veille-ia-gen/veille-ia_ynh/doc/DESCRIPTION.md at b3ff529cfa8f3b7e59cbd16871568541fa153308

Files

T

cyber-mawonaj b3ff529cfa feat(packaging) : package YunoHost veille-ia_ynh (packaging v2, helpers 2.1)

- manifest.toml validé contre le schéma officiel manifest.v2.schema.json
  (sources par tag forge avec placeholder sha256 assumé, nodejs 24, ports,
  system_user, install_dir, data_dir, permission privée all_users, SSO)
- Scripts install/remove/upgrade/backup/restore/change_url (bash -n OK) :
  build au fil de l'eau avec BASE_PATH, ALERTS_TOKEN généré à l'install,
  .env chmod 400, logs journald
- conf/nginx.conf : proxy_params_with_auth (header Ynh-User), conf/systemd.service :
  node via __NODEJS_DIR__, sandboxing modéré, conf/env : DATA_DIR/OLLAMA_URL/ALERTS_TOKEN
- tests.toml minimal, doc DESCRIPTION/ADMIN fr+en
- README racine : dev local, variables d'env, procédure de release, Ollama

2026-08-01 09:25:39 -04:00

647 B

Raw Blame History

Veille IA is a personal watchtower for generative AI models: releases, licenses, rankings, with a recommendation engine based on project profiles.

Model registry: single source of truth (git-versioned YAML), anti-hallucination rule (unverified field = "to be checked")
Project profiles: legal constraints per production context
Recommendation engine: ranked, justified shortlist with warnings
Alert inbox: events classified by a local LLM (Ollama), pushed from Node-RED
Markdown export: guides and rankings regenerated from the registry

Private application (YunoHost SSO), 100% open source (AGPL-3.0), no paid API.

647 B Raw Blame History

647 B

Raw Blame History