While your slides provide the primary narrative, attendees will inevitably ask questions that aren't covered on the screen. The Knowledge Base (KB) gives Aura the context needed to answer them intelligently.
How it Works
When you upload documents to the Knowledge Base, Seminara chunks the text and stores it in a secure Vector Store. When an attendee asks a question, our Intent Engine performs a semantic search (RAG - Retrieval-Augmented Generation) against this vector store to fetch relevant facts before generating Aura's response.
Add to Knowledge Base
You can manage your KB by clicking Add to Knowledge Base in the dashboard. You can add context in three ways:
- File: Upload PDF or TXT files.
- Text: Paste raw text directly into the dashboard.
- URL: Provide a website link and we will automatically scrape the article.
- Best Practices: Upload your company FAQ, technical data sheets, or standard objection-handling scripts.
Boundaries and Hallucinations
Aura is strictly instructed via our internal System Prompts to only answer questions based on the provided slides and the Knowledge Base. If an attendee asks an irrelevant question (e.g., "What is the capital of France?"), Aura will gracefully deflect and steer the conversation back to the presentation.
Frequently Asked Questions (FAQ)
Q: Is my Knowledge Base data used to train public AI models?
A: Absolutely not. Your data is isolated to your organization and is never used to train foundational LLMs.
Q: What is the maximum file size I can upload to the KB?
A: You can upload PDF or TXT files up to 50MB per document. For larger datasets, we recommend breaking them into smaller, topic-specific files.
Q: How long does it take for Aura to process new KB documents?
A: Document processing and vectorization happen in the background and typically complete within 30 seconds. You'll see a "Ready" badge in your dashboard when it's done.
Q: Can I prioritize certain KB documents over others?
A: Yes. You can pin essential documents in the KB manager. The semantic search algorithm assigns a higher weight to pinned documents when fetching context for Aura's answers.