NGIBS is powered by an advanced open-source stack designed for zero-latency, local AI execution and deep web reasoning.
At its heart, NGIBS is not just a software application; it is an orchestrator of modern artificial intelligence frameworks. We leverage several key technologies to make local reasoning possible:
The primary language driving NGIBS. It acts as the glue that binds the user interface, web scraping modules, and AI inference engines together.
Learn about PythonWhile the logic is written in Python, the raw neural network inference relies heavily on optimized C++ (via llama.cpp) to execute billion-parameter LLMs fast on standard CPU/GPUs.
Explore llama.cppThe core user interface is built natively using PyQt6. This ensures deep OS integration, cross-platform stability (Windows/Linux/Mac), and a highly responsive desktop experience.
Learn about PyQt6We use Ollama as our local inference server. NGIBS talks directly to the local API, allowing seamless hot-swapping, pulling, and deletion of GGUF models like Llama 3 and Mistral.
Visit OllamaComplex reasoning loops, tool-calling, and semantic memory retrieval are orchestrated by LangChain. Context vectors are stored securely in a local ChromaDB instance.
For Live Search, NGIBS uses the DuckDuckGo API and Wikipedia modules to fetch URLs, then aggressively parses DOM trees via BeautifulSoup to inject pure text into the context.