How to store text embeddings in custom table in PGVector?

[vinayak-shanawad]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

COLLECTION_NAME = "state_of_the_union_test"

db = PGVector.from_documents(
    embedding=embeddings,
    documents=docs,
    collection_name=COLLECTION_NAME,
    connection_string=CONNECTION_STRING,
)

Description

We're trying to store the text vector embeddings in PGVector and noticed that PGVector has created two tables.

1. langchain_pg_collection: Store the collection details
2. langchain_pg_embedding: Store the embedding details.

As per this PGVector class, I see these tables are hard coded.

Is there any way to store the embeddings in custom tables?

Thanks in advance.

System Info

Platform: Windows
Python version: 3.11
langchain: 0.1.4
pgvector: 0.2.4

[dosubot[bot]]

🤖

Hey there, @vinayak-shanawad! Good to see you back with another intriguing question. Hope you're doing well!

Based on the current implementation of the LangChain framework, there is no built-in way to store text vector embeddings in custom tables with PGVector. The table names 'langchain_pg_collection' and 'langchain_pg_embedding' are hardcoded in the CollectionStore and EmbeddingStore classes respectively, as shown below:

class CollectionStore(BaseModel):
    """Collection store."""
    __tablename__ = "langchain_pg_collection"
    ...

class EmbeddingStore(BaseModel):
    """Embedding store."""
    __tablename__ = "langchain_pg_embedding"
    ...

These classes are used throughout the PGEmbedding class, which handles the storage and retrieval of embeddings. Therefore, without modifying the source code of the LangChain framework, it is not possible to use custom table names.

If you need to use custom table names, you might need to create a custom implementation of the PGEmbedding class, or modify the LangChain source code to allow for custom table names. However, please note that modifying the source code might lead to unexpected behavior and is generally not recommended.

You can find more details about the PGEmbedding class in the LangChain repository.

I hope this answers your question. If you have any further queries, feel free to ask.

Sources

• libs/community/langchain_community/vectorstores/pgembedding.py
• libs/langchain/langchain/vectorstores/pgvector.py

---

This response is meant to be useful and save you time. It is not meant to be a precise solution, but rather a starting point for your own research.

Help me be more useful! Please leave a 👍 if this is helpful and 👎 if it is irrelevant.

If you want to continue the conversation, start your reply with @dosu-bot.

[ucekmez]

hey @dosu-bot! how can we define custom implementation of PGEmbedding with minimal code change and integrate it into existing PGVector class to make custom table names possible for different scenarios, without getting impacted by future langchain updates

[bestickley]

I would like to create these tables in a custom schema (my_schema.langchain_pg_collection) by passing in a configuration object. Would be great if this was a feature.