Inserting data with vectors

To insert vectors along with your data, you can use the insert_with_vector method. This method allows you to insert data along with the vector embeddings. The vector embeddings can be generated using any method you choose, but currently only supporting 1536 dimensions (the same dimensions as OpenAI embedding models for simplicity).

PYTHON

from nexus_python.nexusdb import NexusDB
from openai import OpenAI

# Initialize NexusDB
nexus_db = NexusDB(api_key="your_nexusdb_api_key")

# Initialize OpenAI
client = OpenAI(api_key="your_openai_api_key")

# Example text to insert
text_to_insert = "nexusdb is a great database."

# First, get the vector for the given text
vector = get_embedding(text_to_insert)

insert_response = nexus_db.insert_with_vector(
    relation_name="example_relation",
    text=text_to_insert,
    vectors=vector,
)
print("Insert Response:", insert_response)

SHELL

NEXUSDB_API_KEY = "YOUR_API_KEY"
vectors = list_of_vector_embeddings

curl -X POST https://api.nexusdb.io/query \
-H "Content-Type: application/json" \
-H "API-Key: $NEXUSDB_API_KEY" \
-d '{
"query_type": "Insert",
"relation_name": "example_relation",
"searchable_content": {
        "text": text,
        "vectors": $vectors,
    }
}'

This code will insert the text “nexusdb is a great database.” along with the vector embeddings into the relation “example_relation”. If using the shell script, you might have noticed that this is also the insert method, but with an additional vectors parameter. This is because the insert_with_vector method is a wrapper around the insert method, with the vectors parameter added. You may optionally include the vectors parameter when inserting fields and values, but it is not required. Note that doing this will not automatically link the vectors and other data together, but will link the text and vector embeddings to the specified relation.

Information about referencing other data, metadata filtering, and access keys will be discussed in the “advanced” section, coming soon. For now, you can reference the source code, here.