[bug] IntegrityError When Adding Texts to Vector Store with Metadata #74

New Issue

2026-02-16T05:16:27-05:00

yindo commented

2026-02-16 05:16:27 -05:00

Originally created by @Ocyss on GitHub (May 29, 2025).

I have written the following code based on the example documentation. However, I encountered an error when attempting to add texts to the vector store with metadata.

Code

Here is the relevant code snippet:

pg_engine_entity = PGEngine.from_engine(engine)

await pg_engine_entity.ainit_vectorstore_table(
    table_name=TABLE_NAME_ENTITY,
    vector_size=1024,
    overwrite_existing=True,
    metadata_columns=[
        Column("entity_id", "INTEGER", nullable=False)
    ],
)

vectorstore_entity = await PGVectorStore.create(
    engine=pg_engine_entity,
    table_name=TABLE_NAME_ENTITY,
    embedding_service=openai_embeddings,
)

vid = (
    await vectorstore.aadd_texts(
        [memory_flow.content],
        [
            {
                "entity_id": entity_id
            }
        ],
        [uuid.uuid4()],
    )
)[0]

Error Message

When executing the code, I received the following error:

IntegrityError: (sqlalchemy.dialects.postgresql.asyncpg.IntegrityError) <class 'asyncpg.exceptions.NotNullViolationError'>: null value in column "entity_id" of relation "VectorEntity" violates not-null constraint
DETAIL: Failing row contains (63c145e7-8663-471e-845e-d96714f651e2, 一条测试。, [-0.030918628,0.017731816,-0.03639662,-0.00945296,-0.03324449,-0..., null, null, {"entity_id": 1006}).

Expected Behavior

The text and metadata should be successfully added to the vector store without violating the not-null constraint on the entity_id column.

Actual Behavior

The operation fails with an IntegrityError due to a null value in the entity_id column, despite passing a value in the metadata.

Environment

Library Version:
- SQLModel: 0.0.24
- asyncpg: 0.30.0
- FastAPI: 0.115.12
- LangChain: 0.3.25
- Pydantic: 2.11.4
Database: PostgreSQL
Python Version: 3.13.2
Other Relevant Info: Package manager: uv

Additional Notes

It seems that the metadata value for entity_id is not being correctly mapped to the column in the database.
I suspect there might be an issue with how the metadata is processed or inserted into the table.

Request for Help

Could you please help identify why the entity_id value from the metadata is not being inserted into the database column? Any guidance or fixes would be greatly appreciated!

Additional Question on SQL Construction

I also noticed that the add_embeddings method uses f-strings to manually concatenate SQL statements. Could this potentially lead to security issues, and is there a reason for not using SQLAlchemy's table objects or parameterized queries?

Originally created by @Ocyss on GitHub (May 29, 2025). I have written the following code based on the example documentation. However, I encountered an error when attempting to add texts to the vector store with metadata. ## Code Here is the relevant code snippet: ```python pg_engine_entity = PGEngine.from_engine(engine) await pg_engine_entity.ainit_vectorstore_table( table_name=TABLE_NAME_ENTITY, vector_size=1024, overwrite_existing=True, metadata_columns=[ Column("entity_id", "INTEGER", nullable=False) ], ) vectorstore_entity = await PGVectorStore.create( engine=pg_engine_entity, table_name=TABLE_NAME_ENTITY, embedding_service=openai_embeddings, ) vid = ( await vectorstore.aadd_texts( [memory_flow.content], [ { "entity_id": entity_id } ], [uuid.uuid4()], ) )[0] ``` ## Error Message When executing the code, I received the following error: ``` IntegrityError: (sqlalchemy.dialects.postgresql.asyncpg.IntegrityError) <class 'asyncpg.exceptions.NotNullViolationError'>: null value in column "entity_id" of relation "VectorEntity" violates not-null constraint DETAIL: Failing row contains (63c145e7-8663-471e-845e-d96714f651e2, 一条测试。, [-0.030918628,0.017731816,-0.03639662,-0.00945296,-0.03324449,-0..., null, null, {"entity_id": 1006}). ``` ## Expected Behavior The text and metadata should be successfully added to the vector store without violating the `not-null` constraint on the `entity_id` column. ## Actual Behavior The operation fails with an `IntegrityError` due to a `null` value in the `entity_id` column, despite passing a value in the metadata. ## Environment - **Library Version**: - SQLModel: 0.0.24 - asyncpg: 0.30.0 - FastAPI: 0.115.12 - LangChain: 0.3.25 - Pydantic: 2.11.4 - **Database**: PostgreSQL - **Python Version**: 3.13.2 - **Other Relevant Info**: Package manager: uv ## Additional Notes - It seems that the metadata value for `entity_id` is not being correctly mapped to the column in the database. - I suspect there might be an issue with how the metadata is processed or inserted into the table. ## Request for Help Could you please help identify why the `entity_id` value from the metadata is not being inserted into the database column? Any guidance or fixes would be greatly appreciated! ## Additional Question on SQL Construction I also noticed that the `add_embeddings` method uses f-strings to manually concatenate SQL statements. Could this potentially lead to security issues, and is there a reason for not using SQLAlchemy's table objects or parameterized queries?

yindo closed this issue

2026-02-16 05:16:27 -05:00

yindo commented

2026-02-16 05:16:27 -05:00

@averikitsch commented on GitHub (Jul 2, 2025):

Hi @Ocyss ,
You'll need to tell the vector store about the use of the metadata column "entity_id" like

vectorstore_entity = await PGVectorStore.create(
    engine=pg_engine_entity,
    table_name=TABLE_NAME_ENTITY,
    embedding_service=openai_embeddings,
    metadata_columns=["entity_id"]
)

Let me know if this does not fix your issue.

@averikitsch commented on GitHub (Jul 2, 2025): Hi @Ocyss , You'll need to tell the vector store about the use of the metadata column "entity_id" like ``` vectorstore_entity = await PGVectorStore.create( engine=pg_engine_entity, table_name=TABLE_NAME_ENTITY, embedding_service=openai_embeddings, metadata_columns=["entity_id"] ) ``` Let me know if this does not fix your issue.

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: langchain-ai/langchain-postgres#74