352 task recipes. Each guide is designed around one job done end to end: steps, verification, and the failures that actually bite when you run open-weight AI on your own hardware.
Pull with Ollama and verify a local response.
Reduce context instead of guessing at random settings.
Turn your setup into comparable local evidence.
Make Ollama reachable for apps and tools.
Showing the first 48 matches. Tighten a filter to narrow the list.
Get a runtime installed and a first model answering.
Pick, quantize, load, and serve models; tune throughput.
Retrieval, tool-calling, and agent loops on local models.
Day-to-day operator workflows at the terminal.
Keep it up: observability, reliability, cost control.
Focused extras that don't fit a single domain.