Ollama

The Ollama plugin enables local AI capabilities using Ollama, with no cloud API required.

Installation

pnpm add @teamflojo/floimg-ollama

Prerequisites

Install and run Ollama locally:

# Install Ollama (macOS)
brew install ollama

# Pull required models
ollama pull llava      # For vision
ollama pull llama3.2   # For text

# Start the Ollama server
ollama serve

Registration

import createClient from '@teamflojo/floimg';
import ollama from '@teamflojo/floimg-ollama';

const floimg = createClient();

floimg.registerGenerator(ollama({
  baseUrl: 'http://localhost:11434'  // Default Ollama URL
}));

Vision Analysis (LLaVA)

Analyze images using LLaVA locally.

const analysis = await floimg.analyzeImage({
  blob: image,
  prompt: 'Describe this image in detail'
});

console.log(analysis.text);
// "The image shows a cozy living room with..."

Vision Parameters

Parameter	Type	Required	Description
`model`	string	No	Vision model (default: `llava`)
`prompt`	string	Yes	Question or instruction about the image

Text Generation (Llama)

Generate text using Llama locally.

const result = await floimg.generateText({
  prompt: 'Write a caption for a beach sunset photo',
  model: 'llama3.2'
});

console.log(result.text);
// "Golden hour magic at the beach..."

Text Parameters

Parameter	Type	Required	Description
`model`	string	No	Text model (default: `llama3.2`)
`prompt`	string	Yes	The prompt for text generation

Available Models

Model	Type	Size	Use Case
`llava`	Vision	7B	Image analysis, description
`llava:13b`	Vision	13B	Higher quality analysis
`llama3.2`	Text	3B	Fast text generation
`llama3.2:7b`	Text	7B	Better quality text

Pull additional models with:

ollama pull llava:13b
ollama pull llama3.2:7b

Example Pipeline

import createClient from '@teamflojo/floimg';
import ollama from '@teamflojo/floimg-ollama';
import qr from '@teamflojo/floimg-qr';

const floimg = createClient();
floimg.registerGenerator(ollama());
floimg.registerGenerator(qr());

// 1. Analyze an image locally
const analysis = await floimg.analyzeImage({
  blob: productImage,
  prompt: 'Describe this product for an e-commerce listing'
});

// 2. Generate a caption
const caption = await floimg.generateText({
  prompt: `Write a short marketing tagline based on: ${analysis.text}`
});

// 3. Create a QR code linking to the product
const qrCode = await floimg.generate({
  generator: 'qr',
  params: { data: 'https://example.com/product/123' }
});

console.log(caption.text);

Benefits

Privacy: All processing happens locally
No API Costs: No per-request charges
Offline: Works without internet connection
Fast: No network latency for local inference