Communication Dans Un Congrès Année : 2025

A retrieval-augmented-generation pipeline to help users query system-provided documentation

Résumé

The increasing integration of AI into computing workflows demands a re-evaluation of traditional operating system design. In environments like Debian, users are often faced with a vast ecosystem of command-line tools, each accompanied by extensive manual pages (man pages) detailing usage, flags, and parameters. While comprehensive, these documents are frequently dense, verbose, and not well-suited for rapid onboarding or targeted queries. We propose a Retrieval Augmented Generation (RAG) pipeline to bridge this gap, enabling natural language interaction with system documentation. By combining tokenization, embedding, and dense retrieval with a language generation model, our system allows users to query tool usage in plain language and receive concise, contextually relevant responses. This approach streamlines tool discovery and comprehension, and represents a step toward more intelligent, user-aware operating systems.

Fichier principal
Vignette du fichier
hotcrp-final2.pdf (335.2 Ko) Télécharger le fichier
80-a-retrieval-augmented-generation-pipeline-to-help-users-query-system-provided-documentation.pdf (940.9 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
licence

Dates et versions

hal-05333737 , version 1 (27-10-2025)

Licence

Identifiants

  • HAL Id : hal-05333737 , version 1

Citer

Tzolkin Garduno-Alvarado, Gunnar Wolf, Eddie Soulier, Francis Rousseaux, Feliu Sagols. A retrieval-augmented-generation pipeline to help users query system-provided documentation. DebConf25, IRISA, Jul 2025, Brest, France. pp.2. ⟨hal-05333737⟩
1363 Consultations
169 Téléchargements

Partager

  • More