Your first hour

While the index is building🔗

• The learning page shows files, chunks, symbols and languages counting upward. Steady progress is normal. The exact speed depends on repository size.
• Indexing speed also depends on the embedding model you chose in Step 5. Self-served embedding models are usually slower than the platform default.
• If you brought your own hardware, do not be alarmed if your GPU sits idle during indexing. Unless you selected your own embedding provider, the index is built with the platform's embedding model. Your hardware's first heavy workload comes with the first audit or task.
• You can already look around the workspace while indexing runs, but agents give their best results once learning has finished.

Your first task🔗

1. Open your namespace and go to the Tasks view.
2. Create a task and describe what you want changed or investigated, the way you would brief a colleague. Name the behavior, the place in the product, and what "done" means.
3. An agent picks the task up and works it. You can follow its progress from the task view.
4. When the agent finishes, review the result. Code changes arrive as proposed changes on a branch in your git provider. Review them there like any other change.

A normal task takes minutes, not seconds. If a task has shown no activity at all after several minutes, something is wrong. See Health checks and Troubleshooting rather than re-creating the task.

What else is in the workspace🔗

• Goals. Describe a larger outcome and shape it into tasks with the agent, instead of writing every task by hand.
• Audits. Scirix continuously checks your code and documentation for drift, inconsistencies and convention violations, and turns confirmed findings into work items. While a doc-code audit runs, task processing is paused across all agents. A banner says so, and queued tasks stay queued until the audit completes. An audit over a real repository takes a while. This is normal, not a hang.
• The assistant. Ask questions about your codebase in plain language. It answers from the namespace's index rather than from generic knowledge.

Expectations worth setting🔗

• Scirix works from its index. Right after big pushes to your repositories, give it a moment to re-learn before expecting answers about the newest code.
• Sessions expire. If you leave Scirix open for a long time and pages stop responding, sign out and back in. See Troubleshooting.