Project knowledge: Keeping context in one place

Project knowledge is where you set up all the context that makes your Project or assistant actually useful. When you create a Project, the knowledge you add is the backbone. It's what Juma references every time someone starts a chat inside that Project.

What you can add to Project knowledge

Info

This is the big system prompt, the backbone of your Project. Here you add all instructions, persona, goals, rules, and requirements. Everything the AI needs to know about how to behave in this Project goes here.

You can write this yourself, or you can use the magic wand feature. Just write a couple of sentences about what you want the assistant to do, click the magic wand button, and Juma will gather workspace analytics, analyze the context and structure, and create a detailed system prompt for you. You can then edit, add, or remove things before saving.

Data sources

This is where you add the files, pages, and connections that give your Project the information it needs.

Files and images: Upload brand guidelines, target audience profiles, customer data, templates, and anything else the AI needs to reference.

Supported document types: pdf, doc, docx, xls, xlsx, pptx, txt, htm, html, csv, json, md, cpp, hpp, c

Supported image types: jpg, jpeg, png, webp, svg, heic, heif, gif

Integrations: Connect and upload context directly from Notion, Google Drive, SharePoint, OneDrive, or HubSpot. If you connect files through integrations, you can click the refresh button to automatically update them when the source changes.

Web pages: Paste a URL and it will be saved as context for the Project.

Chats: If you've had a really good brainstorming session in a chat, you can add that whole chat as context for the AI to keep in mind.

Brand voice

Add your voice, tone, and style guidelines here. This keeps everything on brand across all chats in the Project.

Examples

The AI works best when you show it what good looks like. Add examples of ideal outputs here. This will drastically improve the quality of the AI's responses. If you show the AI model what good looks like, it will match that.

Automations

You can automate 1,000+ workflows with n8n. We cover this in a separate article, so check that out for more details.

What makes good Project knowledge

Relevance first: Only include information the AI actually needs. Ask yourself, "Will this help the AI do its job better?" If not, leave it out.

Keep it current: Regularly review and update your Project knowledge. Remove outdated materials. If you've connected files through integrations, use the refresh button to keep them up to date.

Organize strategically: Use clear file names and structure information logically. Make it easy to find and update.

Avoid contradictions: Don't upload conflicting guidelines. If you have multiple versions of something, choose one source of truth.

Test and refine: Start with core context, test the AI's responses, add more only if needed, and remove what isn't being used.

What not to do

Don't overwhelm with volume: More isn't always better. Quality beats quantity.

Don't include bad data: Wrong information is worse than no information.

Don't create confusion: Too many conflicting sources hurt performance.

How your team uses Project knowledge

Once you set up your Project knowledge thoughtfully and keep it clean, your whole team can enter your shared Projects and reuse this knowledge inside the chats they create. They won't need to re-upload files or re-explain context. Everything is already there, ready to use.