Local AI glossary
42 pages. Plain definitions, each followed by why the term matters when the model is running on your own machine.
- Air-gapped
- Evaluation (eval)
- Chunking
- Context overflow
- Context window
- CUDA
- Do Not Track
- EasyList
- EasyPrivacy
- Edge inference
- Embedding
- Fingerprinting
- GGUF
- GPU layers
- Hallucination
- Inference
- KV cache
- llama.cpp
- Local LLM
- Mixture of experts
- Model card
- NPU
- On-device AI
- Parameter count
- Perplexity
- Prompt injection
- Quantisation
- RAG (retrieval-augmented generation)
- Safetensors
- Sandboxing
- Sec-GPC (Global Privacy Control)
- System prompt
- Telemetry
- Temperature
- Time to first token
- Tokenizer
- Tokens per second
- Unified memory
- VRAM
- Vulkan
- WebView2
- Zero egress