Code Sandbox & File Generation

Overview

When you ask Evelant for something like "make me a slide deck about this" or "plot this data as a bar chart," it doesn't just describe what that would look like — it writes and runs real code in a secure, isolated sandbox and hands you back an actual file to download. You never need to know any Python or ask for code explicitly; just ask for the result you want.

Why this matters

Most AI chat tools can only describe what a chart or a spreadsheet would look like — you still have to build it yourself. Because Evelant can actually run code, it can hand you the finished file instead of a description of one: a real .pptx you can open in PowerPoint, a real .xlsx with working formulas, a real chart image, not a text explanation of what the chart would show.

How to use it

There's no special mode to turn on. Just ask for what you want in plain language:

• "Turn this into a PowerPoint deck with one slide per section."
• "Make a spreadsheet of these numbers with a totals row."
• "Plot this as a line chart."
• "Summarize this as a one-page PDF."
• "Package these files into a zip."

Evelant recognizes that the request needs a real file, writes the code to produce it, and runs it automatically. You'll see a terminal-style card appear in the chat while it works, followed by the result.

What you'll see

1. A card showing Executing Python — running in a secure isolated sandbox, with any printed output shown live.
2. Once finished: an exit status, how long it took, and — if the run produced any — a list of output files, each with a Download button.
3. Images and CSV files preview directly inside the card; everything else shows a filename and size before you download it.
4. If something goes wrong, the error output is shown instead, so you can see what happened and ask Evelant to try again or adjust the approach.

Supported file types

It can also just run code for calculations, data analysis, or quick scripts — the file types above are for when the goal is a document or image you can keep.

A worked example

Say you have a paragraph of quarterly sales numbers pasted into the chat and want a quick visual for a meeting.

1. Paste the numbers into the chat.
2. Ask: "Plot this as a bar chart with quarters on the x-axis."
3. A card appears showing the code running in the sandbox.
4. Once finished, a .png chart appears in the output files list — click Download, or if it's an image it previews right there in the card.
5. If the chart needs a tweak, just ask: "make the bars a different color" or "add a title." Each request is a normal follow-up message.

The same pattern works for a slide deck ("turn this into a 6-slide deck, one slide per product line") or a spreadsheet ("put this in a spreadsheet with a totals column") — describe the outcome, and a downloadable file comes back.

Example requests

• "Turn this data into a bar chart."
• "Make a 5-slide deck summarizing this article."
• "Put these numbers in a spreadsheet with a total row."
• "Convert this list into a CSV."
• "Calculate the average and standard deviation of these numbers."
• "Zip these three files together."

Tips

• Describe the end result, not the code. "A 5-slide deck comparing these three options" works better than trying to specify implementation details — Evelant handles the how.
• Point it at real data. Paste in the numbers, or reference a page/attachment already in the conversation, and ask it to build the file from that rather than placeholder data.
• Ask for revisions the same way you'd ask for anything else. "Make the chart a pie chart instead" or "add a slide summarizing the conclusion" works as a normal follow-up message.
• Runs have a time limit. Very heavy computations (huge datasets, long-running loops) may hit the sandbox's execution timeout — ask for a smaller or simplified version if that happens.

Troubleshooting

The run finished but there's no file to download. The code executed but didn't end up writing the specific file format requested. Ask again, optionally being more specific about the exact format you want (e.g. "as a .pptx, not a PDF").

"was not produced" notice. This means the run completed without errors but didn't create the file you asked for — Evelant deliberately doesn't substitute a different format in its place, since that would hand you something other than what you asked for. Ask again.

The download button doesn't do anything. This is rare and usually transient — try clicking again, or ask Evelant to regenerate the file.

A run seems to run too long or gets cut off. Sandbox runs are capped in both how long a single run can take and how many runs you can trigger per minute, to keep it responsive for everyone. Simplify the request (smaller dataset, fewer slides) and try again.

Frequently asked questions

Do I need to know how to code? No — you describe the outcome, Evelant writes and runs the code.

Is my data safe during this? Code runs in an isolated sandbox created fresh for each run — it doesn't have access to your account, your files elsewhere, or anything beyond what's in the conversation.

Can I see the actual code that ran? The card shows the run's printed output (stdout), not the source code itself — the focus is the result, not the implementation.

Does this cost more than a normal message? It's part of the same message exchange, not a separate purchase — see Plans & Limits for how credits work across different kinds of requests.

Can it read data from a page or file I've attached and use that? Yes — reference the page or attachment in your request ("using the numbers from this attached spreadsheet, make a chart") and Evelant builds the file from that content.

Can I ask for multiple files in one request? Yes — for example, "give me both a PDF summary and a spreadsheet of the raw numbers" can produce two separate output files in the same run, or ask for them as a single .zip.

What happens to the sandbox after a run finishes? Each run starts fresh and is discarded afterward — nothing persists between requests except the files you actually download.

Will a generated file still be in the chat after I refresh? Yes. The file card is saved with the message, so it comes back when you reload or reopen the conversation. Files larger than about 2 MB in total are not stored; for those, and for messages from before results were saved, Evelant runs the code again when you open your own conversation and saves the result. In a shared room the code is never re-run on your behalf, so a file that was not saved with the message is not regenerated there.

In the browser extension

File requests work in the Evelant extension too: in the floating chatbox and in both side panels. Evelant runs the script on its own servers before it answers, so you get the finished reply and the same file card, with a download button for each file. An .html file (a Plotly chart, for example) also has a Preview button that shows it right in the chat.

A few differences from the web app:

• You need a current extension build. An older install still gets the file, but shows no card: the reply tells you to update the extension or open the chat on the web to download it.
• Large files (roughly 180 KB or more) are not kept in the extension's own storage. They are on the card while the reply is on screen and always available from the same chat on the web.
• You must be signed in. Guests are asked to sign in instead of getting a file.

Related features

• Illustrations & Interactive Visuals — for live 3D models, hand-drawn diagrams, and interactive previews of generated HTML charts (like Plotly).
• Image Generation — for pictures generated directly from a description, rather than charts or documents built from data.
• Structured Data Extraction — pull tabular data off a page first, then ask the sandbox to turn it into a formatted file.
• Interactive Charts — for pie and donut charts you explore inside the chat, instead of a chart image you download.
• File & Image Attachments — for having Evelant read a file you already have, rather than generate a new one.