket.destroy();
 });
 
+test('Test connection on a Bifrost model checks the model and the key, and sends no fake request', async () => {
+  const before = received.length;
+  const r = await post('/api/test-upstream', { baseURL: `http://127.0.0.1:${upstreamPort}/intact/v1`, apiKey: 'k', model: 'claude/claude-opus-5', mode: 'convert' });
+  const j = await r.json();
+  assert.equal(j.ok, true);
+  assert.equal(j.outFormat, 'bifrost');
+  assert.match(j.sample, /claude-cli\//);
+  const sent = received.slice(before).map(x => x.url);
+  assert.ok(!sent.some(u => u.includes('/messages') || u.includes('/chat/completions')), `sent ${sent}`);
+});
+
 test('Admin test-upstream reports latency and a sample from the upstream', async () => {
   const r = await post('/api/test-upstream', { baseURL: `http://127.0.0.1:${upstreamPort}/chat/v1`, apiKey: 'k', model: 'm', mode: 'convert' });
   assert.equal(r.status, 200);