Ask OpenRouter (NVDA Add-on)

This NVDA add-on allows you to interact with Artificial Intelligence models provided by the OpenRouter platform directly from your screen reader.

The add-on supports both:

• Automatic random selection of free models
• Manual selection of any available model (including paid ones)

Key Features

• Quick Access: Open the chat interface anytime with a global shortcut.
• Conversation Management: Start a fresh conversation or continue your previous exchange.
• Smart Free Model Rotation: Automatically selects a random free model to optimize daily usage quotas.
• Manual Model Selection: Choose a specific model (including paid models) from the settings panel.
• Accessible Results: View responses in a clear, easy-to-navigate window with optional full history display.

Configuration: Obtaining and Installing Your API Key

To use this add-on, you must have an API key from OpenRouter.

Even when using free models, the key is required to identify your requests.

1. How to get an API key

1. Go to OpenRouter.ai.
2. Create an account by clicking "Sign up" (you can sign in with a GitHub, Google or MetaMask account, or your email address).
3. Once logged in, navigate to the "Keys" section in your dashboard, or go directly to: https://openrouter.ai/keys
4. Click the "Create Key" button.
5. Give your key a name (for example: "My OpenRouter API key") and click "Create".
6. Important: Your key will be displayed only once. Copy it immediately and store it in a secure place.

2. Setting up the key in NVDA

1. Open the NVDA menu (NVDA + N).
2. Navigate to Preferences, then Settings.
3. In the categories list, select "Ask OpenRouter".
4. Paste your API key into the "OpenRouter API Key" field.
5. Press OK to save.

Show API Key

In the NVDA settings panel, just after the "OpenRouter API Key" field, there is a checkbox labeled:

"Show API key"

If checked, the characters of the API key become visible.
By default, they are hidden for security reasons.

Model Selection Settings

In the Ask OpenRouter settings category, you will find a new option:

"Use all models, including paid ones"

This option controls how models are selected.

When the option is UNCHECKED (default behavior)

• The add-on automatically selects a random free model for each new conversation.
• It rotates between available free models.
• This helps distribute usage and avoid rate limits.

When the option is CHECKED

When this option is enabled, a list of available models automatically appears after the checkbox.

• The list is sorted in ascending order based on prompt token pricing (cost per input token), from lowest to highest.
• Only non-deprecated models with valid providers are displayed.

What can you do when this option is enabled?

• Choose any available model.
• Use paid models (if you have sufficient OpenRouter credits).
• Select the model that best fits your needs.
• Keep using the same selected model for your conversations (no automatic rotation).

What is a prompt token?

A prompt token represents a small unit of text sent to the model (your question or input).

Models are usually billed separately for:

• Input tokens (prompt)
• Output tokens (completion)

How to Use

Opening the Chat Dialog

Press:

Ctrl + Alt + A

You can change this gesture in: NVDA menu → Preferences → Input Gestures → Ask OpenRouter

Main Interface

The dialog contains three buttons:

1. New Chat – Starts a brand new conversation.
2. Continue Chat – Resumes the previous conversation (keeps history).
3. Close – Closes the dialog (Escape also works).

Entering Your Prompt

After selecting "New Chat" or "Continue Chat":

• A multiline text field appears.
• Pressing Enter inserts a new line.
• To send your message:
  • Press Tab to reach the OK button.
  • Press Enter.

Reading the Response

After processing, a results window appears containing:

• "You said:" followed by your message.
• "The model replied:" followed by the response.
• A "Copy" button to copy the response.

If full history display is enabled, each exchange is clearly separated by headings, making it easy to navigate using your NVDA's quick navigation keys.

Display Options

If you prefer to only display the latest response instead of the full conversation history:

1. Open NVDA menu (NVDA + N).
2. Go to Preferences → Settings.
3. Select Ask OpenRouter.
4. Uncheck: "Display the full chat history for continuous discussions"
5. Press OK.

Unassigned Scripts

The following scripts do not have gestures assigned.
You can define them in:

Preferences → Input Gestures → Ask OpenRouter

Available scripts:

• Open the add-on settings panel
• Start a new chat directly
• Continue an existing chat directly

Free Models, Paid Models and Quotas

Free Model Usage

When "Use all models, including paid ones" is unchecked:

• Only models labeled as free on OpenRouter are used.
• Free models have:
  • Limited daily quotas
  • Shared rate limits
  • Possible temporary unavailability

The add-on automatically rotates between free models to improve availability.

Paid Model Usage

When "Use all models, including paid ones" is checked:

• The add-on uses the exact model you selected.
• This may include paid models.
• You must have sufficient OpenRouter credits.
• Provider rate limits may apply.

Errors such as:

• 402 (insufficient credits)
• 429 (rate limited)
• 404 (model not allowed by privacy settings)

are displayed directly to inform you of the issue.

Privacy Settings Reminder

If you use free models and receive an error mentioning:

"No endpoints found matching your data policy"

You may need to adjust your OpenRouter privacy settings:

https://openrouter.ai/settings/privacy

Ensure that public/free model endpoints are allowed.

Compatibility

• This add-on is compatible with the versions of NVDA ranging from 2025.1 and beyond.

Changes for 20260221.0.0

• Added manual selection of any available model from the settings panel
• Added ability to use paid models

Changes for 20260217.0.0

• Initial version