Pavelevich/llm-checker v3.4.0

on GitHub

v3.4.0 — Ollama Runtime Capacity Planner

Highlights

• Added new ollama-plan command to compute safe Ollama runtime settings from local models + detected hardware.
• Added planner recommendations for:
  • OLLAMA_NUM_CTX
  • OLLAMA_NUM_PARALLEL
  • OLLAMA_MAX_LOADED_MODELS
  • queue / keep-alive / flash-attention env vars
• Added fallback profile + memory risk scoring for safer defaults under pressure.
• Added planner tests and CLI smoke coverage for ollama-plan --help.

Validation

• Full suite passing: 18/18 tests.

Notes

• npm publish attempted for 3.4.0; account requires OTP-protected publish.