From 5f9a8af27c66d8730b9e7cf3c29cac1f2bd237cf Mon Sep 17 00:00:00 2001
From: Mohammad Hijjawi <51020382+MohammadHijjawi97@users.noreply.github.com>
Date: Thu, 22 Oct 2026 19:42:21 +0100
Subject: [PATCH] fix(python): keep embedding functions registered without an
 alias readable (#4356)

`EmbeddingFunctionRegistry.register()` documents that the class name is
used when no alias is given, and it does register the class under that
name. But it stored the alias itself (`None`) as
`__embedding_function_registry_alias__`, which is what
`function_to_metadata()` writes into the table schema. A table using
such a function gets `"name": null` in its `embedding_functions`
metadata, and any later read of the functions fails:

```python
@get_registry().register()
class MyEmbed(TextEmbeddingFunction): ...

f = MyEmbed.create()
class Schema(LanceModel):
    text: str = f.SourceField()
    vector: Vector(4) = f.VectorField()

tbl = db.create_table("t", schema=Schema)
tbl.add([{"text": "hi"}])   # KeyError: None
tbl.search("hi")            # KeyError: None
```

This stores the key the class is registered under instead, so the class
name round-trips. For aliased classes the key is the alias, so their
metadata is unchanged. The TypeScript registry already falls back to
`ctor.name` in the same way.

Added `test_embedding_function_registered_without_alias`, which creates
a table with such a function, reopens it, and adds a row. It fails with
`KeyError` before the change. `test_embeddings.py`, `test_pydantic.py`,
`test_table.py` and `test_db.py` pass, and `ruff format --check` / `ruff
check` are clean.
---
 python/python/lancedb/embeddings/registry.py |  4 ++-
 python/python/tests/test_embeddings.py       | 31 ++++++++++++++++++++
 2 files changed, 34 insertions(+), 1 deletion(-)

diff --git a/python/python/lancedb/embeddings/registry.py b/python/python/lancedb/embeddings/registry.py
index 1dc7ffcb8..3b940fe2e 100644
--- a/python/python/lancedb/embeddings/registry.py
+++ b/python/python/lancedb/embeddings/registry.py
@@ -73,7 +73,9 @@ class EmbeddingFunctionRegistry:
                 raise KeyError(f"{cls.__name__} was already registered")
             key = alias or cls.__name__
             self._functions[key] = cls
-            cls.__embedding_function_registry_alias__ = alias
+            # Store the name the class is registered under, so the table
+            # metadata written for it can be looked up again when reading.
+            cls.__embedding_function_registry_alias__ = key
             return cls
 
         return decorator
diff --git a/python/python/tests/test_embeddings.py b/python/python/tests/test_embeddings.py
index c42172f77..51907f141 100644
--- a/python/python/tests/test_embeddings.py
+++ b/python/python/tests/test_embeddings.py
@@ -64,6 +64,37 @@ def test_embedding_function(tmp_path):
     assert np.allclose(actual, expected)
 
 
+def test_embedding_function_registered_without_alias(tmp_path):
+    registry = EmbeddingFunctionRegistry.get_instance()
+
+    @registry.register()
+    class NoAliasEmbeddingFunction(TextEmbeddingFunction):
+        def ndims(self):
+            return 2
+
+        def generate_embeddings(self, texts):
+            return [np.array([1.0, 2.0]) for _ in texts]
+
+    assert registry.get("NoAliasEmbeddingFunction") is NoAliasEmbeddingFunction
+
+    func = NoAliasEmbeddingFunction.create()
+
+    class Schema(LanceModel):
+        text: str = func.SourceField()
+        vector: Vector(2) = func.VectorField()
+
+    db = lancedb.connect(tmp_path)
+    db.create_table("test", schema=Schema)
+
+    # The class name is written to the table metadata, so the function can be
+    # resolved again when the table is reopened.
+    table = lancedb.connect(tmp_path).open_table("test")
+    table.add([{"text": "hello"}])
+    conf = table.embedding_functions["vector"]
+    assert isinstance(conf.function, NoAliasEmbeddingFunction)
+    assert table.to_arrow()["vector"].to_pylist() == [[1.0, 2.0]]
+
+
 def test_instructor_ndims_uses_instruction():
     instructor = get_registry().get("instructor").create()
     model = MagicMock()