BlackCortex

BlackCortex Lite is a free, self-hosted AI chat application that runs entirely on the user's machine without cloud dependency. This solution executes language models on local hardware, accessible from any browser within the home network without installing software on client devices. The application supports local models through Ollama with compatibility for NVIDIA, AMD Vulkan, Intel Arc, and CPU-only operation, while integrating Google Gemini free tier as the sole cloud option. Core capabilities include PDF document chat, voice input, multi-user management with independent histories, and a model catalog featuring hardware compatibility scoring.

★★★★★

5.0(1 ratings)

• Local models via Ollama engine. BlackCortex Lite executes language models directly on user hardware utilizing the Ollama engine. GPU acceleration support includes NVIDIA cards, AMD with Vulkan, Intel Arc, and exclusive CPU operation. Each model downloads and stores locally without any data leaving the home network. This function removes dependency on external servers and guarantees complete conversation privacy.
• Google Gemini free tier integration. The application includes support for the free Google Gemini API, representing the only cloud service present in the Lite version. Users can switch between local models and Gemini according to their requirements while keeping API keys separate for each user account within the system.
• Multi-user access with personal history. Each person accessing BlackCortex Lite from their browser receives a dedicated space with independent conversation history and stored API keys. The system distinguishes between users without complex configurations, preserving privacy for each household member.
• PDF document chat functionality. This feature enables users to upload PDF files and ask questions about their content. The system extracts text from the PDF and combines it with the selected language model to generate contextually accurate responses based on document information.
• Voice input capability. Devices equipped with microphones can utilize voice input on both desktop computers and mobile devices. This function transcribes user speech into text for submission as a message to the AI model, facilitating interaction without typing.
• Model catalog with hardware compatibility scoring. The interface displays an available model catalog where each card shows a compatibility score based on detected hardware. Card colors indicate expected performance levels, helping users select the most suitable model for their system.
• PIN protection for model management. Model installation and deletion operations require a configurable PIN code. This measure prevents accidental or unauthorized changes to the available model set, maintaining system configuration integrity.
• Local network access without client installation. Any device connected to the same home network can access BlackCortex Lite through a web browser without additional software installation. The machine running the application acts as a server to which other devices connect, including phones, tablets, and laptop computers.
• Local storage using SQLite database. All information generated by the application, including conversations, user configurations, API keys, and document metadata, saves to a file named blackcortex_lite.db. This file resides exclusively on the machine running the program, enabling simple backups and complete data control.
• Execution via executable file on Windows. The current version for Windows 10 and 11 distributes as a single executable file. When launched, the application starts a local web server accessible through a browser without complex installation processes or system registry modifications.
• Environment warning for trusted home networks only. The application is designed exclusively for trusted home networks. It does not incorporate SSL encryption, role-based access control, or audit logging, therefore deployment in public or corporate environments receives no recommendation.

The development history of BlackCortex Lite begins with the first public version v1.0.0, aimed at providing a completely local AI chat solution for home networks. The developers created the application for households where multiple members want access to language models without relying on cloud services or subscriptions. The program is written primarily in the Go programming language, leveraging its concurrency performance and ease of deploying standalone executables. The web interface was built with HTML, JavaScript, and native CSS, while integration with Ollama and AI models occurs through standard APIs. The development team maintains the project as free software, with plans for future versions such as Lite+ that will include support for OpenAI, Claude, xAI Grok, DeepSeek, Qwen, web search, and document analysis.

Alternatives to BlackCortex: