Esta guía está disponible en inglés. La navegación del sitio sigue en español.
AI Actions
AI Actions
CyberPaste includes AI-powered actions that process your clipboard content, summarizing, translating, explaining code, and fixing grammar. They work with any OpenAI-compatible API provider.
Overview
The AI actions process clip content through customizable prompts:
- Right-click any clip → AI Actions → Choose action
- Or use the detail panel AI buttons
- Results appear in a dialog with a copy option
Workflow
- Select a clip (text, code, URL, or HTML/RTF, which are stripped to plain text)
- Open AI actions (right-click or detail panel)
- Choose an action (Summarize, Translate, etc.)
- AI processes the content
- View result in dialog
- Copy the result
ℹ️ An API key is required. Images are not supported; AI works on text content only.
ℹ️ The last result of each action is cached in the clip's metadata (
ai_summary,ai_translation,ai_explanation,ai_grammar_fix).
Smart Actions
Built-in Actions
| Action | Default Prompt |
|---|---|
| Summarize | "Summarize this content concisely." |
| Translate | Bidirectional: Spanish ↔ English; any other language is translated to English |
| Explain Code | "Explain this code snippet." |
| Fix Grammar | "Fix grammar and spelling." |
Customization
Each of the four actions can be customized:
- Custom display name (shown in menus; defaults to the localized label)
- Custom prompt (used as the system prompt)
ℹ️ You cannot create new actions beyond these four, and there are no per-action provider settings; the provider is global (Settings → AI).
Supported Providers
Built-in Providers
Selecting a provider auto-fills its base URL and default model; you only add your API key. Any model from the provider's list (or any other name) can be typed into the model field.
| Provider | Default Base URL | Default Model |
|---|---|---|
| OpenAI | https://api.openai.com/v1 |
gpt-5.6-luna |
| DeepSeek | https://api.deepseek.com |
deepseek-v4-flash |
| Kimi (Moonshot AI) | https://api.moonshot.ai/v1 |
kimi-k2.6 |
| Gemini | https://generativelanguage.googleapis.com/v1beta/openai |
gemini-3.5-flash |
| Grok (xAI) | https://api.x.ai/v1 |
grok-4.6 |
| Meta (Muse Spark) | https://api.meta.ai/v1 |
muse-spark-1.3 |
Available Models per Provider
| Provider | Models in the app |
|---|---|
| OpenAI | GPT-5.6 Luna (fast & cheap, default) · GPT-5.6 Terra (balanced) · GPT-5.6 Sol (flagship) · GPT-6 Astra (most capable) |
| DeepSeek | V4 Flash (default) · V4 Pro |
| Kimi | K2.6 (default) · K2.5 · K3 (flagship) |
| Gemini | 3.1 Flash-Lite · 3.5 Flash (default) · 3.8 Flash · 3.1 Pro |
| Grok | 4.6 (default) · 4.5 · 4.3 |
| Meta | Muse Spark 1.3 (latest, default) · 1.2 · 1.1 |
Any other model name can be entered manually (the field switches to custom mode when the value is not in the list).
Custom Provider
CyberPaste talks to every provider through the OpenAI Chat Completions format (POST {base_url}/chat/completions), so anything speaking that protocol works:
| Endpoint | Base URL | Model |
|---|---|---|
| Ollama (local, fully offline) | http://localhost:11434/v1 |
Any pulled model (e.g., llama3) |
| Any OpenAI-compatible API | Your endpoint URL | Your model name |
Provider Configuration
| Setting | Description |
|---|---|
| Provider | Select from list or custom |
| API Key | Your provider API key |
| Model | Model name (e.g., "gpt-4o", "llama3") |
| Base URL | Custom API endpoint URL |
Configuration
Provider Settings
- Go to Settings → AI
- Select your provider
- Enter API key
- Choose model (or enter custom name)
- For custom providers, enter base URL
Custom Prompts
Each action uses a prompt that is sent as the system prompt; the clip content is always sent as the user message:
- Default prompts are pre-configured
- Edit prompts in Settings → AI (Enter confirms each field)
- The v1.20 prompt editor adds rename/reset controls and descriptions for every field, with truthful default placeholders
Example Custom Prompt
Since the clip content arrives as the user message, your prompt just describes what to do with it:
Summarize the following text in 3 bullet points.
AI Result Dialog
After processing, the result dialog shows:
- The action title
- AI-generated result
- Copy button to copy the result
ℹ️ There are no Replace or Regenerate buttons; run the action again from the context menu to regenerate, and paste or edit manually to replace content.
Tips for Better Results
- Be specific in prompts — Your prompt is the system message; include desired format, length, and style
- Use code blocks — For code explanation, specify the language in your prompt
- Chunk large content — Very long clips may exceed the provider's token limits
- Use local models — Ollama for privacy-sensitive content
Privacy Note
When using AI:
- Clip content is sent to the configured API provider
- No data is sent anywhere else
- For maximum privacy, use a local Ollama instance (Custom provider →
http://localhost:11434/v1) - API keys are stored locally only