Knowledge Base

The knowledge base is how Zenmako's AI learns from your connected tools. When you connect a service like Asana, Trello, or HubSpot, Zenmako indexes the relevant data so the AI can answer questions quickly without making API calls every time.

What is the Knowledge Base?

The knowledge base is an indexed collection of information from your connected tools. It includes projects, tasks, documents, contacts, and other relevant data that the AI uses to understand your business context.

When you ask a question like "What tasks are due this week?" or "Who is our contact at Acme Corp?", the AI searches the knowledge base for answers rather than querying external APIs in real time. This provides:

Each connection has its own knowledge context, and all contexts are combined to give the AI a comprehensive view of your work.

How It Works

When you connect a tool, Zenmako automatically reads and indexes relevant information from that service. The indexing process varies by connector but typically includes:

The knowledge base is automatically compiled into a condensed context that the AI references when answering questions. This context is typically 2,000-3,000 tokens, providing enough information for accurate answers without overwhelming the AI.

Automatic Updates

When you first connect a tool, Zenmako indexes your data immediately. The knowledge base stays synchronized as your data changes, ensuring the AI has access to current information.

Triggering Knowledge Updates

While the knowledge base updates automatically, you may want to manually refresh it after making significant changes in an external tool.

To manually trigger a knowledge update:

The refresh process runs in the background. You will see a confirmation message when it starts, and the "Last updated" timestamp will change once complete.

When to Manually Refresh

Manual refreshes are useful after:

For routine changes like creating tasks or updating records, automatic synchronization is usually sufficient.

Viewing Knowledge Context

You can see exactly what the AI has learned from each connection.

To view the knowledge context:

The knowledge context shows a structured summary of what the AI knows about your data. For example, a project management connection might display:

Projects:
Project Alpha (In Progress) - 12 active tasks, 3 team members
Project Beta (Planning) - 5 active tasks, 2 team members
Marketing Campaign Q1 (Complete) - Archived
Key team members:
Jane Smith (Product Lead)
John Doe (Engineering)
Sarah Wilson (Design)
Upcoming deadlines:
"Launch MVP" due Jan 30
"Design Review" due Feb 1

The "Last updated" timestamp shows when this context was last generated.

Editing Knowledge Context

You can add custom context or notes that the AI should know about your business. This is useful for information that does not exist in your connected tools but is important for accurate answers.

Adding Custom Context

To edit the knowledge context:

Examples of Custom Context

Custom context helps the AI understand business-specific knowledge:

Project Apollo is our top priority for Q1.
All Apollo-related tasks should be treated as high priority.

John Smith is our key contact at Acme Corp.
Acme Corp is considering a contract renewal in March.

Tasks tagged "urgent" should always be flagged immediately.
Marketing approval is required before any external communications.

The design team is currently understaffed - Sarah is handling both UI and UX.
Engineering sprints run Monday to Friday with standup at 9am.

Editing the Learning Prompt

The Learning Prompt controls how Zenmako generates the knowledge context. You can customize it to focus on specific aspects of your data.

To edit the learning prompt:

For example, if you want the AI to focus on deadlines and blockers, you might add instructions like:

Prioritize information about upcoming deadlines and blocked tasks.
Include team member availability and workload where possible.

Preserving Edits

When you manually edit the knowledge context, Zenmako tracks that it has been modified. Future automatic refreshes will regenerate the context based on the learning prompt, so you may want to incorporate important custom information into the learning prompt itself for persistence.

Best Practices

Keep Connections Active

Active connections ensure the knowledge base has fresh data. If you disable a connection, the AI will still have the last indexed data but will not receive updates.

Add Custom Context for Business-Specific Knowledge

The AI cannot infer business priorities or relationships that are not explicitly documented. Use custom context to fill in these gaps:

Trigger Updates After Major Changes

While automatic sync handles routine updates, manual refreshes ensure the AI has the latest information after significant changes:

Review Knowledge Context Periodically

Check what the AI has learned to ensure accuracy:

Troubleshooting

Knowledge Context is Empty

If the knowledge context shows no data:

AI Seems to Have Outdated Information

If the AI references old data:

Custom Edits Are Not Persisting

If your manual edits disappear after a refresh:

Next Steps