Implement vector search experimental feature v2 (v1.6)

meilisearch/config.py

@@ -38,6 +38,7 @@ class Paths:
         separator_tokens = "separator-tokens"
         non_separator_tokens = "non-separator-tokens"
         swap = "swap-indexes"
+        embedders = "embedders"
 
     def __init__(
         self,

meilisearch/index.py

@@ -10,7 +10,7 @@
 from meilisearch.config import Config
 from meilisearch.errors import version_error_hint_message
 from meilisearch.models.document import Document, DocumentsResults
-from meilisearch.models.index import Faceting, IndexStats, Pagination, TypoTolerance
+from meilisearch.models.index import Embedders, Faceting, IndexStats, Pagination, TypoTolerance
 from meilisearch.models.task import Task, TaskInfo, TaskResults
 from meilisearch.task import TaskHandler
 
@@ -1757,6 +1757,71 @@ def reset_non_separator_tokens(self) -> TaskInfo:
 
         return TaskInfo(**task)
 
+    # EMBEDDERS SUB-ROUTES
+
+    def get_embedders(self) -> Embedders | None:
+        """Get embedders of the index.
+
+        Returns
+        -------
+        settings:
+            The embedders settings of the index.
+
+        Raises
+        ------
+        MeilisearchApiError
+            An error containing details about why Meilisearch can't process your request. Meilisearch error codes are described here: https://www.meilisearch.com/docs/reference/errors/error_codes#meilisearch-errors
+        """
+        response = self.http.get(self.__settings_url_for(self.config.paths.embedders))
+
+        if not response:
+            return None
+
+        return Embedders(embedders=response)
+
+    def update_embedders(self, body: Union[Mapping[str, Any], None]) -> TaskInfo:
+        """Update embedders of the index.
+
+        Parameters
+        ----------
+        body: dict
+            Dictionary containing the embedders.
+
+        Returns
+        -------
+        task_info:
+            TaskInfo instance containing information about a task to track the progress of an asynchronous process.
+            https://www.meilisearch.com/docs/reference/api/tasks#get-one-task
+
+        Raises
+        ------
+        MeilisearchApiError
+            An error containing details about why Meilisearch can't process your request. Meilisearch error codes are described here: https://www.meilisearch.com/docs/reference/errors/error_codes#meilisearch-errors
+        """
+        task = self.http.patch(self.__settings_url_for(self.config.paths.embedders), body)
+
+        return TaskInfo(**task)
+
+    def reset_embedders(self) -> TaskInfo:
+        """Reset embedders of the index to default values.
+
+        Returns
+        -------
+        task_info:
+            TaskInfo instance containing information about a task to track the progress of an asynchronous process.
+            https://www.meilisearch.com/docs/reference/api/tasks#get-one-task
+
+        Raises
+        ------
+        MeilisearchApiError
+            An error containing details about why Meilisearch can't process your request. Meilisearch error codes are described here: https://www.meilisearch.com/docs/reference/errors/error_codes#meilisearch-errors
+        """
+        task = self.http.delete(
+            self.__settings_url_for(self.config.paths.embedders),
+        )
+
+        return TaskInfo(**task)
+
     @staticmethod
     def _batch(
         documents: Sequence[Mapping[str, Any]], batch_size: int

meilisearch/models/index.py

@@ -1,6 +1,6 @@
 from __future__ import annotations
 
-from typing import Any, Dict, Iterator, List, Optional
+from typing import Any, Dict, Iterator, List, Optional, Union
 
 from camel_converter import to_snake
 from camel_converter.pydantic_base import CamelBase
@@ -46,3 +46,26 @@ class TypoTolerance(CamelBase):
     disable_on_attributes: Optional[List[str]] = None
     disable_on_words: Optional[List[str]] = None
     min_word_size_for_typos: Optional[MinWordSizeForTypos] = None
+
+
+class OpenAiEmbedder(CamelBase):
+    source: str = "openAi"
+    model: Optional[str] = None  # Defaults to text-embedding-ada-002
+    api_key: Optional[str] = None  # Can be provided through a CLI option or environment variable
+    document_template: Optional[str] = None
+
+
+class HuggingFaceEmbedder(CamelBase):
+    source: str = "huggingFace"
+    model: Optional[str] = None  # Defaults to BAAI/bge-base-en-v1.5
+    revision: Optional[str] = None
+    document_template: Optional[str] = None
+
+
+class UserProvidedEmbedder(CamelBase):
+    source: str = "userProvided"
+    dimensions: int
+
+
+class Embedders(CamelBase):
+    embedders: Dict[str, Union[OpenAiEmbedder, HuggingFaceEmbedder, UserProvidedEmbedder]]

tests/conftest.py

@@ -129,12 +129,21 @@ def index_maker(index_uid=common.INDEX_UID, documents=small_movies):
 
 @fixture(scope="function")
 def index_with_documents_and_vectors(empty_index, small_movies):
-    small_movies[0]["_vectors"] = [0.1, 0.2]
+    small_movies[0]["_vectors"] = {"default": [0.1, 0.2]}
 
     def index_maker(index_uid=common.INDEX_UID, documents=small_movies):
         index = empty_index(index_uid)
-        task = index.add_documents(documents)
-        index.wait_for_task(task.task_uid)
+        settings_update_task = index.update_embedders(
+            {
+                "default": {
+                    "source": "userProvided",
+                    "dimensions": 2,
+                }
+            }
+        )
+        index.wait_for_task(settings_update_task.task_uid)
+        document_addition_task = index.add_documents(documents)
+        index.wait_for_task(document_addition_task.task_uid)
         return index
 
     return index_maker

tests/index/test_index_search_meilisearch.py

@@ -459,13 +459,9 @@ def test_attributes_to_search_on_search_no_match(index_with_documents):
     assert response["hits"] == []
 
 
-@pytest.mark.xfail(
-    strict=True, reason="https://github.com/meilisearch/meilisearch-python/issues/901"
-)
 @pytest.mark.usefixtures("enable_vector_search")
 def test_vector_search(index_with_documents_and_vectors):
     response = index_with_documents_and_vectors().search(
-        "How to Train Your Dragon", opt_params={"vector": [0.1, 0.2]}
+        "", opt_params={"vector": [0.1, 0.2], "hybrid": {"semanticRatio": 1.0}}
     )
-    assert response["hits"][0]["id"] == "287947"
     assert response["vector"] == [0.1, 0.2]

tests/settings/test_settings_embedders.py

@@ -0,0 +1,52 @@
+import pytest
+
+from meilisearch.models.index import Embedders
+
+NEW_EMBEDDERS = {
+    "default": {
+        "source": "userProvided",
+        "dimensions": 1,
+    }
+}
+
+
+@pytest.mark.usefixtures("enable_vector_search")
+def test_get_default_embedders(empty_index):
+    """Tests getting default embedders."""
+    response = empty_index().get_embedders()
+
+    assert response is None
+
+
+@pytest.mark.usefixtures("enable_vector_search")
+def test_update_embedders_with_user_provided_source(empty_index):
+    """Tests updating embedders."""
+    index = empty_index()
+    response_update = index.update_embedders(NEW_EMBEDDERS)
+    update = index.wait_for_task(response_update.task_uid)
+    response_get = index.get_embedders()
+    print(response_get)
+    assert update.status == "succeeded"
+    assert response_get == Embedders(embedders=NEW_EMBEDDERS)
+
+
+@pytest.mark.usefixtures("enable_vector_search")
+def test_reset_embedders(empty_index):
+    """Tests resetting the typo_tolerance setting to its default value."""
+    index = empty_index()
+
+    # Update the settings
+    response_update = index.update_embedders(NEW_EMBEDDERS)
+    update1 = index.wait_for_task(response_update.task_uid)
+    # Get the setting after update
+    response_get = index.get_embedders()
+    # Reset the setting
+    response_reset = index.reset_embedders()
+    update2 = index.wait_for_task(response_reset.task_uid)
+    # Get the setting after reset
+    response_last = index.get_embedders()
+
+    assert update1.status == "succeeded"
+    assert response_get == Embedders(embedders=NEW_EMBEDDERS)
+    assert update2.status == "succeeded"
+    assert response_last is None