Automating Tasks with the Ollama API

It's just an API

Every guide so far has pointed you at Ollama through something else - Open WebUI's chat window, Continue in VS Code, OpenClaw's Discord bot. Underneath all of them is the same thing: Ollama runs a plain REST API on your machine, and anything that can send an HTTP request can talk to it directly.

That means a shell script, a cron job, or a five-line Python file can all use your local model exactly the same way a chat app does - no separate setup, no new install.

Calling it with curl

The simplest possible call, straight from a terminal:

curl http://localhost:11434/api/generate -d '{
  "model": "llama3.2",
  "prompt": "Summarize this in one sentence: local AI runs entirely on your own hardware.",
  "stream": false
}'

That returns a JSON object with the model's response in a response field, plus some metadata about timing. The "stream": false line matters - without it, Ollama sends the answer back as a series of partial JSON chunks as it's generated, which is great for a live-typing chat window and awkward for a script that just wants one clean answer.

Calling it from Python

Ollama publishes an official Python library that wraps the same API:

pip install ollama

import ollama

response = ollama.generate(model='llama3.2', prompt='Summarize this in one sentence: local AI runs entirely on your own hardware.')
print(response['response'])

This is the same request the curl example made, just without hand-building JSON. If you'd rather not add a dependency, plain requests works identically since it's still just HTTP under the hood:

import requests

r = requests.post('http://localhost:11434/api/generate', json={
    'model': 'llama3.2',
    'prompt': 'Summarize this in one sentence: local AI runs entirely on your own hardware.',
    'stream': False
})
print(r.json()['response'])

A real example

A genuinely useful small automation: a script that reads a text file and writes a one-sentence summary of it, runnable on a schedule or whenever you drop a new file somewhere.

import ollama

with open('notes.txt') as f:
    text = f.read()

response = ollama.generate(
    model='llama3.2',
    prompt=f'Summarize this in one sentence:\n\n{text}'
)
print(response['response'])

What actually goes wrong

The response is a wall of separate JSON objects, not one clean answer

This means "stream": false got left out (or stream=False in Python). Without it, Ollama streams one JSON object per generated chunk - fine for a live UI, unusable for a script expecting a single response.

Connection refused

Ollama isn't running. Confirm with:

curl http://localhost:11434/api/tags

If that also fails, start Ollama before troubleshooting the script itself.

The script is slow on the first call, then fine after

Ollama loads a model into memory on first use and keeps it loaded for a few minutes afterward. The first request in a while pays that load time; requests shortly after don't. This is normal behavior, not a bug in your script.

Common questions