zerithdb-python 0.1.0

Python SDK for ZerithDB - Headless AI agent node for P2P mesh network.

ZerithDB Python SDK (AI Agent Node)

The Python SDK for ZerithDB allows headless Python scripts to join the WebRTC mesh network just like a normal user!

Since ZerithDB syncs via a P2P mesh, you don't need a REST API to interact with the database. You can build a zerithdb-python package that runs heavy machine learning workloads (e.g., PyTorch, Transformers) on a GPU server, syncing data seamlessly.

How it works

1. A browser user inserts a row into the database:

   app.db("jobs").insert({ text: "Translate this", status: "pending" });
   

2. The Python script (running on a server) connects to the same WebRTC room and syncs this data via P2P.

3. The script processes the text and updates the row:

   async def on_job_added(job):
       result = my_ml_model.process(job["text"])
       await db.table("jobs").update(job["id"], { "status": "done", "result": result })
   

Why it's cool

Developers can add heavy ML processing to their apps without building APIs, webhooks, or queues. The Python script is just another "peer" in the room!

Getting Started

pip install zerithdb

Basic Example

import asyncio
from zerithdb import ZerithClient

async def main():
    # Initialize the client and connect to a room
    client = ZerithClient("wss://signal.zerithdb.dev")
    await client.connect("my-app-room-id")

    # Listen for new data
    @client.on("jobs:inserted")
    async def handle_job(job):
        if job["status"] == "pending":
            print(f"Processing job: {job['text']}")
            # Run your ML model here...
            await client.update("jobs", job["id"], {"status": "done", "result": "Translated text!"})

    # Keep the agent running
    await client.wait_until_disconnected()

if __name__ == "__main__":
    asyncio.run(main())

Architecture

• Uses aiortc for WebRTC connections and data channels.
• Uses websockets for the initial signaling phase.
• Communicates using the standard ZerithDB CRDT sync protocol.