Python 3.13.0rc1 : Testing python with Ollama local install.

I was very busy with development and testing for about two weeks and my laptop was stuck and I was working hard... Today I managed to test local background clipping on my laptop with a local Ollama installation separated by a Python module but with processing from the Python script. I also used Microsoft's Copilot artificial intelligence for python and it works well even though it is not theoretically specialized in development. The source code is quite large but the result is very good and fast:

import subprocess
import os
import json
from PIL import Image, ImageOps

class OllamaProcessor:
    def __init__(self, config_file):
        self.config_file = config_file
        self.model_methods = self.load_config()

    def load_config(self):
        try:
            with open(self.config_file, 'r') as file:
                config = json.load(file)
            print("Configuration loaded successfully.")
            return config
        except FileNotFoundError:
            print(f"Configuration file {self.config_file} not found.")
            raise
        except json.JSONDecodeError:
            print(f"Error decoding JSON from the configuration file {self.config_file}.")
            raise

    def check_ollama(self):
        try:
            result = subprocess.run(["ollama", "--version"], capture_output=True, text=True, check=True)
            print("Ollama is installed. Version:", result.stdout)
        except subprocess.CalledProcessError as e:
            print("Ollama is not installed or not found in PATH. Ensure it's installed and accessible.")
            raise
... 

Here is the result obtained after finishing running in the command line:

python ollama_test_001.py
Configuration file ollama_config.json created successfully.
Configuration loaded successfully.
Ollama is installed. Version: ollama version is 0.5.7

Available models: ['NAME']
pulling manifest
pulling 170370233dd5... 100% ▕██████████████▏ 4.1 GB
pulling 72d6f08a42f6... 100% ▕██████████████▏ 624 MB
pulling 43070e2d4e53... 100% ▕██████████████▏  11 KB
pulling c43332387573... 100% ▕██████████████▏   67 B
pulling ed11eda7790d... 100% ▕██████████████▏   30 B
pulling 7c658f9561e5... 100% ▕██████████████▏  564 B
verifying sha256 digest
writing manifest
success
Model llava pulled successfully for method process_images_in_folder.

Some "Command failed ..." but the result is cutting well and it has transparency !

Python 3.11.11 : Colab simple test with mistralai - part 048.

Two notebook on my colab repo project

One with xterm colab festure and another with hull-convex known as lattrice with OllamaFunctions.

Python 3.12.1 : testing with ollama - part 001.

You need to download the ollama from the official website.

This allow you to use commands on the console: ollama --help into new command shell.

Use the pip tool to install ollma python package:

pip install ollama
Collecting ollama
  Downloading ollama-0.2.1-py3-none-any.whl.metadata (4.2 kB)
...
Installing collected packages: httpx, ollama
  Attempting uninstall: httpx
    Found existing installation: httpx 0.26.0
    Uninstalling httpx-0.26.0:
      Successfully uninstalled httpx-0.26.0
Successfully installed httpx-0.27.0 ollama-0.2.1

This command will install llama3

ollama run llama3

Let's see a basic python script with mistral model:

import ollama
from rich.console import Console
console = Console()
print(ollama.pull("mistral"))
#with console.pager(styles=True):
#	console.print(ollama.show("mistral"))
with console.pager(styles=True):
	console.print(ollama.list())

Another python script with llama3 model:

import ollama

stream = ollama.chat(
    model='llama3',
    messages=[{'role': 'user', 'content': 'Tell me the sizes of Earth?'}],
    stream=True,
)

for chunk in stream:
  print(chunk['message']['content'], end='', flush=True)

The result is this:

What a great question!

The size of Earth can be measured in various ways, depending on what aspect you're interested in. Here are some common sizes and dimensions of our planet ...