Optimize Rovo Chat performance and reduce latency
Platform Notice: Cloud Only - This article only applies to Atlassian apps on the cloud platform.
Summary
Users may occasionally experience significant variance in Rovo Chat response times (ranging from 30 seconds to several minutes). This is often not a backend service failure but a result of prompt complexity and skill orchestration overhead. This article provides actionable steps to optimize prompts for faster, more consistent results.
Diagnosis: Why is Rovo Chat slow?
Rovo Chat is a general-purpose AI interface. When a prompt is complex or multi-step, Rovo must:
Evaluate Skills: Decide which tools (searching, reading, analyzing) are needed for each step.
Fetch Data: Retrieve and ingest content from external URLs or Confluence pages.
Process Iteratively: Multi-step prompts (e.g., "Read Page A to audit Page B") require sequential processing, which compounds latency.
Solution
Optimization Steps
1. Inline Your Instructions
Instead of asking Rovo to fetch instructions from a separate Confluence page or URL, paste the full logic directly into the chat prompt.
Benefit: Removes the "fetch and ingest" step, significantly reducing initial latency.
2. Be Explicit About Scope
Tell Rovo exactly what to ignore or focus on.
Example: "Audit this page for spelling only. Do not search for external references or cross-check with other spaces."
Benefit: Reduces the number of "skills" Rovo attempts to invoke.
3. Limit Knowledge Sources
Narrowing the search space helps Rovo process information faster.
Filter Apps: Use the source filter to toggle off unused apps (e.g., disable Slack or Jira if you only need Confluence data).
Target Spaces: Explicitly mention the specific Confluence space(s) in your prompt.
4. Transition to Custom Rovo Agents
For repeatable, complex workflows (like documentation audits), a Custom Rovo Agent is the recommended solution.
Pre-configured Knowledge: You can embed your audit logic directly into the Agent's Instructions.
Dedicated Sources: Set specific pages as Knowledge Sources so the Agent "knows" them without needing to fetch them dynamically.
Benefit: Provides the most consistent performance and highest quality output.
Related Resources
Was this helpful?