repository·main·Indexed 19 days ago
https://github.com/npc-worldwide/incognideA unified workspace integrating chat, code editing, document management (Office, PDF, LaTeX), web browsing, and 3D model viewing. It features intelligent context, composable automations via Jinx workflows, and support for both Cloud AI (OpenAI, Anthropic, Gemini) and Local AI providers (Ollama, LM Studio, llama.cpp, oMLX). The platform includes tools for knowledge graph exploration, agent memory management, and NPC persona customization.
You can control the behavior and history of your AI personas through several interfaces:
A Jinx is a reusable automation template that combines natural language prompts with code execution.
Key features include:
Incognide provides tools for research and knowledge management:
To install Incognide, download the installer for your specific platform (Linux, macOS, or Windows) from enpisi.com/incognide, run the installer, and launch the application.
# No specific command provided; download from enpisi.com/incognideIncognide uses an external Python virtual environment (venv) for heavy tasks like fine-tuning to keep the main executable small. To set this up:
<workspace>/.venv.PyTorch (CPU): torch torchvision torchaudioPyTorch (CUDA): torch torchvision torchaudio + CUDA indexDiffusers (Image Gen): diffusers transformers accelerate safetensorsTransformers (LLM): transformers accelerate safetensors sentencepieceWhisper (Speech): openai-whisperOn the first launch, a five-step wizard guides you through initial configuration:
~/.incognide).system/zsh/bash on macOS/Linux, powershell on Windows), and opt-in to local activity tracking for next-action predictions.You can automate workflows by scheduling jinxes and agents to run via Cron jobs. Common automated tasks include:
You can configure guidance context to focus these extractions on specific topics. Scheduling can be initiated directly from the Memory Manager or the Knowledge Graph Editor.
To set up Incognide manually, ensure you have the following prerequisites:
npcpy (Core Python library)npcts (React component library, installed via npm)Follow these steps to clone, install, and run the application:
# 1. Clone and install dependencies
git clone https://github.com/npc-worldwide/incognide.git
cd incognide
npm install
# 2. Run the components
python incognide_serve.py # Start the Backend
npm run dev # Start the Frontend (Vite)
npm start # Start the Electron appIncognide can connect to local LLM providers by probing specific ports and binary locations. The Model Manager allows you to Start/Stop these services.
| Provider | Server Port | Binary / App Checks |
|---|---|---|
| Ollama | 11434 | ollama in PATH; /Applications/Ollama.app on macOS |
| LM Studio | 1234 | lms in PATH; /Applications/LM Studio.app on macOS |
| llama.cpp | 8080 | llama-server, llama-cli, or koboldcpp in PATH |
| oMLX | 8000 | omlx in PATH; /Applications/oMLX.app (macOS only) |
open -a Ollama (macOS) or ollama serve (Linux). Stop via pkill -f 'ollama serve' or the Windows system tray.lms server start / lms server stop if the lms CLI is present; otherwise, use the Developer tab in the app.llama-server -m <model.gguf> --port 8080). Stop via pkill -f llama-server.open -a oMLX (macOS) and manage the server via the menu-bar icon.Note: GGUF/GGML model files can be loaded directly without a server only if llama.cpp or koboldcpp is installed.
# Example llama.cpp start command provided by Incognide
llama-server -m <model.gguf> --port 8080To use Cloud AI, add your API keys during the setup wizard's Cloud keys step. You can also add or update them later via:
Keys are stored in ~/.incogniderc using the format: export <PROVIDER>_API_KEY=....
If you have npcsh installed, you can install Incognide directly to ~/.incognide using the following command within the npcsh environment:
npcsh> /incognidevThe StudioContext object is passed to every action handler, allowing developers to interact with the Studio's environment. Key capabilities include:
performSplit, closeContentPane, setRootLayoutNode, and setActiveContentPaneId allow for structural changes to the workspace.contentDataRef provides a mutable reference to the current content data, and updateContentPane allows refreshing content.handleAddTab, handleTabClose, handleTabSelect, and toggleZenMode manage the lifecycle and visibility of UI elements.generateId for creating unique identifiers and findPanePath for locating specific nodes within the layout tree.export interface StudioContext {
rootLayoutNode: any;
contentDataRef: React.MutableRefObject<Record<string, any>>;
activeContentPaneId: string;
setActiveContentPaneId: (id: string) => void;
setRootLayoutNode: (node: any) => void;
performSplit: (targetPath: number[], side: string, contentType: string, contentId: string, targetPaneId?: string) => void;
closeContentPane: (paneId: string, nodePath: number[]) => void;
updateContentPane: (paneId: string, contentType: string, contentId: string, skipMessageLoad?: boolean) => void;
handleAddTab?: (paneId: string, contentType: string) => void;
handleTabClose?: (paneId: string, tabIndex: number) => void;
handleTabSelect?: (paneId: string, tabIndex: number) => void;
toggleZenMode?: (paneId: string) => void;
generateId: () => string;
findPanePath: (node: any, paneId: string, path?: number[]) => number[] | null;
notifyPaneUpdate?: (paneId: string) => void;
windowId?: string;
currentPath?: string;
}