onbuzz 4.9.13 → 4.10.0

package/src/services/embeddings/__tests__/azureCustomProvider.test.js

@@ -1,232 +1,232 @@
-/**
- * AzureCustomEmbeddingProvider tests.
- *
- * Pins:
- *   - URL normalization: baseURL with/without trailing slash, already
- *     -suffixed URLs, raw input — all converge on `<base>/embeddings`.
- *   - Required constructor args (endpoint, deployment, getApiKey).
- *   - getInfo() reports azure-custom + correct dims + user's deployment.
- *   - Happy path POSTs `{input, model, encoding_format:'float'}` with
- *     bearer auth.
- *   - Missing key → AUTH_ERROR, no HTTP call.
- *   - Fingerprint encodes model + dims so swapping deployment triggers
- *     re-embed via the existing VectorStore drift logic.
- *   - Custom dimensions honoured.
- *   - Default dimensions kick in when none supplied.
- *   - Vendor key rotation is read via the callback on each call.
- */
-
-import { describe, test, expect, jest, beforeEach } from '@jest/globals';
-import { EmbeddingError, EMBEDDING_ERROR_CODES } from '../providerInterface.js';
-import {
-  AzureCustomEmbeddingProvider,
-  normalizeEmbeddingsUrl,
-} from '../providers/azureCustomProvider.js';
-
-function makeMockFetch({ status = 200, json = null, text = '', throwError = null } = {}) {
-  return jest.fn(async () => {
-    if (throwError) throw throwError;
-    return {
-      ok: status >= 200 && status < 300,
-      status,
-      async json() {
-        if (json === null) throw new Error('no json');
-        return json;
-      },
-      async text() { return text; },
-    };
-  });
-}
-
-describe('normalizeEmbeddingsUrl', () => {
-  test('appends /embeddings to a baseURL with trailing slash', () => {
-    expect(normalizeEmbeddingsUrl('https://x.azure.com/openai/v1/'))
-      .toBe('https://x.azure.com/openai/v1/embeddings');
-  });
-
-  test('appends /embeddings to a baseURL without trailing slash', () => {
-    expect(normalizeEmbeddingsUrl('https://x.azure.com/openai/v1'))
-      .toBe('https://x.azure.com/openai/v1/embeddings');
-  });
-
-  test('leaves URL untouched when it already ends with /embeddings', () => {
-    expect(normalizeEmbeddingsUrl('https://x.azure.com/openai/v1/embeddings'))
-      .toBe('https://x.azure.com/openai/v1/embeddings');
-  });
-
-  test('strips trailing slashes even on already-suffixed URLs', () => {
-    expect(normalizeEmbeddingsUrl('https://x.azure.com/openai/v1/embeddings/'))
-      .toBe('https://x.azure.com/openai/v1/embeddings');
-  });
-
-  test('non-string / empty input returned as-is (defensive)', () => {
-    expect(normalizeEmbeddingsUrl(null)).toBeNull();
-    expect(normalizeEmbeddingsUrl('')).toBe('');
-    expect(normalizeEmbeddingsUrl('  ')).toBe('  ');
-  });
-});
-
-describe('AzureCustomEmbeddingProvider — constructor', () => {
-  test('rejects missing endpoint / deployment / getApiKey', () => {
-    expect(() => new AzureCustomEmbeddingProvider({})).toThrow(/endpoint/);
-    expect(() => new AzureCustomEmbeddingProvider({ endpoint: 'https://x' })).toThrow(/deployment/);
-    expect(() => new AzureCustomEmbeddingProvider({
-      endpoint: 'https://x', deployment: 'm',
-    })).toThrow(/getApiKey/);
-  });
-
-  test('default dimensions = 1536 when none supplied', () => {
-    const p = new AzureCustomEmbeddingProvider({
-      endpoint: 'https://x/v1/',
-      deployment: 'embed-v-4-0',
-      getApiKey: () => 'k',
-    });
-    expect(p.getInfo().dimensions).toBe(1536);
-  });
-
-  test('custom dimensions honoured', () => {
-    const p = new AzureCustomEmbeddingProvider({
-      endpoint: 'https://x/v1/',
-      deployment: 'embed-v-4-0',
-      dimensions: 1024,
-      getApiKey: () => 'k',
-    });
-    expect(p.getInfo().dimensions).toBe(1024);
-  });
-
-  test('invalid dimensions fall back to default', () => {
-    const p = new AzureCustomEmbeddingProvider({
-      endpoint: 'https://x/v1/',
-      deployment: 'embed-v-4-0',
-      dimensions: -5,
-      getApiKey: () => 'k',
-    });
-    expect(p.getInfo().dimensions).toBe(1536);
-  });
-});
-
-describe('AzureCustomEmbeddingProvider — getInfo + health', () => {
-  test('getInfo reports azure-custom + the user\'s deployment as modelId', () => {
-    const p = new AzureCustomEmbeddingProvider({
-      endpoint: 'https://x.azure.com/openai/v1/',
-      deployment: 'embed-v-4-0',
-      getApiKey: () => 'k',
-    });
-    expect(p.getInfo()).toMatchObject({
-      name: 'azure-custom',
-      modelId: 'embed-v-4-0',
-      dimensions: 1536,
-      isLocal: false,
-      distinguishesQueryAndDocument: false,
-    });
-  });
-
-  test('health() reflects missing key with a remediation message', async () => {
-    const getApiKey = jest.fn(() => null);
-    const p = new AzureCustomEmbeddingProvider({
-      endpoint: 'https://x/v1/', deployment: 'm', getApiKey,
-    });
-    const h = await p.health();
-    expect(h.ok).toBe(false);
-    expect(h.reason).toMatch(/API key/);
-  });
-});
-
-describe('AzureCustomEmbeddingProvider — HTTP call shape', () => {
-  let fetchImpl, getApiKey, provider;
-
-  beforeEach(() => {
-    fetchImpl = makeMockFetch({
-      status: 200,
-      json: { data: [{ embedding: Array(1536).fill(0.01) }] },
-    });
-    getApiKey = jest.fn(() => 'sk-azure-test');
-    provider = new AzureCustomEmbeddingProvider({
-      endpoint: 'https://loxia-ai-resource.services.ai.azure.com/openai/v1/',
-      deployment: 'embed-v-4-0',
-      getApiKey,
-      fetchImpl,
-    });
-  });
-
-  test('POSTs to the normalized URL with bearer auth', async () => {
-    await provider.embedDocuments(['hello']);
-    const [url, init] = fetchImpl.mock.calls[0];
-    expect(url).toBe('https://loxia-ai-resource.services.ai.azure.com/openai/v1/embeddings');
-    expect(init.headers.Authorization).toBe('Bearer sk-azure-test');
-  });
-
-  test('sends {input, model, encoding_format:"float"} body', async () => {
-    await provider.embedDocuments(['hello']);
-    const [, init] = fetchImpl.mock.calls[0];
-    expect(JSON.parse(init.body)).toEqual({
-      input: ['hello'],
-      model: 'embed-v-4-0',
-      encoding_format: 'float',
-    });
-  });
-
-  test('missing API key → AUTH_ERROR, no HTTP call', async () => {
-    getApiKey.mockReturnValue(null);
-    await expect(provider.embedDocuments(['x'])).rejects.toMatchObject({
-      code: EMBEDDING_ERROR_CODES.AUTH_ERROR,
-      transient: false,
-      provider: 'azure-custom',
-    });
-    expect(fetchImpl).not.toHaveBeenCalled();
-  });
-
-  test('API key is read fresh per call (rotation works)', async () => {
-    await provider.embedDocuments(['a']);
-    getApiKey.mockReturnValue('sk-azure-rotated');
-    await provider.embedDocuments(['b']);
-    expect(fetchImpl.mock.calls[1][1].headers.Authorization).toBe('Bearer sk-azure-rotated');
-  });
-
-  test('batches respect maxBatchSize', async () => {
-    // The mock returns one embedding per input it sees, so the per-batch
-    // count always matches what the base class expects.
-    const dynamicFetch = jest.fn(async (_url, init) => {
-      const body = JSON.parse(init.body);
-      const count = Array.isArray(body.input) ? body.input.length : 1;
-      return {
-        ok: true,
-        status: 200,
-        async json() {
-          return { data: Array(count).fill({ embedding: Array(1536).fill(0.01) }) };
-        },
-        async text() { return ''; },
-      };
-    });
-    provider = new AzureCustomEmbeddingProvider({
-      endpoint: 'https://x/v1/',
-      deployment: 'embed-v-4-0',
-      getApiKey,
-      fetchImpl: dynamicFetch,
-      maxBatchSize: 2,
-    });
-    await provider.embedDocuments(['a', 'b', 'c']);  // 2-chunk batch then single
-    expect(dynamicFetch).toHaveBeenCalledTimes(2);
-    const batches = dynamicFetch.mock.calls.map(c => JSON.parse(c[1].body).input.length);
-    expect(batches).toEqual([2, 1]);
-  });
-});
-
-describe('AzureCustomEmbeddingProvider — fingerprint', () => {
-  test('fingerprint encodes model + dimensions so swaps trigger re-embed', () => {
-    const a = new AzureCustomEmbeddingProvider({
-      endpoint: 'https://x/v1/', deployment: 'embed-v-4-0', dimensions: 1536, getApiKey: () => 'k',
-    });
-    const b = new AzureCustomEmbeddingProvider({
-      endpoint: 'https://x/v1/', deployment: 'embed-v-4-0', dimensions: 1024, getApiKey: () => 'k',
-    });
-    const c = new AzureCustomEmbeddingProvider({
-      endpoint: 'https://x/v1/', deployment: 'text-embedding-3-small', dimensions: 1536, getApiKey: () => 'k',
-    });
-    expect(a.getModelFingerprint()).toBe('azure-custom:embed-v-4-0@1536d');
-    expect(b.getModelFingerprint()).toBe('azure-custom:embed-v-4-0@1024d');
-    expect(c.getModelFingerprint()).toBe('azure-custom:text-embedding-3-small@1536d');
-    expect(a.getModelFingerprint()).not.toBe(b.getModelFingerprint());
-    expect(a.getModelFingerprint()).not.toBe(c.getModelFingerprint());
-  });
-});
+/**
+ * AzureCustomEmbeddingProvider tests.
+ *
+ * Pins:
+ *   - URL normalization: baseURL with/without trailing slash, already
+ *     -suffixed URLs, raw input — all converge on `<base>/embeddings`.
+ *   - Required constructor args (endpoint, deployment, getApiKey).
+ *   - getInfo() reports azure-custom + correct dims + user's deployment.
+ *   - Happy path POSTs `{input, model, encoding_format:'float'}` with
+ *     bearer auth.
+ *   - Missing key → AUTH_ERROR, no HTTP call.
+ *   - Fingerprint encodes model + dims so swapping deployment triggers
+ *     re-embed via the existing VectorStore drift logic.
+ *   - Custom dimensions honoured.
+ *   - Default dimensions kick in when none supplied.
+ *   - Vendor key rotation is read via the callback on each call.
+ */
+
+import { describe, test, expect, jest, beforeEach } from '@jest/globals';
+import { EMBEDDING_ERROR_CODES } from '../providerInterface.js';
+import {
+  AzureCustomEmbeddingProvider,
+  normalizeEmbeddingsUrl,
+} from '../providers/azureCustomProvider.js';
+
+function makeMockFetch({ status = 200, json = null, text = '', throwError = null } = {}) {
+  return jest.fn(async () => {
+    if (throwError) throw throwError;
+    return {
+      ok: status >= 200 && status < 300,
+      status,
+      async json() {
+        if (json === null) throw new Error('no json');
+        return json;
+      },
+      async text() { return text; },
+    };
+  });
+}
+
+describe('normalizeEmbeddingsUrl', () => {
+  test('appends /embeddings to a baseURL with trailing slash', () => {
+    expect(normalizeEmbeddingsUrl('https://x.azure.com/openai/v1/'))
+      .toBe('https://x.azure.com/openai/v1/embeddings');
+  });
+
+  test('appends /embeddings to a baseURL without trailing slash', () => {
+    expect(normalizeEmbeddingsUrl('https://x.azure.com/openai/v1'))
+      .toBe('https://x.azure.com/openai/v1/embeddings');
+  });
+
+  test('leaves URL untouched when it already ends with /embeddings', () => {
+    expect(normalizeEmbeddingsUrl('https://x.azure.com/openai/v1/embeddings'))
+      .toBe('https://x.azure.com/openai/v1/embeddings');
+  });
+
+  test('strips trailing slashes even on already-suffixed URLs', () => {
+    expect(normalizeEmbeddingsUrl('https://x.azure.com/openai/v1/embeddings/'))
+      .toBe('https://x.azure.com/openai/v1/embeddings');
+  });
+
+  test('non-string / empty input returned as-is (defensive)', () => {
+    expect(normalizeEmbeddingsUrl(null)).toBeNull();
+    expect(normalizeEmbeddingsUrl('')).toBe('');
+    expect(normalizeEmbeddingsUrl('  ')).toBe('  ');
+  });
+});
+
+describe('AzureCustomEmbeddingProvider — constructor', () => {
+  test('rejects missing endpoint / deployment / getApiKey', () => {
+    expect(() => new AzureCustomEmbeddingProvider({})).toThrow(/endpoint/);
+    expect(() => new AzureCustomEmbeddingProvider({ endpoint: 'https://x' })).toThrow(/deployment/);
+    expect(() => new AzureCustomEmbeddingProvider({
+      endpoint: 'https://x', deployment: 'm',
+    })).toThrow(/getApiKey/);
+  });
+
+  test('default dimensions = 1536 when none supplied', () => {
+    const p = new AzureCustomEmbeddingProvider({
+      endpoint: 'https://x/v1/',
+      deployment: 'embed-v-4-0',
+      getApiKey: () => 'k',
+    });
+    expect(p.getInfo().dimensions).toBe(1536);
+  });
+
+  test('custom dimensions honoured', () => {
+    const p = new AzureCustomEmbeddingProvider({
+      endpoint: 'https://x/v1/',
+      deployment: 'embed-v-4-0',
+      dimensions: 1024,
+      getApiKey: () => 'k',
+    });
+    expect(p.getInfo().dimensions).toBe(1024);
+  });
+
+  test('invalid dimensions fall back to default', () => {
+    const p = new AzureCustomEmbeddingProvider({
+      endpoint: 'https://x/v1/',
+      deployment: 'embed-v-4-0',
+      dimensions: -5,
+      getApiKey: () => 'k',
+    });
+    expect(p.getInfo().dimensions).toBe(1536);
+  });
+});
+
+describe('AzureCustomEmbeddingProvider — getInfo + health', () => {
+  test('getInfo reports azure-custom + the user\'s deployment as modelId', () => {
+    const p = new AzureCustomEmbeddingProvider({
+      endpoint: 'https://x.azure.com/openai/v1/',
+      deployment: 'embed-v-4-0',
+      getApiKey: () => 'k',
+    });
+    expect(p.getInfo()).toMatchObject({
+      name: 'azure-custom',
+      modelId: 'embed-v-4-0',
+      dimensions: 1536,
+      isLocal: false,
+      distinguishesQueryAndDocument: false,
+    });
+  });
+
+  test('health() reflects missing key with a remediation message', async () => {
+    const getApiKey = jest.fn(() => null);
+    const p = new AzureCustomEmbeddingProvider({
+      endpoint: 'https://x/v1/', deployment: 'm', getApiKey,
+    });
+    const h = await p.health();
+    expect(h.ok).toBe(false);
+    expect(h.reason).toMatch(/API key/);
+  });
+});
+
+describe('AzureCustomEmbeddingProvider — HTTP call shape', () => {
+  let fetchImpl, getApiKey, provider;
+
+  beforeEach(() => {
+    fetchImpl = makeMockFetch({
+      status: 200,
+      json: { data: [{ embedding: Array(1536).fill(0.01) }] },
+    });
+    getApiKey = jest.fn(() => 'sk-azure-test');
+    provider = new AzureCustomEmbeddingProvider({
+      endpoint: 'https://loxia-ai-resource.services.ai.azure.com/openai/v1/',
+      deployment: 'embed-v-4-0',
+      getApiKey,
+      fetchImpl,
+    });
+  });
+
+  test('POSTs to the normalized URL with bearer auth', async () => {
+    await provider.embedDocuments(['hello']);
+    const [url, init] = fetchImpl.mock.calls[0];
+    expect(url).toBe('https://loxia-ai-resource.services.ai.azure.com/openai/v1/embeddings');
+    expect(init.headers.Authorization).toBe('Bearer sk-azure-test');
+  });
+
+  test('sends {input, model, encoding_format:"float"} body', async () => {
+    await provider.embedDocuments(['hello']);
+    const [, init] = fetchImpl.mock.calls[0];
+    expect(JSON.parse(init.body)).toEqual({
+      input: ['hello'],
+      model: 'embed-v-4-0',
+      encoding_format: 'float',
+    });
+  });
+
+  test('missing API key → AUTH_ERROR, no HTTP call', async () => {
+    getApiKey.mockReturnValue(null);
+    await expect(provider.embedDocuments(['x'])).rejects.toMatchObject({
+      code: EMBEDDING_ERROR_CODES.AUTH_ERROR,
+      transient: false,
+      provider: 'azure-custom',
+    });
+    expect(fetchImpl).not.toHaveBeenCalled();
+  });
+
+  test('API key is read fresh per call (rotation works)', async () => {
+    await provider.embedDocuments(['a']);
+    getApiKey.mockReturnValue('sk-azure-rotated');
+    await provider.embedDocuments(['b']);
+    expect(fetchImpl.mock.calls[1][1].headers.Authorization).toBe('Bearer sk-azure-rotated');
+  });
+
+  test('batches respect maxBatchSize', async () => {
+    // The mock returns one embedding per input it sees, so the per-batch
+    // count always matches what the base class expects.
+    const dynamicFetch = jest.fn(async (_url, init) => {
+      const body = JSON.parse(init.body);
+      const count = Array.isArray(body.input) ? body.input.length : 1;
+      return {
+        ok: true,
+        status: 200,
+        async json() {
+          return { data: Array(count).fill({ embedding: Array(1536).fill(0.01) }) };
+        },
+        async text() { return ''; },
+      };
+    });
+    provider = new AzureCustomEmbeddingProvider({
+      endpoint: 'https://x/v1/',
+      deployment: 'embed-v-4-0',
+      getApiKey,
+      fetchImpl: dynamicFetch,
+      maxBatchSize: 2,
+    });
+    await provider.embedDocuments(['a', 'b', 'c']);  // 2-chunk batch then single
+    expect(dynamicFetch).toHaveBeenCalledTimes(2);
+    const batches = dynamicFetch.mock.calls.map(c => JSON.parse(c[1].body).input.length);
+    expect(batches).toEqual([2, 1]);
+  });
+});
+
+describe('AzureCustomEmbeddingProvider — fingerprint', () => {
+  test('fingerprint encodes model + dimensions so swaps trigger re-embed', () => {
+    const a = new AzureCustomEmbeddingProvider({
+      endpoint: 'https://x/v1/', deployment: 'embed-v-4-0', dimensions: 1536, getApiKey: () => 'k',
+    });
+    const b = new AzureCustomEmbeddingProvider({
+      endpoint: 'https://x/v1/', deployment: 'embed-v-4-0', dimensions: 1024, getApiKey: () => 'k',
+    });
+    const c = new AzureCustomEmbeddingProvider({
+      endpoint: 'https://x/v1/', deployment: 'text-embedding-3-small', dimensions: 1536, getApiKey: () => 'k',
+    });
+    expect(a.getModelFingerprint()).toBe('azure-custom:embed-v-4-0@1536d');
+    expect(b.getModelFingerprint()).toBe('azure-custom:embed-v-4-0@1024d');
+    expect(c.getModelFingerprint()).toBe('azure-custom:text-embedding-3-small@1536d');
+    expect(a.getModelFingerprint()).not.toBe(b.getModelFingerprint());
+    expect(a.getModelFingerprint()).not.toBe(c.getModelFingerprint());
+  });
+});