Project Knowledge
How to Structure Good Knowledge
The quality of your agent’s responses depends directly on how well your project knowledge is written. Even powerful AI agents won’t perform well if the content they’re trained on is unclear, bloated, or outdated. This guide shows you how to write, structure, and organize your GRUP project content for maximum clarity, performance, and brand alignment.
Principles of Good Knowledge
1. Be Clear and Concise
Avoid long-winded marketing phrases or jargon
Write in simple, direct language
Focus on facts, functions, and user-facing outcomes
2. Use Bullet Points
Bullet points are easier for agents to parse
Use them for features, steps, lists, and processes
Example:
Instead of: “Our staking system rewards loyal holders with multipliers and boosts across the ecosystem.”
Use:
Users earn rewards for staking
Staking increases based on holding duration
Boosts apply to in-game perks and governance weight
3. Organize by Topic
Break your content into labeled sections:
Overview
Token Utility
Product Features
FAQ
Roadmap
Links
This structure helps the agent respond more accurately and consistently.
4. Prioritize What Matters Most
Put high-priority content first. If there’s too much content, agents may not always “see” deeper sections.
Lead with your most asked questions
Explain what your product does in the first 2-3 lines
Link to external docs where more depth is needed
5. Keep It Up to Date
Outdated info leads to wrong replies. Make it a habit to:
Update your roadmap section after releases
Revise tokenomics if your model changes
Remove links to deprecated platforms or tools
Formats You Can Use in GRUP
You can input knowledge in multiple formats:
Plain text (recommended)
Markdown with headers, bold text, and bullets
File uploads (PDFs, .txt, .docx)
URLs to external content or docs
What to Avoid
Walls of unformatted text
Repetitive or vague marketing phrases
Over-explaining without actionable info
Technical terms without explanation
Any other questions? Get in touch