fix: support NewAPI config test token limit

Author: ZengLiangYi

server/src/routes/config.test.ts

@@ -0,0 +1,102 @@
+import assert from 'node:assert/strict';
+import type { AddressInfo } from 'node:net';
+import { mkdtempSync } from 'node:fs';
+import { tmpdir } from 'node:os';
+import { join } from 'node:path';
+import test from 'node:test';
+import Fastify from 'fastify';
+
+process.env.DATA_DIR = mkdtempSync(join(tmpdir(), 'chatcrystal-config-route-test-'));
+
+const { updateConfig } = await import('../config.js');
+const { configRoutes } = await import('./config.js');
+
+test('config test uses an OpenAI-compatible max_output_tokens value accepted by NewAPI', async () => {
+  const upstream = Fastify();
+  let capturedMaxOutputTokens: unknown;
+
+  upstream.post('/v1/responses', async (req, reply) => {
+    const body = req.body as { max_output_tokens?: number; model?: string };
+    capturedMaxOutputTokens = body.max_output_tokens;
+
+    if (
+      typeof body.max_output_tokens !== 'number' ||
+      body.max_output_tokens < 16
+    ) {
+      reply.status(400);
+      return {
+        error: {
+          message: `Invalid 'max_output_tokens': integer below minimum value. Expected a value >= 16, but got ${body.max_output_tokens} instead.`,
+          type: 'invalid_request_error',
+          param: 'max_output_tokens',
+          code: 'integer_below_minimum',
+        },
+      };
+    }
+
+    return {
+      id: 'resp_test',
+      created_at: 1_700_000_000,
+      model: body.model ?? 'gpt-5.4',
+      output: [
+        {
+          type: 'message',
+          id: 'msg_test',
+          role: 'assistant',
+          content: [{ type: 'output_text', text: 'OK', annotations: [] }],
+        },
+      ],
+      usage: { input_tokens: 1, output_tokens: 1 },
+    };
+  });
+
+  upstream.post('/v1/embeddings', async () => ({
+    data: [{ embedding: [0.1, 0.2, 0.3] }],
+    usage: { prompt_tokens: 1 },
+  }));
+
+  await upstream.listen({ host: '127.0.0.1', port: 0 });
+  const address = upstream.server.address() as AddressInfo;
+  const baseURL = `http://127.0.0.1:${address.port}/v1`;
+
+  updateConfig({
+    llm: {
+      provider: 'custom',
+      baseURL,
+      apiKey: 'test-api-key',
+      model: 'gpt-5.4',
+    },
+    embedding: {
+      provider: 'custom',
+      baseURL,
+      apiKey: 'test-api-key',
+      model: 'text-embedding-ada-002',
+    },
+  });
+
+  const app = Fastify();
+  await app.register(configRoutes);
+
+  try {
+    const response = await app.inject({
+      method: 'POST',
+      url: '/api/config/test',
+    });
+    const body = response.json();
+
+    assert.equal(response.statusCode, 200);
+    assert.equal(body.data.llm.connected, true, body.data.llm.error);
+    assert.equal(body.data.llm.response, 'OK');
+    assert.equal(
+      body.data.embedding.connected,
+      true,
+      body.data.embedding.error,
+    );
+    assert.equal(body.data.connected, true);
+    assert.equal(typeof capturedMaxOutputTokens, 'number');
+    assert.ok(capturedMaxOutputTokens >= 16);
+  } finally {
+    await app.close();
+    await upstream.close();
+  }
+});

server/src/routes/config.ts

@@ -8,6 +8,9 @@ import { clearEmbeddingIndex } from "../services/embedding.js";
 import { getLanguageModel } from "../services/llm.js";
 import { getProvider, listProviders } from "../services/providers.js";
 
+// OpenAI Responses-compatible gateways such as NewAPI reject lower values.
+const CONFIG_TEST_MAX_OUTPUT_TOKENS = 16;
+
 export async function configRoutes(app: FastifyInstance) {
 	// List available providers
 	app.get("/api/providers", async () => {
@@ -149,7 +152,7 @@ export async function configRoutes(app: FastifyInstance) {
 			const llmResult = await generateText({
 				model,
 				prompt: "Reply with exactly: OK",
-				maxOutputTokens: 5,
+				maxOutputTokens: CONFIG_TEST_MAX_OUTPUT_TOKENS,
 			});
 			result.llm = { connected: true, response: llmResult.text.trim() };
 		} catch (err) {