Guide

Set up AI with your own key or a local model

FlowReader has no model of its own. The assistant uses a provider you connect. That’s either a cloud provider with your API key, or a model running on your computer.

Before you start

  • FlowReader Pro on a signed-in account. AI chat needs FlowReader Pro, also with a local model.
  • The browser extension for Chromium or Firefox, or the desktop app.
  • For a cloud provider, an API key from Anthropic, OpenAI, Google, Hugging Face or OpenCode Go. The provider bills you for what you use.
  • For a local model, Ollama or LM Studio on the same computer, with at least one model downloaded.

Steps

Connect a cloud provider

  1. Open Settings, then AI features, then Overview & models. Ctrl+Shift+, (Cmd+Shift+, on a Mac) opens Settings.
  2. Under Providers, in the Online list, choose Configure next to your provider.
  3. Choose Get an API key to open the provider’s key page. Create a key there and copy it.
  4. Paste the key into the key field, for example Anthropic API key, and choose Connect.
  5. When the status reads Connected, pick a model under Model. Choosing a model makes this provider the one chat uses.

To check the whole path, choose Test connection. It sends one short request.

Connect a local model

  1. Start Ollama, or start the local server in LM Studio, and make sure a model is downloaded.
  2. In Overview & models, in the On this device list, choose Configure next to Ollama or LM Studio.
  3. Leave Endpoint at the default address, http://localhost:11434 for Ollama and http://localhost:1234 for LM Studio, unless the runtime listens somewhere else.
  4. Choose Connect, then pick one of your models under Model.

Ollama and LM Studio must run on the same computer. A runtime on another machine in your local network, a gateway, or another OpenAI-compatible endpoint can be added under Custom provider.

The AI provider settings. Online providers such as Anthropic, OpenAI and Google and the on-device LM Studio and Ollama are not set up, and a custom Home lab provider is marked Active.

Ask the first question

  1. Open Assistant in the sidebar.
  2. Ask about something in your library. Answers cite the items they draw on.

What you should see

The top of Overview & models names the provider and model in use, followed by what the assistant’s tools may do. The provider’s row reads Ready with the model, and the provider chat uses is marked Active.

In Assistant, a question gets an answer with citations that open the passage in the reader.

Where your key is kept

Your key is stored encrypted on this device and is not part of your synced library.

With Online Sync on, FlowReader copies the key to your other devices through its own server. The server can read the key while copying it. A Monitor that uses AI runs on FlowReader’s server with your key.

To stop using a key, remove it in Overview & models and revoke it on the provider’s key page.

Common problems

  • “AI provider setup required”

    Cause: No provider is connected yet, or the connected provider has no model chosen.

    Fix: Choose Configure AI in the message and follow the steps above.

  • “Activate FlowReader Pro to use AI chat.”

    Cause: Pro is not active on the account you are signed in with. This applies to local models too.

    Fix: Start a trial or activate Pro under Plan in Settings › Account & sync, then ask again.

  • “… did not accept this key. Check that you copied all of it, or create a new one.”

    Cause: The key is incomplete, mistyped or revoked.

    Fix: Copy the whole key again from the provider, paste it and choose Connect. If it still fails, create a new key.

  • “… accepted the key but refused access.”

    Cause: The key is valid, but the provider account has no credit, no billing set up, or a plan without API access.

    Fix: Sort out billing or the plan with the provider, then choose Connect again.

  • “Could not reach Ollama at http://localhost:11434. Make sure it is running, then connect again.”

    Cause: The local runtime is not running, or it listens on a different address.

    Fix: Start Ollama or the LM Studio server, check the address in Endpoint, and choose Connect.

  • “Only loopback endpoints are supported for local runtimes.”

    Cause: The Endpoint for Ollama or LM Studio points at another machine. These two rows only connect to the same computer, such as http://localhost.

    Fix: Set Endpoint back to a localhost address, or add the other machine under Custom provider.

  • “… answered but has no models.”

    Cause: The runtime is running but has no model downloaded or loaded.

    Fix: Download or load a model in Ollama or LM Studio, then choose Connect again.

  • “The selected model is unavailable. Refresh the model list in Settings.”

    Cause: The provider has retired or renamed the model.

    Fix: Choose Refresh models in Overview & models and pick another model.

All help topicsStill stuck?Contact support