A lightweight, CPU-only AI assistant to help novice users download and configure a small fast local AI on their system. This is designed as a helper app to be included with other apps to guide users through possibly complex situations it has limited tooling
BootstrapAI provides a simple terminal interface that guides users through:
- Downloading small, CPU-friendly AI models
- Configuring the inference engine
- Chatting with the local model — no cloud, no GPU required
┌──────────────┐ ┌──────────────────┐ ┌───────────────┐
│ ConsoleUI │────▶│ Application │────▶│ IInference │
│ (CLI menu) │ │ (orchestrator) │ │ Engine (IF) │
└──────────────┘ └──────────────────┘ └───────┬───────┘
│ │ │
▼ ▼ ▼
┌──────────────┐ ┌──────────────────┐ ┌───────────────┐
│ DownloadMgmt │ │ ConfigManager │ │ LiggufEngine │
│ (libcurl) │ │ (JSON on disk) │ │ (CPU-only) │
└──────────────┘ └──────────────────┘ └───────────────┘
The inference engine is behind the IInferenceEngine interface so it can be swapped out easily — BootstrapAI is not tied to ligguf forever.
- C++17 compiler (GCC 9+, Clang 10+)
- CMake ≥ 3.16
- libcurl (for model downloads)
- nlohmann/json (header-only, for config)
- Google Test (for unit tests, optional)
# Configure
cmake -B build
# Build
cmake --build build
# Run
./build/BootstrapAIcmake --build build && cd build && ctestEarly development. See project.ai for the task list.