Managing Workspace Memory & Knowledge Base
Learn how to manage documentation, upload PDFs, and ground your agents' memory.
An AI agent is only as smart as the data it can access. At Septra, we use a Retrieval-Augmented Generation (RAG) pipeline to ground our agents in your private business reality. The Knowledge Base is your workspace's institutional memory, serving as the central library of facts.
---
What is in the Knowledge Base?
Your agents search this repository before drafting any copy, analyzing budgets, or responding to chat queries. We recommend populating it with:
- Brand Guidelines: Voice tones, messaging frameworks, design guides, and writing style documents.
- Product Catalogs: Description lists, pricing tiers, specifications, and service agreements.
- Operational Guidelines: Standard Operating Procedures (SOPs), onboarding steps, and company wikis.
- Case Studies & Collateral: Past success stories, conversion templates, and sales presentations.
---
Adding Files to Your Memory
Adding documents is straightforward and supports multiple formats:
- Supported Formats: You can upload
.pdf,.md(Markdown),.txt, and.docxfiles. - Uploading: Navigate to the Knowledge Base section in the dashboard and drop files directly into the upload area.
- Automated Ingestion: Once uploaded, Septra's processing pipeline automatically reads the files, performs Semantic Chunking (breaking text into contextually complete ideas rather than arbitrary lines), and generates vector search indexes.
- Processing Time: The upload-to-active latency is under 5 minutes. Once indexed, your agents automatically gain access to the new facts.
---
Best Practices for Document Quality
To ensure your agents provide highly accurate answers, follow these guidelines when writing your company files:
- Avoid Ambiguity: Write clear, declarative sentences. Instead of saying "Our rates vary," write "Our standard consulting rate is $150 per hour."
- Use Structure: Use headings (
#,##) and bulleted lists. This helps our chunking engine cluster related information together. - Keep Files Updated: Regularly audit your documents. If your product specs change, overwrite the old file with the new specifications.
- Add Metadata: Use keywords or descriptive filenames (e.g.,
brand-voice-guide-2026.pdfinstead ofdoc123.pdf).
---
Auditing Agent Logic with Citations
Every time an agent uses a piece of information from your Knowledge Base, it grounds the answer by displaying a Citation.
Citations show the exact document name and text segment used, giving you absolute transparency over why the agent provided a specific answer. This ensures you can audit the AI's logic easily and catch any data gaps or outdated training materials.