Skip to main content

​
Overview

The SolVec class is the main entry point for the VecLabs Python SDK. It manages connections to the Solana network and provides access to vector collections.

​
Constructor

from solvec import SolVec

sv = SolVec(
    network="devnet",
    wallet="~/.config/solana/id.json",
    rpc_url=None
)

​
Parameters

​
network
str
default:"devnet"
Solana network to connect to. Options:
  • "mainnet-beta" - Production Solana mainnet
  • "devnet" - Solana devnet for testing
  • "localnet" - Local Solana validator
​
wallet
str | None
default:"None"
Path to your Solana keypair JSON file. Required for write operations (upsert, delete).Example: "~/.config/solana/id.json"
​
rpc_url
str | None
default:"None"
Custom RPC URL. If not provided, uses default RPC for the selected network.Use this for premium RPC providers like Helius or QuickNode for better performance.

​
Raises

  • ValueError - If an invalid network name is provided

​
Methods

​
collection()

Get or create a vector collection. Equivalent to Pinecone’s Index(). 
col = sv.collection(
    name="agent-memory",
    dimensions=1536,
    metric="cosine"
)

​
Parameters

​
name
str
required
Collection name (maximum 64 characters). Used as a unique identifier for the collection.
​
dimensions
int
default:"1536"
Vector dimension. Default is 1536, which matches OpenAI’s text-embedding-3-small and text-embedding-ada-002 models.Common dimensions:
  • 1536 - OpenAI embeddings
  • 768 - Sentence transformers
  • 384 - MiniLM models
​
metric
str | DistanceMetric
default:"DistanceMetric.COSINE"
Distance metric for similarity search. Options:
  • "cosine" or DistanceMetric.COSINE - Cosine similarity (recommended for most use cases)
  • "euclidean" or DistanceMetric.EUCLIDEAN - Euclidean distance
  • "dot" or DistanceMetric.DOT - Dot product

​
Returns

​
collection
SolVecCollection
A SolVecCollection instance for performing vector operations.

​
Example

from solvec import SolVec

sv = SolVec(network="devnet", wallet="~/.config/solana/id.json")

# Create collection for OpenAI embeddings
col = sv.collection("agent-memory", dimensions=1536)

# Create collection with Euclidean distance
col = sv.collection(
    "image-vectors",
    dimensions=768,
    metric="euclidean"
)

​
list_collections()

List all collection names in the current session. 
collections = sv.list_collections()

​
Returns

​
collections
list[str]
List of collection names that have been accessed in this session.

​
Example

sv = SolVec(network="devnet")

col1 = sv.collection("memories")
col2 = sv.collection("facts")

print(sv.list_collections())  # Output: ['memories', 'facts']

​
Class Attributes

​
RPC_URLS

Default RPC URLs for each Solana network: 
RPC_URLS = {
    "mainnet-beta": "https://api.mainnet-beta.solana.com",
    "devnet": "https://api.devnet.solana.com",
    "localnet": "http://localhost:8899",
}

​
Complete Example

from solvec import SolVec
import openai

# Initialize SolVec client
sv = SolVec(
    network="devnet",
    wallet="~/.config/solana/id.json"
)

# Get or create a collection
col = sv.collection("agent-memory", dimensions=1536)

# Generate embeddings with OpenAI
text = "User prefers dark mode"
response = openai.embeddings.create(
    model="text-embedding-3-small",
    input=text
)
embedding = response.data[0].embedding

# Store vector
col.upsert([{
    "id": "mem_001",
    "values": embedding,
    "metadata": {"text": text, "category": "preference"}
}])

# Query for similar memories
results = col.query(vector=embedding, top_k=5)
for match in results.matches:
    print(f"{match.id}: {match.metadata['text']} (score: {match.score})")

​
Notes

  • Collections are cached in memory per SolVec instance
  • Calling collection() multiple times with the same name returns the same instance
  • The SDK is designed to be Pinecone-compatible for easy migration

Build docs developers (and LLMs) love

Get started for freeTalk to us