Skip to main content

​
What is AI Providers?

AI Providers is a configuration hub for managing AI settings in Obsidian. Think of it as a central control panel where you can store your API keys and AI provider settings once, then share them across all your Obsidian plugins that need AI capabilities.
Important: AI Providers is a configuration tool—it doesn’t do any AI processing itself. It helps other plugins connect to AI services more easily by managing settings in one place.

​
Why AI Providers?

Without AI Providers, each AI-powered plugin would need its own settings page for API keys, endpoints, and model selection. This creates several problems:
  • Duplicate configuration: Enter the same API keys multiple times
  • Harder maintenance: Update credentials separately in each plugin
  • Inconsistent experience: Different UIs and setup flows
  • Development overhead: Plugin developers must build provider integrations from scratch

For Users

Configure AI providers once, use them everywhere. Switch between OpenAI, Ollama, Claude, and 15+ other providers without reconfiguring each plugin.

For Developers

Skip building provider integrations. Use the AI Providers SDK to instantly support all providers with just a few lines of code.

​
Key Benefits

​
Centralized Configuration

Manage all your AI providers in one settings panel:
  • Store API keys securely
  • Configure provider URLs and endpoints
  • Select models from refreshable lists
  • Enable/disable providers as needed

​
Wide Provider Support

AI Providers supports 18+ AI providers out of the box:

Cloud Providers

  • OpenAI
  • Anthropic
  • Google Gemini
  • OpenRouter
  • Mistral AI
  • Groq
  • Perplexity AI
  • DeepSeek
  • xAI (Grok)

Local & Self-Hosted

  • Ollama
  • LM Studio
  • Open WebUI
  • llama.cpp
  • LocalAI

Additional Services

  • Together AI
  • Fireworks AI
  • Cerebras
  • DeepInfra
  • SambaNova
  • And more…

​
Developer-Friendly SDK

The AI Providers SDK makes it simple to add AI capabilities to your plugin: 
import { waitForAI } from '@obsidian-ai-providers/sdk';

// Wait for AI Providers to load
const aiResolver = await waitForAI();
const aiProviders = await aiResolver.promise;

// Execute a prompt with streaming
const response = await aiProviders.execute({
    provider: aiProviders.providers[0],
    prompt: "Summarize this note",
    onProgress: (chunk, fullText) => {
        console.log(fullText); // Update UI as text streams in
    }
});
The SDK handles all provider-specific API differences, streaming, error handling, and even caching for embeddings.

​
Architecture Overview

AI Providers follows a simple, extensible architecture: 
┌─────────────────────────────────────────┐
│  Your Plugin (uses SDK)                 │
├─────────────────────────────────────────┤
│  @obsidian-ai-providers/sdk             │
│  - waitForAI()                          │
│  - execute(), embed(), retrieve()       │
├─────────────────────────────────────────┤
│  AI Providers Plugin                    │
│  - AIProvidersService                   │
│  - Provider Handlers                    │
│  - Settings Management                  │
│  - Embeddings Cache (IndexedDB)         │
├─────────────────────────────────────────┤
│  External AI APIs                       │
│  - OpenAI, Anthropic, Ollama, etc.      │
└─────────────────────────────────────────┘

​
How It Works

  1. Settings Layer: Users configure providers in the plugin settings UI
  2. Service Layer: AIProvidersService manages provider instances and exposes them via the SDK
  3. Handler Layer: Provider-specific handlers (OpenAI, Anthropic, Ollama) translate requests to the right API format
  4. Transport Layer: FetchSelector chooses the appropriate fetch method (Obsidian, Electron, or native)
  5. Cache Layer: Embeddings are cached in IndexedDB to avoid redundant API calls
AI Providers requires Obsidian 0.15.0 or later. It works on desktop and mobile.

​
Supported Features

  • Text Generation: Stream completions from any provider with execute()
  • Embeddings: Generate vector embeddings with embed() (cached automatically)
  • RAG Search: Semantic search with retrieve() for retrieval-augmented generation
  • Messages API: Support for multi-turn conversations and system prompts
  • Image Analysis: Send images to vision models like GPT-4V
  • Abort Control: Cancel in-progress requests with AbortController
  • Progress Tracking: Monitor embedding and retrieval progress
  • Model Discovery: Fetch available models dynamically from providers

​
Multilingual Support

AI Providers is translated into 11 languages: English, Spanish, French, Italian, Portuguese, German, Russian, Chinese, Japanese, Korean, and Dutch.

​
Next Steps

Installation

Install AI Providers from the Obsidian community plugin store

Quick Start

Set up your first AI provider in under 5 minutes

​
Plugins Using AI Providers

AI Providers is used by plugins like:
  • Local GPT: Privacy-focused AI assistant with local models
  • More plugins integrating soon…
If you’re a plugin developer, check out the SDK documentation to integrate AI Providers into your plugin.

Build docs developers (and LLMs) love

Get started for freeTalk to us