A Resilient RAGOps Architecture for Chatbots over Knowledge Bases Derived from Multimodal Artifacts

• Márcia de Sousa Instituto Atlântico
• Gabriel Oliveira Instituto Atlântico
• Lucas Alves Instituto Atlântico
• Cleilton Rocha Instituto Atlântico

DOI: https://doi.org/10.5753/eniac.2026.30106

Resumo

Retrieval-Augmented Generation (RAG) chatbots rely on operational pipelines to manage knowledge in production, but building reliable knowledge bases from multimodal artifacts and recovering them after failures remain open RAGOps challenges. This paper presents a resilient RAGOps architecture for constructing and restoring such knowledge bases. It combines artifact processing, provenance tracking, embedding versioning, and rollback to transform heterogeneous artifacts into chunks and embeddings while preserving the knowledge lifecycle. Operationally instantiated, the architecture was evaluated through answer quality and resilience protocols. The stable knowledge base yielded an average BERTScore of 0.7860, which decreased to 0.7445 under the contamination scenario (approximately 5.3%) and reached 0.7945 after rollback, a level comparable to the original stable state. These results provide initial evidence of the architecture’s ability to support controlled knowledge-base restoration.

Referências

Abootorabi, M. M., Zobeiri, A., Dehghani, M., Mohammadkhani, M., Mohammadi, B., Ghahroodi, O., Baghshah, M. S., and Asgari, E. (2025). Ask in any modality: A comprehensive survey on multimodal retrieval-augmented generation. Findings of the Association for Computational Linguistics: ACL 2025, pages 16776–16809.

Badgujar, P. (2026). Self-healing data pipelines: Leveraging ai to detect and correct failures in real-time. International Journal of Artificial Intelligence, Data Science, and Machine Learning, 7(1):87–90.

Badillo-Rangel, E. (2024). Enhancing llm performance in specialized spanish domains using rag and peft qlora.

Barnett, S., Kurniawan, S., Thudumu, S., Brannelly, Z., and Abdelrazek, M. (2024). Seven failure points when engineering a retrieval augmented generation system.

De Florio, V. (2013). On the constituent attributes of software and organizational resilience. Interdisciplinary Science Reviews, 38(2):122–148.

Jiao, S., Huang, C., Qi, S., Wang, X., Li, Y., and Yao, L. (2026). Doctor-rag: Failure-aware repair for agentic retrieval-augmented generation. arXiv preprint arXiv:2604.00865.

Lavie, A., Sagae, K., and Jayaraman, S. (2004). The significance of recall in automatic metrics for mt evaluation. In Conference of the Association for Machine Translation in the Americas, pages 134–143. Springer.

Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Küttler, H., Lewis, M., tau Yih, W., Rocktäschel, T., Riedel, S., and Kiela, D. (2021). Retrieval-augmented generation for knowledge-intensive nlp tasks.

Li, S., Zhang, H., Wang, Y., Zhou, M., Ma, Z., and Liang, F. (2025). Universal end-to-end automated evaluation framework for retrieval-augmented generation system. In 2025 IEEE 8th International Conference on Pattern Recognition and Artificial Intelligence (PRAI), pages 867–871. IEEE.

Lin, C.-Y. (2004). Rouge: A package for automatic evaluation of summaries. In Text summarization branches out, pages 74–81.

Meyer, J. F. (2009). Defining and evaluating resilience: A performability perspective. In Proceedings of the International Workshop on Performability Modeling of Computer and Communication Systems (PMCCS), volume 1, page 561.

Papineni, K., Roukos, S., Ward, T., and Zhu, W.-J. (2002). Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th annual meeting of the Association for Computational Linguistics, pages 311–318.

Prajapati, T. (2025). Livevectorlake: A real-time versioned knowledge base architecture for streaming vector updates and temporal retrieval.

Salemi, A. and Zamani, H. (2024). Evaluating retrieval quality in retrieval-augmented generation. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 2395–2400.

Serafim, L. M. et al. (2025). Do humano aos llms: Apontamentos sobre o processamento de linguagem.

Steelman Jr, R. L. (2025). Medallion architecture. In Mastering Snowflake DataOps with DataOps. live: An End-to-End Guide to Modern Data Management, pages 247–264. Springer.

Sun, T., Somalwar, A., and Chan, H. (2024). Multimodal retrieval augmented generation evaluation benchmark. In 2024 Ieee 99th vehicular technology conference (vtc2024-Spring), pages 1–5. IEEE.

Tu, Y., Su, W., Zhou, Y., Liu, Y., and Ai, Q. (2025). Robust fine-tuning for retrieval augmented generation against retrieval defects. In Proceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR ’25, page 1272–1282, New York, NY, USA. Association for Computing Machinery.

Xu, X., Weytjens, H., Zhang, D., Lu, Q., Weber, I., and Zhu, L. (2025). Ragops: Operating and managing retrieval-augmented generation pipelines.

Zhang*, T., Kishore*, V., Wu*, F., Weinberger, K. Q., and Artzi, Y. (2020). Bertscore: Evaluating text generation with bert. In International Conference on Learning Representations.

Zhou, R. (2024). Advanced embedding techniques in multimodal retrieval augmented generation a comprehensive study on cross modal ai applications. Journal of Computing and Electronic Information Management, 13(3):16–22.