Providers and models
Use better-vault with OpenAI, Anthropic Claude, Google Gemini, Ollama or LM Studio, with your own API key, your ChatGPT plan or a local model on your computer.
better-vault works with five providers: you bring the key or the local server, and you pick the provider and model for each question in the prompt bar.
| Provider | Needs | Web mode | Pictures (Infographic) |
|---|---|---|---|
| OpenAI | API key or ChatGPT sign-in | Yes | Yes (GPT Image models) |
| Anthropic | API key | Yes | No |
| Google (Gemini) | API key | Yes | Yes (Gemini image models) |
| Ollama | Ollama running on your machine | No | No |
| LM Studio | LM Studio's local server | No | No |
Add an API key
Get a key
Create one in the provider's console: OpenAI, Anthropic or Google AI Studio.
Paste it in better-vault
Open Settings → AI providers, choose the provider and paste the key. better-vault loads the model list from the provider with your key, so new models appear without an update.
Your key is stored on your device and sent only with the requests that need it. Privacy explains what is sent where.
Use your ChatGPT plan
If you pay for ChatGPT, better-vault can use your plan instead of an OpenAI API key. Chat and every Studio tool then count against the usage included in your plan.
Continue with ChatGPT
Open Settings → AI providers → OpenAI and choose Continue with ChatGPT, then Open ChatGPT. Your browser opens OpenAI's sign-in page.
Sign in and allow it
Sign in to ChatGPT and allow better-vault to use your plan. The page then tells you to go back to better-vault, where OpenAI's models are ready in the prompt bar.
You can review and limit what better-vault uses at ChatGPT → Settings → Usage, and sign out from the same place in better-vault's settings. Pictures for infographics still need an API key, because a ChatGPT plan can't generate images in other apps.
The sign-in works only when better-vault's server runs on the same computer as your browser: in the desktop app, or when you run the web app yourself. The tokens stay with that server, in a private file in its data folder, and never reach the browser.
Use a local model
Install a model server
Install Ollama or LM Studio and download a model. With Ollama, this downloads a small one:
ollama pull llama3.2Run better-vault on the same computer
Use the desktop app, or run the web app yourself as shown in Get started.
Turn the provider on
In Settings → AI providers, enable Ollama (http://localhost:11434) or LM Studio (http://localhost:1234), then pick the model in the prompt bar.
With a local model, your sources never leave your computer. Web mode is not available, because it relies on a provider's built-in search.
Which model should I use?
Any recent large model works for chat and reports, and larger models follow citations more closely. Studio tools that return structured data, such as quizzes, flashcards and Anatomy, need a model that follows JSON schemas well; small local models may need a retry.
For infographics, you choose an image model in the tool's dialog. Its key is sent with that request only.
Desktop app
Install the better-vault desktop app for macOS, Windows or Linux. It has the same interface as the web app, keeps notebooks in a local file and updates itself.
Privacy and your data
better-vault is local-first. Notebooks, sources, chats and keys stay in your browser or in a file on your computer, and nothing is ever kept on a server.