proximadb

Python client SDK for ProximaDB — a cloud-native vector + graph database engineered for AI-first applications (vector search, OLTP/OLAP, graph, document, multimodal) over REST, gRPC, Arrow Flight, and the PostgreSQL wire protocol.

• Distribution: proximadb (PyPI) · import: proximadb_sdk
• In-process engine: proximadb_embedded (native PyO3 wheels) — pip install 'proximadb[embedded]'

Install

pip install proximadb                 # client SDK (REST/gRPC/Flight)
pip install 'proximadb[embedded]'     # + in-process native engine (no server)
pip install 'proximadb[embeddings]'   # + local embedding providers
pip install 'proximadb[codegraph]'    # + shared code->CPG chunker (victor-codegraph)

Extras compose, e.g. pip install 'proximadb[embeddings,langchain,llama_index]'.

Quickstart

from proximadb_sdk import ProximaDBClient

client = ProximaDBClient(url="http://localhost:5678")
client.create_collection("docs", dimension=384)
client.insert("docs", vectors=[...], ids=[...])
hits = client.search("docs", query=[...], top_k=5)

Embedded (in-process, no server):

from proximadb_embedded import Client  # requires the [embedded] extra

db = Client(data_dir="./data")

Transports

The SDK speaks REST (ergonomics), gRPC (typed RPC), and Arrow Flight (zero-copy bulk) on ProximaDB's multiplexed port, plus standard PostgreSQL drivers via pgwire. The REST surface is generated from the canonical OpenAPI spec behind an ergonomic facade.

License

Apache-2.0.