Skip to main content
When you start an analysis, Galagos saves its state continuously in the background. You do not need to keep your browser open. Close the tab, shut your laptop, or switch to a different task — when you return to your project, the analysis will be exactly where you left it. After about thirty seconds of continuous streaming, the chat shows a small tip confirming the same thing: you can safely close the browser; the run continues server-side.

What gets saved

Session persistence covers every part of your working state:
  • Chat history — every message you sent and every response the agent produced, in order.
  • Uploaded files — every file in the project’s sandbox stays intact between sessions.
  • Intermediate outputs — files produced by completed steps remain available to later steps.
  • Background jobs — long downloads, alignments, and structure-prediction runs continue while you are away.
  • The Steps list — every command the agent ran, with stdout, stderr, and exit codes.
  • Analysis context — the agent retains context within a conversation, so you can ask follow-up questions about a result days after the original run finished.

Background jobs

When the agent kicks off a long-running task — pulling an SRA accession, aligning a genome, running AlphaFold — an orange status strip appears above the chat input listing the active process (“1 background process: wget”). You can continue working on other prompts in the same conversation while the job runs. Galagos emails you when:
  • A long-running job finishes.
  • The agent hits a decision it needs you to make.
Both notifications include a direct link back to the project.

Resuming an analysis

1

Close the browser whenever

Before, during, or after a run — it doesn’t matter. The sandbox keeps running.
2

Come back

Open Galagos and log in. From the dashboard, click the project you were working in.
3

Check the status bar

The status bar shows Ready if the agent is idle, or Working if a job is still in progress. The chat scrolls to the latest entry automatically.
4

Continue or download

Ask the next question. Refine the previous result. Download a report. Everything is in the same place.
Long-running analyses — whole-genome alignment, AlphaFold structure prediction, large-scale single-cell clustering — can take hours. You do not need to keep the browser open.

Time-travel editing

Sessions are not append-only. Every message in the chat has a small edit and retry control that lets you:
  • Edit and resend an earlier message — the conversation forks from that point.
  • Retry a failed step.
  • Revert to an earlier checkpoint.
  • Undo a revert if you change your mind.
Time-travel branches from history rather than overwriting it. The original path remains in the run record, which is important for reproducibility — a reviewer can still see what the first attempt looked like.

Built-in reproducibility

Storing the full workflow context is not just a convenience feature — it is the foundation of reproducibility. Because the platform records every tool call, every parameter, and every intermediate output, you or a colleague can:
  • Repeat the analysis exactly — rerun the same workflow on the same files and get the same results.
  • Audit the steps — review what happened at each stage and verify the parameters match your intended methodology.
  • Share the workflow — collaborators with access to the project can open the same chat and see the complete history.
See Reproducibility and audit for the full picture, including the workspace audit log and the reproducibility archive.

Frequently asked questions

The run continues server-side. A dropped connection does not interrupt the workflow. When you reconnect, the chat reflects whatever happened while you were offline.
Sessions persist as long as the project exists and your account is active. There is no expiry that would discard a running or completed analysis.
Yes. You can start a new analysis while a previous one is still running. Each one tracks its own state independently.
No. Intermediate outputs stay in the project’s sandbox and count toward the storage quota. They remain available until you delete them or the project.
Yes. The platform emails you when long-running jobs complete and when the agent needs a decision before it can continue.

Next steps

Reproducibility and audit

The Steps list, the audit log, and the reproducibility archive.

Running an analysis

The conversation loop the agent follows from question to result.

Visualisations and reports

What the agent produces alongside the run record.