The technological solutions in this context are primarily based on Large Language Models (LLM) and approaches such as prompt engineering, fine-tuning, and Retrieval-Augmented Generation (RAG). LLMs are large language models designed to generate responses to user queries. These models can perform this task thanks to the vast amount of data they are trained on. To tailor LLMs to the specific needs of an organization, the RAG approach is used. This method involves providing the model not only with the user's request but also with a set of relevant documents specific to the domain of interest, which can include documents from the organization itself. These documents serve as an additional data source, allowing the model to extract and utilize detailed and targeted knowledge to respond more accurately to the received requests. For adapting LLMs to more detailed tasks, fine-tuning is employed, which involves retraining the model using data specific to the application domain.
The solutions described so far can be easily implemented using Oracle services like OCI Generative AI service. This service integrates LLMs based on Cohere and Meta Llama 2 models. The Oracle environment also provides the capability to customize these models (fine-tuning) and integrate them into enterprise solutions that involve using a RAG system. All of this is supported by dedicated Oracle GPU clusters that ensure privacy, reliability, and security.
Some contexts where AIOps solutions can be adopted include:
- Service desk support: A chatbot capable of synthesizing, categorizing, and proposing solutions within a ticket-based IT support system. The chatbot acts as an AI agent to streamline data processing and assist human IT experts in managing and resolving operational tickets.
- Event and log analysis: A solution capable of monitoring the operational status of an IT system, analyzing logs, detecting unexpected behaviors, and autonomously implementing remediation solutions. Additionally, human IT experts can interact with an integrated chatbot to obtain synthesized information about logs and the monitored system's status.
- IT documentation generation: A system capable of analyzing and documenting the code of a software product or IT infrastructure. This system extracts knowledge and contextual information that can support IT experts during code maintenance, system migrations, and improvement