About Knowledge Documents Used in an Oracle Analytics AI Agent

Knowledge documents allow you to provide internal information such as process documentation, enterprise policies, competitive insights, market data, budgets, objectives, or other domain-specific knowledge.

Your content remains private within your Oracle Analytics instance and is used solely to enrich Oracle Analytics AI Assistant prompts so the agent can produce more precise answers. After you upload knowledge documents, they are parsed and vectorized. A query retrieves and adds only the most relevant portions to the prompt, ensuring that responses remain accurate and contextually aligned.

You must save the agent once before you can upload knowledge documents. As you upload documents, an indexing process is initiated. Your document is ready for use by the agent once its status on the list shows it as ready.

Text-heavy documents with clear punctuation are preferred. Avoid primarily tabular data, which is more difficult to split and vectorize effectively. Watch for conflicting information. If documents contain contradictory facts, for example, one states “return rate objective is 10%,” and another says “5%”, the RAG process may favor whichever information it retrieves first. To ensure reliable answers, avoid or reconcile conflicting information across your documents.

You can define a priority level, High, Medium, or Low, to each uploaded knowledge document. This enables more effective information management across multiple sources; High priority documents receive the greatest weight during the RAG process, while Regular and Low priority files are weighted accordingly in the Agent’s final responses.

You can use documents written in the following supported languages. English, Spanish, French, Arabic, Japanese, Korean, Portuguese, German, Russian, Dutch, Hungarian, Turkish, Polish, Italian, and Thai.