Skip to main content

configuration-and-data

Create, validate, and troubleshoot AiZynthFinder configuration files, stocks, model assets, public-data setup commands, and data-tool commands.

Ir a la instalación

Datos de origen

Repositorio
VectorSpaceLab/AREX-Skill
Última actividad en el origen
26 de agosto de 2026 a las 16:31
Idioma detectado de SKILL.md
inglés
Estrellas
12
Forks
2

Opciones de instalación

De forma predeterminada está seleccionado el prompt que primero revisa el origen. Puedes cambiar a un comando directo o descargar una copia local.

Revisa los archivos de origen

Lee SKILL.md y los archivos complementarios que muestra SkillsMP antes de decidir si quieres instalarlo.

Explorador de archivos
6 archivos

Mostrando SKILL.md

SKILL.md
Instrucciones de origen · Vista previa de solo lectura
name
configuration-and-data
description
Create, validate, and troubleshoot AiZynthFinder configuration files, stocks, model assets, public-data setup commands, and data-tool commands.
disable-model-invocation
true
metadata
{"disco-role":"operating"}
license
MIT
# Configuration and Data Use this sub-skill when the task is about AiZynthFinder YAML configuration, data assets, stock files, policy/filter model references, or the bundled data-preparation CLIs. Stay in this sub-skill for safe static validation and setup guidance; route out before running retrosynthesis or interpreting output. ## Owns - Creating and reviewing YAML sections: `search`, `post_processing`, `expansion`, `filter`, `stock`, and optional `scorer`. - Explaining env-var interpolation with `${VAR}` and the failure mode for missing variables. - Choosing short-form versus full-form syntax for expansion policies, filter policies, and stocks. - Documenting search defaults, `algorithm_config`, bond controls, post-processing keys, and stock stop criteria. - Preparing asset guidance for model checkpoints, template tables, HDF5/CSV/text stocks, MongoDB stocks, and molbloom stocks. - Giving safe recipes for `smiles2stock` and `download_public_data` without executing downloads, Mongo writes, or bloom creation unless explicitly requested. - Running the bundled static validator script to catch schema-shape issues and missing local file references before a search run. ## Routes - Use [planning-workflows](../planning-workflows/SKILL.md) to execute `aizynthcli`, `aizynthapp`, or an actual retrosynthesis run from a config. - Use [route-analysis](../route-analysis/SKILL.md) to inspect outputs, route trees, statistics, clustering, or solved-route quality. - Use [extension-and-development](../extension-and-development/SKILL.md) to implement custom expansion strategies, stock query classes, scorers, filters, plugins, or custom search algorithms. ## References - [Configuration reference](references/configuration.md) covers YAML sections, defaults, env vars, policy/filter/stock syntax, and validation expectations. - [Stocks and assets](references/stocks-and-assets.md) covers model/template/stock file formats, optional dependencies, and service assumptions. - [CLI tools](references/cli-tools.md) covers safe `smiles2stock` and `download_public_data` recipes and when commands mutate external state. - [Troubleshooting](references/troubleshooting.md) covers common configuration, data, optional dependency, and service failures. ## Safe Validation Before running a user-provided config, run the bundled validator from any working directory: ```bash python skills/aizynthfinder/sub-skills/configuration-and-data/scripts/validate_aizynth_config.py path/to/config.yml --json ``` The validator intentionally does not import AiZynthFinder and does not load models, templates, stocks, MongoDB, TensorFlow servers, molbloom filters, or run tree search. Treat `errors` as blockers and `warnings` as setup issues to review with the user. ## Operating Rules - Keep public guidance asset-agnostic: use user-provided or project-relative paths, not local checkout paths or environment prefixes. - Do not claim MongoDB, molbloom, TensorFlow serving, public downloads, or route-distance features work until the user confirms the needed optional dependencies and services. - Do not execute `download_public_data`, Mongo writes, or molbloom creation without explicit user intent because they can use network, storage, or external services. - If a config validates statically but no expansion policy will be selected during planning, route to planning-workflows for the runtime selection step.
Ver en GitHub