Academy
5. Using Knowledge Bases Across the Platform
Connect a knowledge base to ChatPro, OUPI Studio, Deep Research and workflow RAG steps so the AI produces grounded, cited answers.
A knowledge base is a private, indexed collection of your documents. Once created and populated, it becomes a retrieval layer that any AI feature on the platform can query. When the AI answers, it searches your base, pulls the most relevant passages, and grounds its response in your actual content — complete with citations. This means the AI stops guessing and starts referencing. You can connect a knowledge base to four surfaces: the RAG selector in ChatPro's chat form, the RAG toggle in OUPI Studio, Deep Research queries, and dedicated RAG search steps inside workflows. Each surface works the same way: you point it at a base, and every answer becomes document-grounded.
In ChatPro, open the chat form and look for the RAG selector — pick the knowledge base you want the conversation grounded in. From that point, every message you send will trigger a retrieval step behind the scenes, and the AI's replies will cite passages from your documents.
In OUPI Studio, the mechanism is even simpler: enable the RAG toggle on your agent, select the base, and the agent automatically searches it before responding.
For Deep Research, the platform queries your base as part of its broader research pipeline, ensuring your proprietary documents are included alongside any other sources.
In workflows, add a RAG search step and configure it to target your base. The retrieved passages feed into downstream steps, so automated pipelines also produce grounded, cited outputs.
Keep one knowledge base per topic or project. Retrieval is sharper when a base is focused — a catch-all base with mixed subjects returns noisier, less relevant passages. For example, create separate bases for 'HR Policies', 'Product Documentation', and 'Legal Templates'.
Before connecting a base anywhere, verify it actually returns good results. Every knowledge base page includes a built-in search field: type a realistic question (not just keywords) and inspect the passages that come back. If the results miss the mark, you have two levers: add more relevant documents to fill gaps, or revisit your chunking settings — increase chunk size for long structured texts, decrease it for short Q&A-style content. This test-and-tune loop is the fastest way to improve answer quality across every surface that uses the base.
Share a knowledge base with your team so everyone grounds the AI in the same verified information. Open the share dialog on the base page and select team members — they can then use it in their own ChatPro conversations, Studio agents, and workflows without duplicating documents.
Go to your Knowledge Bases page, open an existing base (or create one), and use the built-in search field to test a real question. Then open ChatPro, select that base in the RAG selector, and ask the same question. Compare the cited passages in the AI's answer with the search results you saw on the base page.
You can connect a knowledge base to four surfaces: ChatPro (RAG selector), OUPI Studio (RAG toggle), Deep Research, and workflow RAG search steps. Each makes the AI ground its answers in your documents with citations. Keep bases topic-focused for sharper retrieval, test with the built-in search field before going live, and share bases with teammates so everyone works from the same verified knowledge. Indexing consumes embedding credits proportional to content size, while searching is lightweight.