Przejdź do treści
Biblioteka skilli
Nauka i badania · Analiza

fluent-model-selection

/fluent-model-selection

Dobiera modele fizyczne, solver i role warunków brzegowych dla przypadku Fluent, a każdy wybór uzasadnia numerem reguły i stanem dowodów.

Licencja
MIT
Ostatnia zmiana repozytorium
12.11.2026

Ocena i bezpieczeństwo

90/ 100Klasa: dobry
Format25 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność25 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść25 / 35
Co najmniej 150 słów i do 500 linii, nagłówki, listy kroków, przykłady albo kod.
Higiena15 / 15
Bez zaślepek typu TODO, bez ścieżek Windows, z licencją.
Ryzykobrak sygnałów w skanie
Pliki w paczce
1
Skrypty
0
Komentarze HTML
0
Adresy URL w SKILL.md
0
Przestarzałość

Brak sygnałów. Repozytorium zmienione 12.11.2026, bez odwołań do wycofanych modeli i API.

Licencja

MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.

Co robi i kiedy

Dobiera modele fizyczne, solver i role warunków brzegowych dla przypadku Fluent, a każdy wybór uzasadnia numerem reguły i stanem dowodów.

Oryginalny opis (EN)

Decides Fluent models, solver and boundary-condition roles for a case and justifies each choice with a rule id and an evidence status. Use when physics must be chosen, or the user must be interviewed for a case spec. Not for TUI syntax, journal text, or judging a journal.

Jak działa

  1. When to use
  2. When not to use
  3. Inputs
  4. Procedure
  5. What the statuses mean
  6. Outputs
  7. Refusals
  8. Hand-off
Sekcje z nagłówków SKILL.md, w oryginale · 110 linii, 987 słów
Podgląd SKILL.md
---
name: fluent-model-selection
description: Decides Fluent models, solver and boundary-condition roles for a case and justifies each choice with a rule id and an evidence status. Use when physics must be chosen, or the user must be interviewed for a case spec. Not for TUI syntax, journal text, or judging a journal.
license: MIT
compatibility: Node ≥ 22.19 (bundled with Kady). No network. No Fluent required for this skill.
metadata:
  version: 0.1.0
  pack: fluent-skill-pack
  level: L1
---

# Fluent model selection (L1 decisions)

This skill turns a case description into **engineering decisions**: which physics are active, which
turbulence and solver settings apply, and *why*. It is the L0 router's second stage: it reads the
case spec, applies the pack's rule base, interviews the user for what is still unknown, and hands
each decision back with the rule that produced it, its status (`verified` / `derived` /
`unverified`) and its source. Judgement lives here; the arithmetic lives in the pack's CLIs.

`<home>` below is the pack home directory (`home.json` lives there). Resolve it once from
`$KADY_FLUENT_HOME`, or accept `~/.kady/fluent`; if no home is found the CLIs exit 3 and tell you
to run `install.mjs`.

## When to use

- A case spec exists (or is being assembled) and the user asks what models to run, why this
  turbulence model, whether the flow is compressible, which solver, or how to converge it.
- The user answers some case questions but not all, and you must ask the rest without inventing
  values.
- You need the rationale record that the journal stage and the reviewer will audit: rule ids,
  statuses, sources, and explicit warnings.

## When not to use

- **Command syntax.** Which keystrokes enable k-omega SST, the exact menu spelling, whether a TUI
  path was ever exercised: that is `fluent-tui-knowledge`'s job. This skill names the model; it
  does not guarantee the TUI path.
- **Writing the journal.** Turning decisions into ordered journal lines is
  `fluent-journal-authoring`'s job. This skill never emits journal lines.
- **Judging a journal.** Independent validation belongs to `fluent-journal-validation`, which must
  not be the author.
- Anything needing a Fluent license: this skill is entirely offline.

## Inputs

- `case-spec.json` — a case spec as described in `<home>/data/docs/CASE_SPEC.md`. Only `meta` is
  required; anything absent is treated as unknown, never as a default.
- Optional `base.json` — an existing spec to merge answers over.
- Interview answers keyed by question id or by spec path (from Kady's `interview` tool).
- The pack's own data: `physics-rules.json` (23 rules) and `questions.json` (70 questions). Never
  edit data or rules from this skill; report a wrong rule instead.

## Procedure

1. Inspect the spec: `node <home>/bin/advise.mjs --spec case-spec.json`. This returns
   `recommendations[]` (one entry per recommended spec path, with `id`, `why`, `status`,
   `confidence`, `source`, `recipes[]`), `applied[]`, `skipped[]`, `complete` and `missing[]`.
2. Read `skipped[]` before trusting `complete`. A rule skipped with `no-value:` means the spec lacks
   the field the rule keys on — ask for it or flag it; do not guess the answer. A `mismatch:` skip
   means the rule is genuinely inapplicable. Never silently drop a rule.
3. Get the next questions:
   `node <home>/bin/questions.mjs --spec case-spec.json --limit 5`. Each item carries an
   `interview` object already shaped for Kady's native `interview` tool (enum → `single` with
   options and a recommended value, boolean → yes/no, number → text). Copy
   `questions[].interview` into the `questions` array you pass to `interview`; keep the ids so the
   answers can be fed back. Ask only what blocks the case: `weight: "critical"` items first.
4. Turn answers into a spec:
   `node <home>/bin/spec-build.mjs --answers answers.json --base case-spec.json`. Answers may be
   keyed by question id (`task.dimension`) or by spec path (`fluent.dimension`). The CLI maps each
   answer to its declared `specPath`, coerces it to the question's type, and reports
   `unsupported[]` (unknown key, wrong type, refused overwrite) and `missing[]` (required answers
   still absent). Nothing is invented: an unanswered field stays absent.
5. Re-run `advise.mjs` on the updated spec until `missing` is empty or the remaining gaps are
   deliberate. Then reconcile the decisions with `recommendations[]`: each chosen model must trace
   to a rule id, or be an explicit user choice recorded as such.
6. Report `decisions[]` — one entry per decision: `specPath`, chosen `value`, `ruleId` (or
   `"user"`), `why`, `status`, `confidence`, `source`, and any `recipes[]` worth citing. The caller
   persists them; this skill writes no files.
7. Hand the decisions and their recipes to `fluent-journal-authoring`, then to
   `fluent-journal-validation`. Do not write or judge journals yourself.

## What the statuses mean

- `verified` — observed in a live Fluent session and recorded with its source. Safe to cite.
- `derived` — inferred from the knowledge base (docs, QA logs) plus verified facts. State it as a
  recommendation, not as observed behaviour.
- `unverified` — the pack has not confirmed this path. Say so; suggest a live check.

Never upgrade a status. A `derived` rule stays `derived` in your report even if you agree with it.

## Outputs

- `recommendations[]` / `decisions[]` with rule ids, statuses, sources and recipes — the rationale
  record the router stores.
- The next `interview` questions, with recommended values clearly marked as suggestions.
- An updated spec plus two honest lists: what is `unsupported` and what is `missing`.
- Warnings: conflicting recommendations for one path, unevaluable rules, shape mismatches between
  the documented spec and the interview paths.

## Refusals

- No home / no data → exit 3 with the fix (`install.mjs` or `KADY_FLUENT_HOME`). Do not fake data.
- A rule whose condition cannot be evaluated is reported in `skipped` with its reason, never
  treated as applied or as inapplicable.
- `recommend.actions[]` entries are **recipe ids**, not spec fields. Never write them into the
  spec; pass them to journal authoring or cite them.
- `--profile <id>` when the profile file is absent → exit 3 naming the index entry and the paths
  looked for. Do not reconstruct the profile by hand.
- A recommendation never becomes a fact: if the user has not confirmed the physical parameters
  (Mach, Re, units, temperature), the spec stays incomplete.
- Never raise a `status`, never invent a `source`, never claim a TUI path is verified.

## Hand-off

- TUI syntax, command spelling and live-session checks → `fluent-tui-knowledge`.
- Journal lines from these decisions → `fluent-journal-authoring` (only `verified`/`derived`
  recipes may become lines; `unverified` remains a gap).
- Independent review → `fluent-journal-validation`.
- Answer the user in Russian; keep this skill's own artifacts (specs, JSON) in English.

Wersje

W bazie nie ma innych skilli o nazwie fluent-model-selection.

Migawka bazy z 13.11.2026 · wersja paczki 133d894cc595 (skrót SHA-256 całej paczki)

Pobrania w ostatnich 30 dniach: 0 · ranking

Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.

Podobne skille

  • 90fluent-journal-authoring
    Buduje plik dziennika Fluent z zatwierdzonej specyfikacji przypadku, używając tylko sprawdzonych przepisów, i wypisuje kroki, których nie da się wygenerować.
  • 90fluent-case-report
    Składa istniejące materiały przypadku CFD w raport z sumami sha256 i listą otwartych braków, gdy potrzebne jest czytelne podsumowanie dla człowieka.
  • 95ansys-fluent-mcp
    Steruje sesją solvera ANSYS Fluent przez serwer MCP: wczytuje przypadek i siatkę, zmienia ustawienia, uruchamia iteracje i odczytuje wyniki.
  • 85fluent-cfd
    Prowadzi symulacje CFD w Ansys Fluent i PyFluent: konfigurację siatki, warunków brzegowych i turbulencji, zbieżność oraz walidację wyników.
  • 90CFD Fluent Pipeline
    Prowadzi audytowalną symulację CFD w Ansys Fluent i PyFluent: siatkę, warunki brzegowe, model turbulencji, zbieżność i walidację wyników.
  • 96fluent-cli
    Pomaga budować aplikacje i pobierać dane w ServiceNow przez Fluent (now-sdk).

Uwagi i zgłoszenia

Wiadomość trafi do Bartosza Sokolińskiego. Adres e-mail jest potrzebny tylko do odpowiedzi. Szczegóły w polityce prywatności.