test: use inference-testing fakes and add e2e retry, base64, live suites

.github/workflows/ci.yml

@@ -21,8 +21,8 @@ jobs:
           bun-version: latest
       - run: bun install --frozen-lockfile
       - run: bunx tsc --noEmit
-      - run: bun test ./src
-      - run: bunx prettier --check src
+      - run: bun run test
+      - run: bunx prettier --check src tests
       - run: bun run build
       - name: Import packed dist with Node
         run: |

IMPLEMENTATION.md

@@ -101,8 +101,14 @@ package depends only on the published surface).
 
 ## Tests
 
-`bun run test` is `bun test ./src`. Coverage is the client contract: empty
-input, URL and body knobs, auth header presence, batch split and order,
-`batchSize` rejection, malformed / short / duplicate-index replies,
-index-not-position placement, 429 retry, 401 abort, `probeEmbedDims`,
-overridable `Retry-After`.
+`bun run test` runs `bun test ./src ./tests`.
+
+- `src/embed.test.ts` — units on `@intx/inference-testing` fakes: index
+  placement across batches, short reply, base64 decode.
+- `tests/retry.test.ts` — 429 retried once after `Retry-After` on the
+  harness clock; 401 rejects without retry.
+- `tests/base64.test.ts` — `base64` and `float` agree over a real HTTP stub.
+- `live/embeddings.test.ts` (`bun run test:live`, not in `test` or CI) —
+  round trip against a real `/v1/embeddings` (`EMBEDDING_E2E_BASE_URL`,
+  required; `EMBEDDING_E2E_MODEL`, default `nomic-embed-text`).
+  Skips when unset or unreachable.

live/embeddings.test.ts

@@ -0,0 +1,49 @@
+import { describe, expect, test } from "bun:test";
+
+import { embedTexts, probeEmbedDims } from "../src/index";
+
+// Any OpenAI-compatible `/v1/embeddings` server. Skips unless
+// EMBEDDING_E2E_BASE_URL is set, so it never loads a model on this machine.
+const baseURL = process.env.EMBEDDING_E2E_BASE_URL;
+const model = process.env.EMBEDDING_E2E_MODEL ?? "nomic-embed-text";
+
+// Skips unless the server answers and lists the model.
+const reachable =
+  baseURL !== undefined &&
+  (await fetch(`${baseURL}/models`, {
+    signal: AbortSignal.timeout(1_000),
+  }).then(
+    async (res) => res.ok && (await res.text()).includes(`"${model}`),
+    () => false,
+  ));
+const config = { baseURL: baseURL ?? "", model };
+
+function cosine(a: number[], b: number[]): number {
+  let dot = 0;
+  let na = 0;
+  let nb = 0;
+  a.forEach((x, i) => {
+    const y = b[i] ?? 0;
+    dot += x * y;
+    na += x * x;
+    nb += y * y;
+  });
+  return dot / Math.sqrt(na * nb);
+}
+
+describe.skipIf(!reachable)(`live /v1/embeddings at ${config.baseURL}`, () => {
+  test("embeds three texts in input order at the probed width", async () => {
+    const texts = ["the cat sat", "quarterly revenue grew", "rust borrowck"];
+    const dims = await probeEmbedDims(config);
+    const vectors = await embedTexts(texts, config);
+
+    expect(vectors).toHaveLength(texts.length);
+    for (const vector of vectors) expect(vector).toHaveLength(dims);
+
+    // Order: each batched vector matches that text embedded alone.
+    for (const [i, text] of texts.entries()) {
+      const [alone] = await embedTexts([text], config);
+      expect(cosine(vectors[i] ?? [], alone ?? [])).toBeGreaterThan(0.99);
+    }
+  }, 120_000); // a cold local model can take a while to page in
+});

package.json

@@ -17,7 +17,8 @@
     "build": "tsc -p tsconfig.build.json",
     "prepack": "bun run build",
     "typecheck": "tsc --noEmit",
-    "test": "bun test ./src",
+    "test": "bun test ./src ./tests",
+    "test:live": "bun test ./live",
     "format": "prettier -w ."
   },
   "dependencies": {