debug

closes #1084

docs/src/embeddings/default_embedding_functions.md

@@ -224,7 +224,6 @@ This embedding function supports ingesting images as both bytes and urls. You ca
 !!! info
     LanceDB supports ingesting images directly from accessible links.
 
-
 ```python
 
 db = lancedb.connect(tmp_path)
@@ -290,4 +289,67 @@ print(actual.label)
 
 ```
 
+### Imagebind embeddings
+We have support for [imagebind](https://github.com/facebookresearch/ImageBind) model embeddings. You can download our version of the packaged model via - `pip install imagebind-packaged==0.1.2`.
+
+This function is registered as `imagebind` and supports Audio, Video and Text modalities(extending to Thermal,Depth,IMU data):
+
+| Parameter | Type | Default Value | Description |
+|---|---|---|---|
+| `name` | `str` | `"imagebind_huge"` | Name of the model. |
+| `device` | `str` | `"cpu"` | The device to run the model on. Can be `"cpu"` or `"gpu"`. |
+| `normalize` | `bool` | `False` | set to `True` to normalize your inputs before model ingestion. |
+
+Below is an example demonstrating how the API works:
+
+```python
+db = lancedb.connect(tmp_path)
+registry = EmbeddingFunctionRegistry.get_instance()
+func = registry.get("imagebind").create()
+
+class ImageBindModel(LanceModel):
+    text: str
+    image_uri: str = func.SourceField()
+    audio_path: str
+    vector: Vector(func.ndims()) = func.VectorField()
+
+# add locally accessible image paths
+text_list=["A dog.", "A car", "A bird"]
+image_paths=[".assets/dog_image.jpg", ".assets/car_image.jpg", ".assets/bird_image.jpg"]
+audio_paths=[".assets/dog_audio.wav", ".assets/car_audio.wav", ".assets/bird_audio.wav"]
+
+# Load data
+inputs = [
+    {"text": a, "audio_path": b, "image_uri": c}
+    for a, b, c in zip(text_list, audio_paths, image_paths)
+]
+
+#create table and add data
+table = db.create_table("img_bind", schema=ImageBindModel)
+table.add(inputs)
+```
+
+Now, we can search using any modality:
+
+#### image search
+```python
+query_image = "./assets/dog_image2.jpg" #download an image and enter that path here
+actual = table.search(query_image).limit(1).to_pydantic(ImageBindModel)[0]
+print(actual.text == "dog")
+```
+#### audio search
+
+```python
+query_audio = "./assets/car_audio2.wav" #download an audio clip and enter path here
+actual = table.search(query_audio).limit(1).to_pydantic(ImageBindModel)[0]
+print(actual.text == "car")
+```
+#### Text search
+You can add any input query and fetch the result as follows:
+```python
+query = "an animal which flies and tweets" 
+actual = table.search(query).limit(1).to_pydantic(ImageBindModel)[0]
+print(actual.text == "bird")
+```
+
 If you have any questions about the embeddings API, supported models, or see a relevant model missing, please raise an issue [on GitHub](https://github.com/lancedb/lancedb/issues).

python/python/lancedb/embeddings/imagebind.py

@@ -31,7 +31,7 @@ class ImageBindEmbeddings(EmbeddingFunction):
     six different modalities: images, text, audio, depth, thermal, and IMU data
 
     to download package, run :
-        `pip install imagebind@git+https://github.com/raghavdixit99/ImageBind`
+        `pip install imagebind-packaged==0.1.2`
     """
 
     name: str = "imagebind_huge"

python/python/lancedb/embeddings/openai.py

@@ -113,5 +113,5 @@ class OpenAIEmbeddings(TextEmbeddingFunction):
         if self.organization:
             kwargs["organization"] = self.organization
         if self.api_key:
-            kwargs["api_key"] = self
+            kwargs["api_key"] = self.api_key
         return openai.OpenAI(**kwargs)

python/python/lancedb/query.py

@@ -271,7 +271,8 @@ class LanceQueryBuilder(ABC):
         and also the "_distance" column which is the distance between the query
         vector and the returned vectors.
         """
-        raise NotImplementedError
+        # raise NotImplementedError
+        self.to_arrow()
 
     def to_list(self) -> List[dict]:
         """
@@ -434,12 +435,12 @@ class LanceVectorQueryBuilder(LanceQueryBuilder):
         self._vector_column = vector_column
         self._prefilter = False
 
-    def metric(self, metric: Literal["L2", "cosine"]) -> LanceVectorQueryBuilder:
+    def metric(self, metric: Literal["L2", "cosine", "dot"]) -> LanceVectorQueryBuilder:
         """Set the distance metric to use.
 
         Parameters
         ----------
-        metric: "L2" or "cosine"
+        metric: "L2" or "cosine" or "dot"
             The distance metric to use. By default "L2" is used.
 
         Returns

python/python/lancedb/remote/table.py

@@ -68,10 +68,16 @@ class RemoteTable(Table):
 
     def list_indices(self):
         """List all the indices on the table"""
-        print(self._name)
         resp = self._conn._client.post(f"/v1/table/{self._name}/index/list/")
         return resp
 
+    def index_stats(self, index_uuid: str):
+        """List all the indices on the table"""
+        resp = self._conn._client.post(
+            f"/v1/table/{self._name}/index/{index_uuid}/stats/"
+        )
+        return resp
+
     def create_scalar_index(
         self,
         column: str,
@@ -290,6 +296,7 @@ class RemoteTable(Table):
         return LanceVectorQueryBuilder(self, query, vector_column_name)
 
     def _execute_query(self, query: Query) -> pa.Table:
+        print("query metric", query.metric)
         if (
             query.vector is not None
             and len(query.vector) > 0

python/python/lancedb/table.py

@@ -118,7 +118,8 @@ def _append_vector_col(data: pa.Table, metadata: dict, schema: Optional[pa.Schem
     functions = EmbeddingFunctionRegistry.get_instance().parse_functions(metadata)
     for vector_column, conf in functions.items():
         func = conf.function
-        if vector_column not in data.column_names:
+        no_vector_column = vector_column not in data.column_names
+        if no_vector_column or pc.all(pc.is_null(data[vector_column])).as_py():
             col_data = func.compute_source_embeddings_with_retry(
                 data[conf.source_column]
             )
@@ -126,9 +127,16 @@ def _append_vector_col(data: pa.Table, metadata: dict, schema: Optional[pa.Schem
                 dtype = schema.field(vector_column).type
             else:
                 dtype = pa.list_(pa.float32(), len(col_data[0]))
-            data = data.append_column(
-                pa.field(vector_column, type=dtype), pa.array(col_data, type=dtype)
-            )
+            if no_vector_column:
+                data = data.append_column(
+                    pa.field(vector_column, type=dtype), pa.array(col_data, type=dtype)
+                )
+            else:
+                data = data.set_column(
+                    data.column_names.index(vector_column),
+                    pa.field(vector_column, type=dtype),
+                    pa.array(col_data, type=dtype),
+                )
     return data
 
 
@@ -1514,7 +1522,7 @@ class LanceTable(Table):
 
     def _execute_query(self, query: Query) -> pa.Table:
         ds = self.to_lance()
-
+        print("metric:", query.metric)  
         return ds.to_table(
             columns=query.columns,
             filter=query.filter,

python/python/tests/test_embeddings.py

@@ -11,6 +11,7 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 import sys
+from typing import List, Union
 
 import lance
 import lancedb
@@ -23,6 +24,8 @@ from lancedb.embeddings import (
     EmbeddingFunctionRegistry,
     with_embeddings,
 )
+from lancedb.embeddings.base import TextEmbeddingFunction
+from lancedb.embeddings.registry import get_registry, register
 from lancedb.pydantic import LanceModel, Vector
 
 
@@ -112,3 +115,34 @@ def test_embedding_function_rate_limit(tmp_path):
     table.add([{"text": "hello world"}])
     table.add([{"text": "hello world"}])
     assert len(table) == 2
+
+
+def test_add_optional_vector(tmp_path):
+    @register("mock-embedding")
+    class MockEmbeddingFunction(TextEmbeddingFunction):
+        def ndims(self):
+            return 128
+
+        def generate_embeddings(
+            self, texts: Union[List[str], np.ndarray]
+        ) -> List[np.array]:
+            """
+            Generate the embeddings for the given texts
+            """
+            return [np.random.randn(self.ndims()).tolist() for _ in range(len(texts))]
+
+    registry = get_registry()
+    model = registry.get("mock-embedding").create()
+
+    class LanceSchema(LanceModel):
+        id: str
+        vector: Vector(model.ndims()) = model.VectorField(default=None)
+        text: str = model.SourceField()
+
+    db = lancedb.connect(tmp_path)
+    tbl = db.create_table("optional_vector", schema=LanceSchema)
+
+    # add works
+    expected = LanceSchema(id="id", text="text")
+    tbl.add([expected])
+    assert not (np.abs(tbl.to_pandas()["vector"][0]) < 1e-6).all()