Feat: destination vector db add option for embedding without field names

airbyte_cdk/destinations/vector_db_based/config.py

@@ -108,6 +108,12 @@ class ProcessingConfigModel(BaseModel):
         always_show=True,
         examples=["text", "user.name", "users.*.name"],
     )
+    omit_field_names_from_embeddings: bool = Field(
+        default=False,
+        title="Omit field names from embeddings",
+        description="Do not include the field names in the text that gets embedded. By default field names are embedded e.g 'user: name, user.email: [email protected]'. If set to true, only the values are embedded e.g. 'name, [email protected]'.",
+        always_show=True,
+    )
     metadata_fields: Optional[List[str]] = Field(
         default=[],
         title="Fields to store as metadata",

airbyte_cdk/destinations/vector_db_based/document_processor.py

@@ -5,7 +5,7 @@
 import json
 import logging
 from dataclasses import dataclass
-from typing import Any, Dict, List, Mapping, Optional, Tuple
+from typing import Any, Dict, List, Mapping, Optional, Tuple, Union
 
 import dpath
 from airbyte_cdk.destinations.vector_db_based.config import (
@@ -125,6 +125,7 @@ def __init__(self, config: ProcessingConfigModel, catalog: ConfiguredAirbyteCata
         self.text_fields = config.text_fields
         self.metadata_fields = config.metadata_fields
         self.field_name_mappings = config.field_name_mappings
+        self.omit_field_names_from_embeddings = config.omit_field_names_from_embeddings
         self.logger = logging.getLogger("airbyte.document_processor")
 
     def process(self, record: AirbyteRecordMessage) -> Tuple[List[Chunk], Optional[str]]:
@@ -162,10 +163,21 @@ def _generate_document(self, record: AirbyteRecordMessage) -> Optional[Document]
         relevant_fields = self._extract_relevant_fields(record, self.text_fields)
         if len(relevant_fields) == 0:
             return None
-        text = stringify_dict(relevant_fields)
+        if not self.omit_field_names_from_embeddings:
+            text = stringify_dict(relevant_fields)
+        else:
+            text = self._extract_values_from_dict(relevant_fields)
         metadata = self._extract_metadata(record)
         return Document(page_content=text, metadata=metadata)
 
+    def _extract_values_from_dict(self, data: Union[dict, list, Any]) -> str:
+        if isinstance(data, dict):
+            return "\n".join(self._extract_values_from_dict(value) for value in data.values())
+        elif isinstance(data, list):
+            return "\n".join(self._extract_values_from_dict(item) for item in data)
+        else:
+            return str(data)
+
     def _extract_relevant_fields(
         self, record: AirbyteRecordMessage, fields: Optional[List[str]]
     ) -> Dict[str, Any]:

unit_tests/destinations/vector_db_based/config_test.py

@@ -242,6 +242,13 @@ def test_json_schema_generation():
                         "type": "array",
                         "items": {"type": "string"},
                     },
+                    "omit_field_names_from_embeddings": {
+                        "title": "Omit field names from embeddings",
+                        "description": "Do not include the field names in the text that gets embedded. By default field names are embedded e.g 'user: name, user.email: [email protected]'. If set to true, only the values are embedded e.g. 'name, [email protected]'.",
+                        "default": False,
+                        "always_show": True,
+                        "type": "boolean",
+                    },
                     "metadata_fields": {
                         "title": "Fields to store as metadata",
                         "description": "List of fields in the record that should be stored as metadata. The field list is applied to all streams in the same way and non-existing fields are ignored. If none are defined, all fields are considered metadata fields. When specifying text fields, you can access nested fields in the record by using dot notation, e.g. `user.name` will access the `name` field in the `user` object. It's also possible to use wildcards to access all fields in an object, e.g. `users.*.name` will access all `names` fields in all entries of the `users` array. When specifying nested paths, all matching values are flattened into an array set to a field named by the path.",

unit_tests/destinations/vector_db_based/document_processor_test.py

@@ -193,6 +193,7 @@ def test_complex_text_fields():
         "non.*.existing",
     ]
     processor.metadata_fields = ["non_text", "non_text_2", "id"]
+    processor.omit_field_names_from_embeddings = False
 
     chunks, _ = processor.process(record)
 
@@ -213,6 +214,56 @@ def test_complex_text_fields():
     }
 
 
+def test_complex_text_fields_omit_field_names():
+    processor = initialize_processor()
+
+    record = AirbyteRecordMessage(
+        stream="stream1",
+        namespace="namespace1",
+        data={
+            "id": 1,
+            "nested": {
+                "texts": [
+                    {"text": "This is the text"},
+                    {"text": "And another"},
+                ]
+            },
+            "non_text": "a",
+            "non_text_2": 1,
+            "text": "This is the regular text",
+            "other_nested": {"non_text": {"a": "xyz", "b": "abc"}},
+        },
+        emitted_at=1234,
+    )
+
+    processor.text_fields = [
+        "nested.texts.*.text",
+        "text",
+        "other_nested.non_text",
+        "non.*.existing",
+    ]
+    processor.metadata_fields = ["non_text", "non_text_2", "id"]
+    processor.omit_field_names_from_embeddings = True
+
+    chunks, _ = processor.process(record)
+
+    assert len(chunks) == 1
+    assert (
+        chunks[0].page_content
+        == """This is the text
+And another
+This is the regular text
+xyz
+abc"""
+    )
+    assert chunks[0].metadata == {
+        "id": 1,
+        "non_text": "a",
+        "non_text_2": 1,
+        "_ab_stream": "namespace1_stream1",
+    }
+
+
 def test_no_text_fields():
     processor = initialize_processor()