Home

AIBrain documentation

AIBrain is a local Windows research interface for GGUF language models already installed by Ollama. It provides streaming chat and a native OpenGL connectome driven by real-time llama.cpp inference telemetry.

This documentation describes the software as implemented. Real-time means the active backend measures raw next-token logit statistics, context position, stream timing, output progress, and recent output repetition while a response is generated. The direct GGUF path does not expose hidden states, attention matrices, MLP activations, or literal model-neuron activity. The graph is a display layout for the measured channels, not a reconstruction of model topology.

Start here

1. Getting started — requirements, managed installation, virtual-environment setup, and first launch.
2. User guide — chat, playback, the visualizer, and controls.
3. Models and inference — Ollama discovery, GGUF validation, local loading, and generation settings.

Reference

• Connectome and analysis
• JSON reference
• Configuration
• Native acceleration
• Architecture
• Troubleshooting
• Wiki publishing

Scope

AIBrain is intentionally a local, text-only application. It does not provide cloud APIs, image chat, tools, retrieval augmentation, or a browser-based UI. The loaded GGUF must be compatible with the installed llama-cpp-python backend.