AI Providers

Ariv supports four AI providers, each with different strengths. This guide helps you understand the trade-offs so you can pick the right one for your workflow. You can switch providers at any time — just update the settings and your existing tags, searches, and Ask Brain history continue to work.

Google Gemini

Recommended default for most users.

Models

Model	Speed	Quality	Notes
Gemini 2.0 Flash	Fast	Good	Default. Best balance of speed and cost
Gemini 2.5 Flash	Moderate	Better	Improved reasoning over 2.0 Flash
Gemini 2.5 Pro	Slower	Best	Most capable Gemini model

Setup

1. Go to aistudio.google.com.
2. Sign in with your Google account.
3. Create an API key.
4. In Ariv: set ai.provider to Google Gemini and paste your key into ai.gemini.apiKey.

Why Choose Gemini

• Free tier available. Google offers a generous free tier for the Gemini API, making it the easiest way to try Ariv’s AI features without committing to a paid account.
• Fast responses. Gemini 2.0 Flash is one of the quickest models available, which matters for a snappy auto-tagging and Ask Brain experience.
• Good all-rounder. Handles tagging, summarization, and Q&A well across a range of note types.

Tip

Start with Gemini 2.0 Flash. If you find the quality lacking on complex Ask Brain queries, try upgrading to Gemini 2.5 Flash or Gemini 2.5 Pro for those sessions.

---

OpenAI

Models

Model	Speed	Quality	Notes
GPT-4o	Moderate	Best	Most capable, handles complex questions
GPT-4o Mini	Fast	Good	Lower cost, good for routine tasks
o3-mini	Slower	Best	Strong reasoning, good for analytical queries

Setup

1. Go to platform.openai.com.
2. Sign in and navigate to the API keys section.
3. Create a new secret key.
4. In Ariv: set ai.provider to OpenAI and paste your key into ai.openai.apiKey.

Why Choose OpenAI

• Mature ecosystem. If you already have an OpenAI API account with billing set up, connecting it to Ariv takes seconds.
• Strong at structured tasks. GPT-4o is particularly good at extracting structured information from messy notes — useful for auto-tagging and action item detection.
• Reasoning models. o3-mini excels at analytical questions where you need the AI to think through a problem rather than just retrieve information.

Note

OpenAI requires a paid API account. There is no free tier for API usage. Check platform.openai.com/pricing for current rates.

---

Anthropic (Claude)

Models

Model	Speed	Quality	Notes
Claude Sonnet 4.5	Moderate	Best	Balanced quality and speed
Claude Haiku 4.5	Fast	Good	Quick responses, lower cost

Setup

1. Go to console.anthropic.com.
2. Sign in and create an API key from the dashboard.
3. In Ariv: set ai.provider to Anthropic and paste your key into ai.anthropic.apiKey.

Why Choose Anthropic

• Excellent writing quality. Claude tends to produce well-structured, nuanced responses — especially noticeable in Ask Brain answers and note summarization.
• Good at following instructions. Claude is strong at adhering to the specific formatting and tagging conventions Ariv uses, leading to clean auto-tag suggestions.
• Strong with long context. Claude handles long notes well, which matters when Ariv sends note content for AI tagging or Ask Brain context.

Note

Anthropic requires a paid API account. Visit console.anthropic.com for pricing details.

---

Ollama (Local)

For privacy-focused users and offline work.

Ollama runs AI models entirely on your machine. No data ever leaves your computer.

Setup

1. Install Ollama from ollama.ai.
2. Pull a model: open a terminal and run ollama pull llama3.1:8b (or whichever model you prefer).
3. Make sure Ollama is running (it starts a local server automatically).
4. In Ariv: set ai.provider to Ollama, then configure:
   • ai.ollama.url — the server address (default: http://localhost:11434)
   • ai.ollama.model — the model name (e.g., llama3.1:8b)

Popular Models

Model	Size	Quality	Notes
llama3.1:8b	~4.7GB	Good	Best balance of quality and resource usage
mistral	~4.1GB	Good	Fast, solid general-purpose model
phi3	~2.3GB	Moderate	Lightweight, runs on modest hardware

Why Choose Ollama

• Total privacy. Your notes never leave your computer. Zero network requests for AI features.
• No API costs. Once you download a model, usage is completely free forever.
• Works offline. Use AI features on a plane, in a cafe with no Wi-Fi, or anywhere else without internet.

Caution

Local models require decent hardware to run well. Expect to need 8GB+ RAM for a good experience, and 16GB+ for larger models. Response times depend on your machine’s CPU (or GPU if available) — cloud providers will generally respond faster.

Ollama Troubleshooting

• “Connection refused” error: Make sure Ollama is running. Open a terminal and run ollama serve if it’s not started.
• Slow responses: Try a smaller model like phi3 or ensure no other heavy processes are consuming system resources.
• Model not found: Run ollama list in a terminal to see which models are installed, and ollama pull <model> to download new ones.

---

Custom Models

All four providers regularly release new models. If you want to use a model that isn’t listed in Ariv’s dropdown, you can specify it directly:

1. Set your provider and API key as usual.
2. Enter the model’s ID in ai.customModel.
3. This overrides the dropdown selection.

This is useful for:

• Accessing newer models that haven’t been added to Ariv’s dropdown yet.
• Using fine-tuned models if your provider supports them.
• Specifying model variants (e.g., specific version tags or quantization levels for Ollama).

Note

When using a custom model, make sure the model ID matches exactly what your provider expects. Check your provider’s documentation for the correct model identifier string.

---

Switching Providers

You can change providers at any time without losing anything. Tags, Ask Brain history, and search indexes are stored locally and are independent of which AI provider generated them. Simply update ai.provider and the corresponding API key in Settings, and Ariv will start using the new provider for all future AI requests.

---

Related: AI Setup — step-by-step configuration guide | Auto-Tagging — how AI refines tag suggestions | Semantic Search — local AI-powered meaning search