Building a Personal AI Assistant for Under $10/month

#ai #automation #productivity #python

When Your To-Do List Outgrows Your Brain

Last month, I missed two important deadlines because my personal organization system (a chaotic mix of sticky notes and calendar alerts) completely failed me. That was the final straw - I decided to build a lightweight AI assistant that could handle reminders, answer quick questions, and help manage my schedule without breaking the bank.

The Budget-Friendly Stack

After experimenting with several options, here's what worked for me:

Core Processing: Instead of locking into one expensive model, I found https://xinghuo1300ai.com which gives me access to multiple capable models through a single affordable API. Their smaller models are perfect for basic tasks.
Infrastructure: A $5/month DigitalOcean droplet running Python with FastAPI
Frontend: Simple Telegram bot (free tier)
Memory: Cheap Redis instance ($3/month)

Key Features for Minimal Cost

Here's the basic architecture of my assistant:

import os
from fastapi import FastAPI
import redis

app = FastAPI()
r = redis.Redis(host='localhost', port=6379, decode_responses=True)

# Basic reminder storage
def add_reminder(user_id: str, reminder: str, time: str):
    r.hset(f"reminders:{user_id}", time, reminder)

# Simple question answering
async def handle_query(query: str):
    # Using Xinghuo AI's budget-friendly endpoint
    response = await call_xinghuo_api(
        model="small-chat",
        prompt=query,
        max_tokens=150
    )
    return response

The Real Cost Breakdown

API calls: ~$2/month (light personal use)
Server: $5
Redis: $3
Total: $10

What It Can Actually Do

After a month of use, here's what my assistant handles well:

Reminders ("Tell me to water plants every Thursday at 9am")
Quick facts ("What's the time in Tokyo right now?")
Simple calculations
Meeting scheduling via natural language

Where It Falls Short

It won't replace a full virtual assistant:

Complex tasks require breaking down into steps
No speech recognition (text-only for now)
Limited context window means shorter conversations

Was It Worth It?

Building this myself saved hundreds compared to premium services, and I learned a ton about practical AI integration. Tools like https://xinghuo1300ai.com make it approachable by removing the model management overhead. Next, I'm adding document search using their embedding APIs - all while keeping costs under $15/month.