rem-vectordb 0.2.0

Official Python SDK for the REM distributed vector database

REM Vector Database - Python SDK

Official Python SDK for REM Network — the decentralized vector database for AI applications. A Pinecone-compatible API powered by 2,000+ distributed miners on the Sui blockchain.

Installation

pip install rem-vectordb

With LangChain support:

pip install rem-vectordb[langchain]

With LlamaIndex support:

pip install rem-vectordb[llamaindex]

Quick Start

from rem import REM

client = REM(api_key="rem_your_api_key")

# Create a collection
collection = client.create_collection("my-docs", dimension=1536)

# Upsert vectors
collection.upsert([
    {"id": "doc1", "values": [0.1, 0.2, ...], "metadata": {"title": "Hello"}},
    {"id": "doc2", "values": [0.3, 0.4, ...], "metadata": {"title": "World"}},
])

# Semantic search
results = collection.query(vector=[0.1, 0.2, ...], top_k=10)
for match in results.matches:
    print(f"{match.id}: {match.score:.4f}")

Features

Encrypted Metadata (AES-256-GCM)

Protect sensitive metadata fields with per-namespace encryption. Miners never see your plaintext data.

collection = client.create_collection(
    name="secure-docs",
    dimension=1536,
    encrypted_fields=["text", "user_id", "email"]
)

Hybrid Search (Vector + Keyword)

Combine semantic vector search with BM25 keyword matching via Reciprocal Rank Fusion.

results = collection.query(
    vector=[0.1, 0.2, ...],
    query_text="machine learning",
    hybrid_alpha=0.7,  # 0.0=pure keyword, 1.0=pure vector
    top_k=10,
)

Metadata Filtering

Pinecone-compatible filter operators: $eq, $gt, $gte, $lt, $lte, $in, $nin, $and, $or.

results = collection.query(
    vector=[0.1, 0.2, ...],
    top_k=10,
    filter={
        "$and": [
            {"category": {"$eq": "science"}},
            {"year": {"$gte": 2020}}
        ]
    },
)

Batch Queries

Execute up to 10 queries in a single API call for recommendation systems and AI agents.

results = collection.query_batch([
    {"vector": [0.1, ...], "top_k": 5},
    {"vector": [0.3, ...], "top_k": 5, "filter": {"type": "article"}},
    {"query_text": "neural networks", "top_k": 3},
])

Fetch & Delete

Retrieve or remove vectors by ID.

# Fetch vectors
fetched = collection.fetch(ids=["doc1", "doc2"])
for v in fetched.vectors:
    print(f"{v.id}: {v.metadata}")

# Delete vectors
result = collection.delete(ids=["doc1"])
print(f"Deleted {result.deleted_count} vectors")

Async Support

Full async/await interface for high-throughput applications.

from rem import AsyncREM

async with AsyncREM(api_key="rem_xxx") as client:
    collection = await client.create_collection("my-docs", dimension=1536)
    await collection.upsert([...])
    results = await collection.query(vector=[...], top_k=10)

Framework Integrations

LangChain

from langchain_openai import OpenAIEmbeddings
from rem.integrations.langchain import REMVectorStore

store = REMVectorStore(
    api_key="rem_xxx",
    collection_name="docs",
    embedding=OpenAIEmbeddings(),
)

# Add documents
store.add_texts(["Hello world", "REM is great"], metadatas=[{"source": "test"}])

# Similarity search
results = store.similarity_search("greeting", k=5)

# Use in RAG chains
from langchain.chains import RetrievalQA
qa = RetrievalQA.from_chain_type(llm=llm, retriever=store.as_retriever())

LlamaIndex

from llama_index.core import VectorStoreIndex
from rem.integrations.llamaindex import REMVectorStore

vector_store = REMVectorStore(api_key="rem_xxx", collection_name="docs")
index = VectorStoreIndex.from_vector_store(vector_store)
query_engine = index.as_query_engine()
response = query_engine.query("What is REM Network?")

API Reference

Client

client = REM(
    api_key="rem_xxx",                           # Required (starts with rem_)
    base_url="https://api.getrem.online",         # Default
    timeout=30.0,                                  # Seconds
)

Collections

Method	Description
client.create_collection(name, dimension, metric, encrypted_fields)	Create collection
client.get_collection(id)	Get by ID
client.list_collections()	List all
client.delete_collection(id)	Delete

Vectors

Method	Description
collection.upsert(vectors)	Insert/update vectors
collection.query(vector, top_k, filter, query_text, hybrid_alpha)	Search
collection.query_batch(queries)	Batch search (up to 10)
collection.fetch(ids)	Fetch by ID
collection.delete(ids)	Delete by ID
collection.stats()	Collection stats

Distance Metrics

• cosine (default) — Normalized similarity
• euclidean — L2 distance
• dot_product — Inner product

Getting Started

1. Sign up at app.getrem.online to get your API key
2. You get $20 in free API credits
3. pip install rem-vectordb
4. Start building!

Full documentation: app.getrem.online/docs

Links

• Homepage
• Platform & Docs
• Network Explorer
• Discord
• Twitter/X
• Telegram

License

MIT - See LICENSE for details.

Built by BeClever OÜ (Estonia).