From 607cb462adfed97c77d63ddb44e3e5cea42e6b84 Mon Sep 17 00:00:00 2001
From: carlos-alm <127798846+carlos-alm@users.noreply.github.com>
Date: Sun, 3 May 2026 06:17:55 -0600
Subject: [PATCH] fix(search): point jina-code alias at published HF repo

The `jina-code` alias mapped to `Xenova/jina-embeddings-v2-base-code`,
which 404s on Hugging Face. Point it at `jinaai/jina-embeddings-v2-base-code`,
the published code embedding model, drop the stale "requires HF token" note
in the README, and add a regression test for the alias.

Fixes #1025.
---
 README.md                               | 2 +-
 src/domain/search/models.ts             | 2 +-
 tests/search/embedding-strategy.test.ts | 4 ++++
 3 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/README.md b/README.md
index 7aaf2fdc..f2596403 100644
--- a/README.md
+++ b/README.md
@@ -428,7 +428,7 @@ A single trailing semicolon is ignored (falls back to single-query mode). The `-
 | `minilm` | all-MiniLM-L6-v2 | 384 | ~23 MB | Apache-2.0 | Fastest, good for quick iteration |
 | `jina-small` | jina-embeddings-v2-small-en | 512 | ~33 MB | Apache-2.0 | Better quality, still small |
 | `jina-base` | jina-embeddings-v2-base-en | 768 | ~137 MB | Apache-2.0 | High quality, 8192 token context |
-| `jina-code` | jina-embeddings-v2-base-code | 768 | ~137 MB | Apache-2.0 | Best for code search, trained on code+text (requires HF token) |
+| `jina-code` | jina-embeddings-v2-base-code | 768 | ~137 MB | Apache-2.0 | Best for code search, trained on code+text |
 | `nomic` | nomic-embed-text-v1 | 768 | ~137 MB | Apache-2.0 | Good quality, 8192 context |
 | `nomic-v1.5` (default) | nomic-embed-text-v1.5 | 768 | ~137 MB | Apache-2.0 | **Improved nomic, Matryoshka dimensions** |
 | `bge-large` | bge-large-en-v1.5 | 1024 | ~335 MB | MIT | Best general retrieval, top MTEB scores |
diff --git a/src/domain/search/models.ts b/src/domain/search/models.ts
index 54ce5956..920e6e62 100644
--- a/src/domain/search/models.ts
+++ b/src/domain/search/models.ts
@@ -42,7 +42,7 @@ export const MODELS: Record<string, ModelConfig> = {
     quantized: false,
   },
   'jina-code': {
-    name: 'Xenova/jina-embeddings-v2-base-code',
+    name: 'jinaai/jina-embeddings-v2-base-code',
     dim: 768,
     contextWindow: 8192,
     desc: 'Code-aware (~137MB). Trained on code+text, best for code search.',
diff --git a/tests/search/embedding-strategy.test.ts b/tests/search/embedding-strategy.test.ts
index 0f829bc0..f7a79f93 100644
--- a/tests/search/embedding-strategy.test.ts
+++ b/tests/search/embedding-strategy.test.ts
@@ -143,6 +143,10 @@ describe('MODELS contextWindow', () => {
       expect(config.contextWindow, `${key} missing contextWindow`).toBeGreaterThan(0);
     }
   });
+
+  test('jina-code points to the published code embedding model', () => {
+    expect(MODELS['jina-code'].name).toBe('jinaai/jina-embeddings-v2-base-code');
+  });
 });
 
 describe('buildEmbeddings with structured strategy', () => {