llmfit
Installation
SKILL.md
llmfit operator
When to use this skill
Use this skill for source-grounded, node-local model-fit decisions: hardware inventory, fit verdicts, score/ranking output, hardware profiles, plan and storage estimates, and read-only REST queries. It keeps formula-based estimates separate from measured inference.
Route generic service or inference performance tuning to performance-optimization. This checkout has no separate lmstudio-cli or typesafe-ai skill; treat those names only as llmfit provider/API values and use the pinned reference when needed.
Instructions
- Run the offline gate first. From the repository root run
python3 .agent-skills/llmfit/scripts/preflight.py --json; from an installed skill root runpython3 "$SKILLS_ROOT/llmfit/scripts/preflight.py" --json. Add--repo PATH --expect-commit FULL_SHAonly when checking a supplied source checkout. The helper never starts llmfit or invokes network/provider work. - Establish binary and scope. Run
llmfit --versionandllmfit --help; if unavailable, stop at an install/source-build handoff. Do not run an installer, package manager,cargo build, container pull, or model download without explicit approval. - Capture read-only hardware evidence. Run
llmfit system --json; usellmfit doctoronly for a diagnostic request. Load references/upstream.md for the pinned platform probe and capability matrix. - Run a static fit query. Prefer
llmfit recommend --json --limit N(optionally a supported--use-case) orllmfit fit --json -n N; usellmfit info "MODEL"for one row. Keep tok/s, prefill, and TTFT labeled as estimates unless output explicitly reports measured confidence. - Apply the source thresholds.
memory_required / memory_availablegives Perfect at ≤0.60, Good at ≤0.85, Marginal at ≤0.98, and Too Tight above 0.98.recommended_ram_gbdoes not alter the verdict; offload paths cap at Good, and Too Tight is not a runnable claim. Consult the reference for estimation caveats. - Compare capacity and plans consistently. Use global capacity overrides before
fit, or a validated hardware profile for speed-relevant simulation. For a named model usellmfit plan "MODEL" --context 8192 --json, keeping quantization/context fixed when comparingplanandfit. Storage output is capacity arithmetic, not free-disk evidence. - Keep API and side effects distinct. A loopback
serveplus GET routes is read-only after an approved launch;/api/v1/planis still an estimate. Downloads, TUI/provider pulls,llmfit bench, credentials/network, and--sharePR creation require separate approval; use the pinned dry-run command for share payload preview. - Return an evidence-backed conclusion. Include exact commands, host/profile/override inputs, model/quant/context, fit level/run mode, estimate basis/confidence, and whether a measured run occurred. Never report estimated throughput as a benchmark.