THE TOKFIRE FIELD GUIDE
Local AI, put to work.
Six practical guides for the hardware on your desk. Build something useful, understand the technology, and measure what actually works.
06PC / MAC
LOCAL LLM
LOCAL LLM
Your first local AI assistant: start small, make it useful
Turn a PC or Mac into a practical drafting and summarising assistant with Ollama.
Read the guideAsk your own documents: a small, local RAG tutorial
Build a document-questioning prototype with evidence you can inspect—without a cloud vector database.
Read the guideRun a local LLM on Apple silicon with MLX
Start with a 4-bit model, understand unified memory, and test whether prompt caching helps your workload.
Read the guideCan a 16GB GPU run a 27B model? Build a memory budget
What GSQ + RCO changes, what KV cache still costs, and how to test a quantized GGUF honestly.
Read the guideFrom chat to action: build a small local tool agent
A read-only stock lookup shows what an agent actually does—and how to check whether it did the job.
Read the guideOne model, three jobs: measure useful local AI speed
Separate throughput from responsiveness, then test batching, caching and speculative decoding fairly.
Read the guide