Gel

Gel is a powerful data platform built on top of PostgreSQL.

• Think in objects and graphs instead of tables and JOINs.
• Use the advanced Python SDK, integrated GUI, migrations engine, Auth and AI layers, and much more.
• Run locally, remotely, or in a fully managed cloud.

Installation

pip install langchain-gel

Setup

1. Run gel project init
2. Edit the schema. You need the following types to use the LangChain vectorstore:

using extension pgvector;
                                    
module default {
    scalar type EmbeddingVector extending ext::pgvector::vector<1536>;

    type Record {
        required collection: str;
        text: str;
        embedding: EmbeddingVector; 
        external_id: str {
            constraint exclusive;
        };
        metadata: json;

        index ext::pgvector::hnsw_cosine(m := 16, ef_construction := 128)
            on (.embedding)
    } 
}

Note: this is the minimal setup. Feel free to add as many types, properties and links as you want! Learn more about taking advantage of Gel's schema by reading the docs.

3. Run the migration: gel migration create && gel migrate.

Usage

from langchain_gel import GelVectorStore

vector_store = GelVectorStore(
    embeddings=embeddings,
) 

See the full usage example here.