@teamflojo/floimg-ollama

Local AI provider for floimg using Ollama. Run vision and text models locally with no API key required.

Standing on the Shoulders of Giants

This plugin integrates with Ollama, the amazing local AI runtime. We provide a consistent FloImg interface while exposing the full power of Ollama's models.

• Full Ollama power: All models and configuration options work
• Native format: Use Ollama model names and options directly
• Their docs are your docs: See Ollama documentation

FloImg orchestrates the workflow (generate → transform → save). Ollama does what it does best.

Features

• Vision Analysis - Analyze images with LLaVA or other vision models
• Text Generation - Generate text with Llama, Mistral, or other models
• 100% Local - No API keys, no data leaves your machine
• Privacy-First - Perfect for sensitive content

Prerequisites

1. Install Ollama
2. Pull the models you want to use:

# Vision model
ollama pull llava

# Text model
ollama pull llama3.2

Installation

npm install @teamflojo/floimg-ollama

Usage

Vision Analysis

import createClient from "@teamflojo/floimg";
import { ollamaVision } from "@teamflojo/floimg-ollama";

const client = createClient();
client.registerVisionProvider(ollamaVision({ model: "llava" }));

const result = await client.analyzeImage({
  provider: "ollama-vision",
  blob: imageBlob,
  params: {
    prompt: "What objects are in this image?",
    outputFormat: "json",
  },
});

console.log(result.content);

Text Generation

import createClient from "@teamflojo/floimg";
import { ollamaText } from "@teamflojo/floimg-ollama";

const client = createClient();
client.registerTextProvider(ollamaText({ model: "llama3.2" }));

const result = await client.generateText({
  provider: "ollama-text",
  params: {
    prompt: "Write a creative caption for a sunset photo",
    temperature: 0.8,
  },
});

console.log(result.content);

Register Both Providers

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

const client = createClient();
const [vision, text] = ollama({
  baseUrl: "http://localhost:11434",
  visionModel: "llava",
  textModel: "llama3.2",
});

client.registerVisionProvider(vision);
client.registerTextProvider(text);

In Pipelines

await client.run({
  name: "local-image-analysis",
  steps: [
    { kind: "generate", generator: "qr", params: { text: "hello" }, out: "qr" },
    {
      kind: "vision",
      in: "qr",
      provider: "ollama-vision",
      params: { prompt: "What does this QR code contain?" },
      out: "analysis",
    },
    {
      kind: "text",
      in: "analysis",
      provider: "ollama-text",
      params: { prompt: "Summarize this in one sentence" },
      out: "summary",
    },
  ],
});

Configuration

Vision Provider

| Option | Default | Description | | --------- | ------------------------ | ------------------- | | baseUrl | http://localhost:11434 | Ollama server URL | | model | llava | Vision model to use |

Text Provider

| Option | Default | Description | | --------- | ------------------------ | ----------------- | | baseUrl | http://localhost:11434 | Ollama server URL | | model | llama3.2 | Text model to use |

Supported Models

Vision Models

• llava - LLaVA (default)
• llava:13b - LLaVA 13B
• bakllava - BakLLaVA

Text Models

• llama3.2 - Llama 3.2 (default)
• llama3.1 - Llama 3.1
• mistral - Mistral 7B
• mixtral - Mixtral 8x7B
• codellama - Code Llama

License

MIT