Overview
Silentis is a privacy-focused, offline AI assistant designed to provide comprehensive and accurate responses to user queries. It supports multiple AI models optimized for various tasks, such as logical reasoning, coding, and natural language processing. The application runs locally on your device, ensuring data privacy, and includes features like chat history management, document integration, and model customization.
This documentation provides detailed instructions on how to install, configure, and use Silentis, covering all features and functionalities.
Table of Contents
System Requirements
Before installing Silentis, ensure your system meets the following requirements:
- Operating System: Windows, macOS, or Linux
- RAM: Minimum 4GB (8GB recommended for most models)
- Storage: At least 10GB free space for model files
- Processor: Multi-core CPU (GPU optional for acceleration)
- Dependencies: Python 3.8+ (bundled in the compiled version)
- Internet: Required only for downloading models; offline usage supported thereafter
Installation
Silentis is distributed as a compiled executable for ease of use. Follow these steps to install:
- Download the Application:
- Visit silentis.ai to download the latest version of Silentis for your operating system.
- Choose the appropriate package (Windows
.exe
, macOS .dmg
, or Linux .AppImage
).
- Install the Application:
- Windows: Run the
.exe
file and follow the installer prompts.
- macOS: Open the
.dmg
file, drag Silentis to the Applications folder, and launch it.
- Linux: Make the
.AppImage
executable (chmod +x Silentis.AppImage
) and run it.
- Initial Setup:
- Upon first launch, Silentis creates a configuration directory:
- Windows:
%APPDATA%\AIAssistant
- macOS:
~/.config/AIAssistant
- Linux:
~/.local/share/AIAssistant
- The application will prompt you to select a model file or download one (see Model Management).
- Optional Model Download:
- Silentis requires a GGUF model file to function. You can download supported models from the application’s Downloads section or manually from Hugging Face (see supported models below).
Getting Started
- Launch Silentis:
- Open the application. The interface loads in a desktop window powered by PyQt and Flask, displaying the chat interface by default.
- Load a Model:
- Navigate to the Settings section from the sidebar.
- Click Browse to select a GGUF model file or Download to fetch one from the supported list.
- Click Load Model to initialize the AI.
- Start Chatting:
- Return to the Chat section.
- Type a query in the text box and click Send or press
Ctrl+Enter
.
- The AI will respond, and the conversation will be saved automatically.
User Interface Overview
The Silentis interface is divided into several sections, accessible via the sidebar:
- Sidebar:
- Logo and Title: Displays the Silentis logo and name.
- Chat Menu: Dropdown with options for Chat, Chat History, Clear Chat, Export Chat, Export to PDF, and Import Chat.
- Downloads: Access model download options.
- Settings: Configure model, document, API, and theme settings.
- About: Links to the Silentis website.
- Status Bar: Shows model status, GPU usage, document status, memory usage, and software version.
- Main Content:
- Hero Section: Displays the Silentis title and version.
- Interface Section: Contains the active section (Chat, Chat History, Settings, or Downloads).
- Chat Container: Displays conversation messages with user and AI responses.
- Error Container: Shows error messages if any occur.
- Footer: Contains copyright and version information.
- Theme Support:
- Toggle between light and dark themes in the Settings section for better visibility.
Core Features
Chat Interface
The chat interface is the primary way to interact with Silentis.
- Sending Messages:
- Enter your query in the text area (max 4096 characters).
- Click Send or press
Ctrl+Enter
to submit.
- The AI processes the query and streams the response in real-time.
- Message Features:
- Copy: Click the Copy button to copy a message to the clipboard.
- Code Blocks: Code in responses is syntax-highlighted with a Copy Code button.
- Edit: Edit previous messages by clicking the Edit button (available for user messages).
- Timestamps: Each message includes a timestamp for reference.
- Document Context: If a document is uploaded, the AI uses it for relevant queries, indicated by a “Document loaded” status.
- Keyboard Shortcuts:
Ctrl+Enter
: Send message
Ctrl+N
: Start a new chat
Ctrl+L
: Clear current chat
Chat History Management
Silentis automatically saves conversations and allows you to manage them.
- Accessing Chat History:
- Click Chat History in the Chat dropdown.
- View a list of saved chats with titles, timestamps, and actions.
- Features:
- Search: Filter chats by title or content using the search bar.
- Pagination: Navigate through multiple pages of chats.
- Load Chat: Click a chat title to load its messages.
- Edit Title: Click Edit Title to rename a chat.
- Delete: Click Delete to remove a chat (requires confirmation).
- New Chat: Click New Chat or press
Ctrl+N
to start a fresh conversation.
- Clear Chat:
- Select Clear Chat from the Chat dropdown or press
Ctrl+L
to reset the current conversation.
Model Management
Silentis supports multiple AI models, each optimized for specific tasks.
- Supported Models:
ID |
Name |
RAM Required |
License |
Description |
1 |
Reasoner v1 |
8GB |
Apache 2.0 |
Optimized for logical reasoning and coding tasks |
2 |
Llama 3 8B Instruct |
8GB |
LLaMA License |
Great for natural language and instruction following |
3 |
DeepSeek-R1-Distill-Qwen-7B |
8GB |
MIT |
Specialized in knowledge distillation and complex topics |
4 |
Phi-3 Mini Instruct |
4GB |
MIT |
Lightweight model for quick responses |
- Loading a Model:
- Go to Settings > Model Settings.
- Click Browse to select a local GGUF file or navigate to Downloads to fetch a model.
- Click Load Model to initialize it.
- The status bar updates to show the loaded model’s name.
- Downloading Models:
- Go to Downloads.
- Select a download folder by clicking Browse.
- Click Download next to a model to start the process.
- A progress bar shows the download status, and a notification confirms completion.
- Model Parameters:
- Temperature: Adjust creativity (0.1–1.0, default 0.7).
- Max Tokens: Set response length (100–4096, default 2048).
- Use GPU: Enable GPU acceleration if available (requires compatible hardware).
Document Upload
Silentis can process documents to provide context for queries.
- Supported Formats:
.txt
and .pdf
- Uploading a Document:
- Go to Settings > Document Upload.
- Click Browse to select a file.
- Click Upload Document to process it.
- The status bar updates to “Loaded,” and the AI uses the document content when relevant.
- Usage:
- The AI automatically detects if a query relates to the document (e.g., keywords like “document” or matching terms).
- Responses include a “Document loaded” indicator when the document is used.
- Document content is limited to 10,000 characters to optimize performance.
Settings Configuration
Customize Silentis through the Settings section.
- Theme:
- Toggle between light and dark themes for better readability.
- Model Settings:
- Model File: Specify the path to a GGUF file.
- Temperature: Control response randomness.
- Max Tokens: Limit response length.
- Use GPU: Enable/disable GPU acceleration.
- System Prompt: Customize the AI’s behavior (see Custom System Prompts).
- Document Upload:
- Manage document uploads as described above.
- API Settings:
- Enable API: Activate the REST API for external access.
- API Port: Set the port (1024–65535, default 5000).
- API Key: Optional key for authentication.
- Note: Changing API settings requires restarting the application.
- Saving Settings:
- Click Save Settings to apply changes.
- Settings are stored in
config.json
in the configuration directory.
API Access
Silentis includes a REST API for programmatic interaction (disabled by default).
- Enabling the API:
- Go to Settings > API Settings.
- Check Enable API and set a port and optional API key.
- Save settings and restart the application.
- API Endpoints:
- POST /chat: Send a prompt and receive a streamed response.
- Body:
{ "prompt": "Your query" }
- Headers:
{ "X-API-Key": "your_key" }
(if enabled)
- POST /clear_chat: Clear the current chat.
- GET/POST/DELETE /chat_history: Manage chat history.
- POST /edit_message: Edit a specific message.
- POST /export_chat: Export a chat to JSON.
- POST /import_chat: Import a chat from JSON.
- POST /load_model: Load a model file.
- GET/POST /config: Get or update configuration.
- GET /models: List supported models.
- POST /download_model: Download a model.
- POST /upload_document: Upload a document.
- POST /export_pdf: Export chat to PDF.
- Security:
- Use an API key to secure endpoints.
- The API runs locally (
localhost
), so external access requires network configuration.
Advanced Usage
Custom System Prompts
The system prompt defines the AI’s behavior. Silentis automatically sets a prompt based on the loaded model but allows customization.
- Default Prompts:
- Reasoner v1: Focuses on logical reasoning and detailed code explanations.
- Llama 3 8B Instruct: Emphasizes natural, conversational responses.
- DeepSeek-R1-Distill-Qwen-7B: Prioritizes comprehensive, academic answers.
- Phi-3 Mini Instruct: Delivers concise, straightforward responses.
- Customizing:
- Go to Settings > Model Settings.
- Edit the System Prompt text area.
- Example:
You are a technical writer. Provide detailed documentation with examples.
- Click Save Settings to apply.
- Tips:
- Include clear instructions for tone, style, or task focus.
- Use
<|eot_id|>
to mark the end of the prompt for compatibility.
Exporting Chats
Save conversations for backup or sharing.
- Export to JSON:
- Select Export Chat from the Chat dropdown.
- Choose a folder and save the file (e.g.,
chat_1.json
).
- The file includes the chat ID, title, messages, and timestamps.
- Export to PDF:
- Select Export to PDF from the Chat dropdown.
- Choose a folder and save the file (e.g.,
chat_history.pdf
).
- The PDF includes formatted messages with roles and timestamps.
Importing Chats
Restore previously exported conversations.
- Importing:
- Select Import Chat from the Chat dropdown.
- Choose a
.json
file exported from Silentis.
- The chat is added to the history with a new ID and loaded into the chat interface.
Using GPU Acceleration
GPU acceleration can improve performance for supported models.
- Requirements:
- Compatible NVIDIA GPU with CUDA support.
- Appropriate drivers and libraries installed (bundled in the compiled version).
- Enabling:
- Go to Settings > Model Settings.
- Check Use GPU.
- Save settings and reload the model.
- The status bar shows “Active” for GPU usage.
- Note:
- GPU usage increases memory requirements.
- If unavailable, Silentis falls back to CPU automatically.
Troubleshooting
- Model Fails to Load:
- Verify the GGUF file path is correct.
- Ensure sufficient RAM (check model requirements).
- Check the console for detailed errors (run from a terminal if possible).
- No Response from AI:
- Confirm a model is loaded (status bar shows model name).
- Check for errors in the error container.
- Ensure the prompt is valid and not empty.
- Slow Performance:
- Enable GPU acceleration if available.
- Use a lighter model like Phi-3 Mini Instruct.
- Close other resource-intensive applications.
- API Not Working:
- Verify the API is enabled and the port is not in use.
- Check the API key if authentication is enabled.
- Restart the application after changing API settings.
- Document Not Loading:
- Ensure the file is a supported format (
.txt
or .pdf
).
- Check file permissions and path validity.
- Try a smaller file if the content exceeds 10,000 characters.
FAQs
- Q: Do I need an internet connection to use Silentis?
A: Internet is only required to download models. Once downloaded, Silentis operates fully offline.
- Q: Can I use my own models?
A: Yes, any GGUF-compatible model can be loaded by specifying its path in Settings.
- Q: How secure is Silentis?
A: Silentis runs locally, ensuring no data leaves your device. The optional API can be secured with a key.
- Q: What happens if I exceed the character limit for prompts?
A: The input is capped at 4096 characters, and a counter displays the current usage.
- Q: Can I run multiple instances of Silentis?
A: Yes, but ensure each instance uses a different API port to avoid conflicts.
Support and Feedback
For support, visit silentis.ai or contact the Silentis team via the About section. We welcome feedback to improve the application. Please include your operating system, version (v1.0 - Preview), and a detailed description of any issues.
Silentis v1.0 Changelog
Last Updated: April 12, 2025
© 2025 Silentis. All rights reserved.