Max Nardit

AI Setup

Beetroot can transform clipboard text and images using AI -- translate, summarize, fix grammar, read text from photos, describe images, extract data, and more. You bring your own API key (BYOK). Beetroot never stores or transmits your data except to the AI provider you choose, and only when you explicitly ask for a transform. All AI API calls run in native Rust -- no browser JavaScript, no CORS issues.

Version: 1.6.5 Last updated: 2026-04-04

Supported Providers

ProviderSpeedCostBest for
OpenAIFastPaidGeneral transforms, widely used
Google GeminiFastFree tier availableBudget-friendly option
AnthropicFastPaidHigh-quality rewrites
DeepSeekModerateVery cheapDeep reasoning tasks
Local LLMVariesFreeFull privacy, no internet needed

Cloud Provider Setup

OpenAI

Models: gpt-5.4-nano (fast, cheapest) or gpt-5.4-mini (smarter, 2x faster than previous generation)

  1. Go to platform.openai.com and sign in or create an account.
  2. Navigate to API keys and create a new key. It starts with sk-.
  3. In Beetroot, open Settings > AI.
  4. Select OpenAI as your provider.
  5. Paste your API key.
  6. Choose a model -- gpt-5.4-nano is recommended to start (fast and cheap).
  7. Click Test to verify the connection.
  8. Click Save.

If you previously used gpt-5-nano or gpt-5-mini, Beetroot auto-migrates to gpt-5.4 models on launch.

Google Gemini

Models: gemini-2.5-flash-lite (fastest) or gemini-2.5-flash (better reasoning)

  1. Go to aistudio.google.com and sign in with your Google account.
  2. Click "Get API key" and create a key. It starts with AIza.
  3. In Beetroot, open Settings > AI.
  4. Select Google Gemini as your provider.
  5. Paste your API key.
  6. Choose a model -- gemini-2.5-flash-lite is the fastest option.
  7. Click Test to verify.
  8. Click Save.

Gemini offers a generous free tier, making it a great starting point if you want to try AI transforms without spending anything.

Anthropic

Models: claude-haiku-4-5 (fastest) or claude-sonnet-4-6 (best balance of speed and quality)

  1. Go to console.anthropic.com and sign in or create an account.
  2. Navigate to API keys and create one. It starts with sk-ant-.
  3. In Beetroot, open Settings > AI.
  4. Select Anthropic as your provider.
  5. Paste your API key.
  6. Choose a model -- claude-haiku-4-5 is recommended for quick transforms.
  7. Click Test to verify.
  8. Click Save.

DeepSeek

Models: deepseek-chat (everyday tasks) or deepseek-reasoner (deep reasoning, chain-of-thought)

  1. Go to platform.deepseek.com and sign in or create an account.
  2. Create an API key in your dashboard.
  3. In Beetroot, open Settings > AI.
  4. Select DeepSeek as your provider.
  5. Paste your API key.
  6. Choose a model -- deepseek-chat is good for most tasks; deepseek-reasoner is for complex analysis.
  7. Click Test to verify.
  8. Click Save.

DeepSeek offers very competitive pricing. The deepseek-reasoner model shows its thinking process, which is automatically cleaned from the output.

Local LLM Setup (No Internet Required)

Run AI transforms entirely on your computer with no API key and no data leaving your machine.

Option 1: Ollama

Ollama is the easiest way to run local models.

  1. Download and install Ollama from ollama.com.
  2. Open a terminal and pull a model: 
    ollama pull llama3.2
  3. Ollama runs automatically in the background on port 11434.
  4. In Beetroot, open Settings > AI.
  5. Select Local LLM as your provider.
  6. Choose the Ollama preset -- the endpoint fills in automatically.
  7. Click Test -- Beetroot will connect and show a dropdown of your installed models.
  8. Select your model from the dropdown.
  9. Click Save.

Recommended Ollama models for text:

  • llama3.2 -- Good general-purpose model, runs on most hardware
  • mistral -- Fast and capable
  • gemma2 -- Google's open model, good for text tasks

Recommended Ollama models for vision:

  • llava -- General-purpose vision model
  • bakllava -- Better at detailed descriptions
  • moondream -- Lightweight, fast

Option 2: LM Studio

LM Studio provides a graphical interface for managing and running local models.

  1. Download and install LM Studio from lmstudio.ai.
  2. Download a model through the LM Studio interface.
  3. Start the local server (LM Studio runs on port 1234 by default).
  4. In Beetroot, open Settings > AI.
  5. Select Local LLM as your provider.
  6. Choose the LM Studio preset.
  7. Click Test to verify the connection. The loaded model is detected automatically.
  8. Click Save.

Option 3: Custom Endpoint

Any server that speaks the OpenAI-compatible API format works with Beetroot.

  1. Start your model server.
  2. In Beetroot Settings > AI, select Local LLM.
  3. Choose Custom and enter your endpoint URL (e.g., http://127.0.0.1:8080/v1/chat/completions).
  4. Enter the model name.
  5. Click Test and Save.

Model Recommendations

GoalProviderModelWhy
Cheapest cloud optionGoogle Geminigemini-2.5-flash-liteGenerous free tier
Best qualityAnthropicclaude-sonnet-4-6Excellent rewrites and translations
FastestOpenAIgpt-5.4-nanoLowest latency
Deep analysisDeepSeekdeepseek-reasonerChain-of-thought reasoning
Best vision (cloud)Google Geminigemini-2.5-flashFast, well-structured OCR output
Best vision (local)Local LLMQwen 3.5 4B (LM Studio)Accurate OCR, runs on most hardware
Full privacyLocal LLMllama3.2 (Ollama)Nothing leaves your machine
Offline useLocal LLMAny Ollama modelWorks without internet

AI Vision Transforms

Starting in v1.6.5, Beetroot can analyze images from your clipboard history using AI vision models. Right-click any image clip → AI, and choose a vision prompt.

Five built-in vision prompts:

PromptWhat it does
Read TextOCR from photos, screenshots, handwritten notes
Describe ImageGet a text description of what's in the image
Extract DataPull structured data from tables, receipts, prescriptions
Summarize ImageQuick summary of visual content
Translate Image TextTranslate text found in images

You can also create custom vision prompts in Settings > AI.

Vision-capable providers:

  • Cloud: OpenAI GPT-5.4, Anthropic Claude (Haiku/Sonnet), Google Gemini 2.5, DeepSeek
  • Local: Ollama (llava, bakllava, moondream), LM Studio (any vision model)

Tip: Even small local vision models (4B parameters) can read handwritten text accurately. If privacy matters, use Ollama or LM Studio -- no data leaves your machine.

Background AI Processing

AI transforms (both text and vision) run in the background. Click a prompt, the menu closes instantly, and you can keep working. When the result is ready:

  • If Beetroot is visible, the new clip appears at the top of the list.
  • If Beetroot is hidden, a native Windows notification pops up. Click it to bring Beetroot to the front.

You can queue multiple transforms -- they run one by one.

Using AI Transforms

Once a provider is set up, there are several ways to transform text and images:

Method 1: Transform Panel

  1. Select a text clip in the list.
  2. Press Alt+T (or right-click > Transform).
  3. The Transform panel shows built-in text transforms (UPPERCASE, lowercase, Title Case, Trim whitespace, Remove spaces, Single line, Sort lines, Remove duplicates) and your AI prompts below. Use the search box to filter.
  4. Click an AI prompt. The text is sent to your provider.
  5. The transformed result is saved as a new clip in your history.

Method 2: Quick Access from Context Menu

  1. Right-click a text clip.
  2. At the bottom of the context menu, you'll see your Quick Access prompts (up to 5).
  3. Click one to transform the text immediately.

To enable Quick Access on a prompt, go to Settings > AI and check the "Quick Access" box next to the prompts you use most often. Up to 5 prompts can be Quick Access at once.

Method 3: Vision Transforms (Image Clips)

  1. Right-click an image clip.
  2. Hover over the AI submenu.
  3. Choose a vision prompt (Read Text, Describe Image, etc.).
  4. The image is sent to your AI provider. The result is saved as a new text clip.

Built-in Prompts

Beetroot comes with 10 ready-to-use AI prompts:

PromptWhat it does
Fix GrammarCorrects grammatical errors without changing meaning
Any to EnglishDetects the language and translates to English
SummarizeCondenses text into 2-3 key sentences
Make ProfessionalRewrites in a clear, business-appropriate tone
Format as CodeApplies proper code indentation and formatting
Bullet PointsConverts text into a bulleted list
SimplifyRewrites in plain, simple language
Make ShorterCondenses to roughly half the length
Explain ThisExplains in simple terms for anyone
Extract Key DataExtracts names, dates, numbers, and URLs

You can also create your own custom prompts (up to 20 total, including built-ins) in Settings > AI.

Troubleshooting

ProblemSolution
"Set API key in Settings"You have not entered an API key for the selected provider. Go to Settings > AI.
"API key is invalid"Double-check your key. Make sure it matches the selected provider.
"Request timed out (30s)"The provider took too long. Try again or switch to a faster model.
"Empty response"The AI returned nothing. Try a different prompt or model.
"Text too long for AI transform"Clips longer than 50,000 characters cannot be processed. Copy a shorter section.
Local LLM "Failed"Make sure your model server (Ollama or LM Studio) is running and the endpoint is correct.

Privacy

  • Beetroot only sends data to an AI provider when you explicitly request a transform.
  • For text transforms, only the specific clip text is sent. For vision transforms, only the specific image is sent.
  • Your full clipboard history is never transmitted.
  • If you use a Local LLM, no data leaves your machine at all.
  • All AI API calls are made from native Rust code -- no browser JavaScript, no third-party servers in between.
  • API keys are stored locally on your computer and are never transmitted anywhere except to their respective provider.

Last updated: 2026-04-04