Continue from an Earlier Result

Pick up from a table or object saved by an earlier chat instead of running the whole analysis again.

You don't need to redo QC every time you have a new question. Mention a file that an earlier chat saved, and Drylab starts from it, or fork a chat to try another approach from any answer.

When to use it

An earlier chat saved a clustered .h5ad, and now you want marker genes or cell-type labels without repeating QC.
You have a DE table from last week and want pathway enrichment on it.
You want to try a different normalization or clustering resolution without changing the original analysis.
A chat has grown long, and you want to keep working from its results in a fresh, focused chat.

Prompt examples

Single-cell, after a QC chat:

Load @pbmc_clustered.h5ad from my earlier QC chat and find marker genes for each cluster. Don't redo QC or clustering. Show a dot plot of the top 5 markers per cluster.

Bulk RNA-seq, after a DE chat:

Using @de_tumor_vs_normal.csv, run GSEA on the Hallmark gene sets, ranked by the log2 fold change. Save the results as gsea_hallmark.csv and plot the top 10 pathways.

Say what the file already contains:

@pbmc_annotated.h5ad already has cell types in the cell_type column and the condition in the group column. Compare the proportion of each cell type between treated and control.

Start a fresh chat with a short summary:

Continuing a scRNA-seq analysis. @pbmc_clustered.h5ad holds the QC'd cells, clustered with Leiden at resolution 0.8. Goal of this chat: annotate the clusters with the markers in @marker_panel.csv.

Steps

	Mention the saved file in a new chat	Fork the chat
Best for	A new question about a result	Trying another approach from a given answer
What Drylab gets	Only the file you mention	The notebook and context up to that answer, plus any files you pick
The original chat	Unchanged	Unchanged

Mention a saved result

Every chat saves its results in its own folder under History in your project, and the @ picker lists those files too. Each result shows its folder, so you can tell which chat made it.

Choose the same project as the earlier chat, then start a new chat. The @ picker lists the files of the current project, including every chat's folder under History.
Type @f (with a space) and part of the file name, for example @f de_tumor. Check the path under each result: /<project>/History/<chat title>/... shows which chat saved it.
Pick the file. It becomes a chip in your message.
Say what the file already contains and what you want next, then send it.

A file that a chat has just saved can take a minute to appear in the @ picker. If it isn't listed yet, wait a moment and type the name again.

Open the same project folder as the earlier chat, then start a new chat.
Type @ and part of the file name. In the Desktop App, @ searches your whole project folder, including each chat's folder under History.
Pick the file. It becomes a chip in your message.
Say what the file already contains and what you want next, then send it.

To reuse a result in many chats, ask Drylab to save a copy in Project Files in the web app, with a name you'll recognize. See Save Results to the Same Workspace.

Fork a chat

Open the chat and find the answer you want to branch from.
Hover that answer and click Fork under it.
In Fork to new conversation, change the name under New conversation name if you like. It starts as the chat's title followed by "(fork)".
Under Additional files / folders to copy, tick the files the new chat should have, such as the .h5ad you want to keep working on. The notebook and context up to that answer are always copied.
Click Create forked conversation. The new chat opens, and the original stays as it was.

In the Desktop App, the files are copied into the new chat's folder on your computer ("Files will be copied locally on your machine"). Fork a chat in the app where it was created: a Desktop App chat can't be forked from the web app, and a web chat can't be forked from the Desktop App.

Tips

Name the file and its contents. "The cell types are in the cell_type column" saves Drylab from guessing and keeps it from redoing earlier steps.
Say what not to redo, for example "don't redo QC or clustering".
Prefer a new chat for a new question. Short, focused chats make fewer mistakes than one very long chat. See Stay Focused in Long Analyses.
Fork to compare. Fork before changing a parameter, so you can compare both versions side by side.