How to Build a Paid AI Tool API with Ollama + Qwen and Earn Extra Income

Looking for a practical way to make side income from AI without building a full-fledged SaaS?

With Ollama + Qwen, you can turn a low-cost VPS into your own paid AI microservice API.

This guide walks you through every step — from deployment to payment setup — so you can start selling API access to clients or small businesses and generate recurring side income within hours.

1. Prerequisites

1. Choose a Server

• Recommended VPS: LightNode (cheap, hourly billing, great for testing an MVP)

• Minimum Specs:

• 2 vCPU / 4GB RAM / 30GB SSD

• Ubuntu 22.04 OS

• 1TB monthly bandwidth

• Cost: ~$7–10/month or $0.084/hour

Tip: Start small with a CPU VPS (3B model). Upgrade to a GPU VPS later as demand grows.

2. Connect to Your Server via SSH

ssh root@your_server_ip

3. Install Required Packages

sudo apt update && sudo apt upgrade -y
sudo apt install git curl python3-pip docker.io docker-compose -y
sudo systemctl enable docker

Install Ollama and Qwen Model

1. Install Ollama

curl -fsSL https://ollama.com/install.sh | sh

2. Pull the Qwen model (choose 3B or 7B version)

ollama pull qwen2.5:3b
# or
ollama pull qwen2.5:7b

3. Test the model

ollama run qwen2.5:3b

Try:

Hello, introduce yourself

You should receive a response.

3. Build a Paid API Service

We’ll use FastAPI to expose a paid HTTP API.

1. Install packages:

pip install fastapi uvicorn requests

2. Create a file app.py:

from fastapi import FastAPI, Request
import requests

app = FastAPI()
OLLAMA_URL = "http://localhost:11434/api/generate"
VALID_TOKENS = ["abc123", "vipuser"]

@app.post("/qwen")
async def run_qwen(request: Request):
    data = await request.json()
    prompt = data.get("prompt", "")
    token = data.get("token", "")

    if token not in VALID_TOKENS:
        return {"error": "Invalid or missing token. Please recharge."}

    payload = {
        "model": "qwen2.5:3b",
        "prompt": prompt,
        "stream": False
    }

    response = requests.post(OLLAMA_URL, json=payload)
    return response.json()

3. Start the service:

uvicorn app:app --host 0.0.0.0 --port 8000

Test endpoint:

POST http://<your_server_ip>:8000/qwen
{
  "token": "abc123",
  "prompt": "Write a short AI marketing copy."
}

4. Add a Credit System for Paid Access

Start by manually issuing tokens to paying users.
Later, automate with a basic SQLite-based credit system:

import sqlite3

conn = sqlite3.connect('users.db')
cur = conn.cursor()
cur.execute("CREATE TABLE IF NOT EXISTS users (token TEXT, credits INTEGER)")
cur.execute("INSERT INTO users VALUES ('abc123', 100)")
conn.commit()

@app.post("/qwen")
async def run_qwen(request: Request):
    data = await request.json()
    token = data.get("token", "")

    cur.execute("SELECT credits FROM users WHERE token=?", (token,))
    row = cur.fetchone()
    if not row or row[0] <= 0:
        return {"error": "No credits left. Please recharge."}

    cur.execute("UPDATE users SET credits=credits-1 WHERE token=?", (token,))
    conn.commit()

5. Run API in Background

Keep your service running even after closing SSH:

nohup uvicorn app:app --host 0.0.0.0 --port 8000 &

or:

pip install pm2
pm2 start "uvicorn app:app --host 0.0.0.0 --port 8000"

6. Provide API Documentation to Clients

Example usage guide:

Endpoint: http://<your_ip>:8000/qwen
Method: POST
Parameters:
{
  "token": "abc123",
  "prompt": "Write a 50-word SEO blog intro about AI startups."
}
Response:
{
  "response": "Here's your AI-generated text..."
}

Pricing suggestions:

• $5 = 1,000 calls

• $15/month = unlimited access

7. Scale and Optimize for More Revenue

✅ Add Nginx + SSL for secure HTTPS endpoints

✅ Offer multiple models (Qwen-Coder, Qwen-7B, etc.) with tiered pricing

✅ Create a simple front-end landing page for better conversions

✅ Upgrade to GPU VPS for faster responses and more clients

Side Hustle Potential

• Cost: ~$7–10/month LightNode VPS

• Revenue: 10 clients × $5 = $50/month (starter level)

• Scalability: Easily grow to $500–1000/month with better hosting and marketing