🧮 Working with CSV and similar files

What Types of Files Are Supported?

Assist 2.0 can automatically analyse data files including:

• CSV files (Comma-separated values)
• Excel files (.xlsx, .xls, .xlsm)
• TSV files (Tab-separated values)
• JSON files (JavaScript Object Notation)
• Other structured data formats

How Does Assist 2.0 Analyse Data Files?

When you upload a data file, Assist 2.0 works in a secure sandbox environment behind the scenes. It can:

• Read and process your data files
• Run Python code to analyse the data
• Create visualisations and charts
• Perform statistical calculations
• Generate insights from your datasets
• Produce new spreadsheets, decks, or documents from the results

Think of it as having a data scientist automatically examine your files and report back with findings — no setup required.

Where appropriate, Assist 2.0 will also reach for the right skill for the job (for example, a spreadsheet skill when the deliverable is an .xlsx   file). Skills are pre-built best-practice workflows that the agent picks up automatically. You don't need to name them.

[SCREENSHOT: Assist 2.0 chat with a CSV uploaded and the agent showing sandbox/skill activity in the thinking panel]

How It Works (Two Ways)

Method 1: Just Start Talking (Recommended)

The Smart Way — It Just Works:

1. Upload your data file (CSV, Excel, etc.) into the chat
2. Ask naturally about your data:
   • "What's in this spreadsheet?"
   • "Show me sales trends"
   • "Analyse this customer data"
   • "Create a chart from this CSV"
3. Assist 2.0 automatically:
   • Detects you have a data file
   • Recognises you want analysis
   • Spins up the sandbox and runs the work behind the scenes
   • Processes your file and returns results

You'll see the tool and thinking activity while it works. Once the final response is delivered, that activity collapses behind an expander so the chat stays clean.

Method 2: Explicit Request (Advanced)

If you want to be specific, you can say things like:

• "Run a full data analysis on this CSV"
• "Use the spreadsheet skill to clean and reformat this file"
• "Process this dataset in the sandbox and return an Excel summary"

This is usually unnecessary — natural language works fine — but it's useful when you want a particular output format (e.g. "give me back an .xlsx" or "build me a PowerPoint from this data").

What Happens Behind the Scenes?

When you upload a data file and ask for analysis:

1. File Detection: Assist 2.0 identifies your file as analysable data
2. Planning: The agent decides what's needed and, if in Plan Mode, proposes a plan before running
3. Code Generation: Python code is automatically written to process your specific file
4. Execution: The code runs in a secure sandbox environment
5. Analysis: Data is processed, patterns identified, charts created
6. Results: Findings are formatted and returned to you, with any output files made available to download

Plan Mode vs Execute Mode

• Execute Mode: Assist 2.0 gets straight to work as soon as you ask. Best for quick, low-risk analysis.
• Plan Mode: Assist 2.0 proposes a plan first and waits for you to accept or revise it before running. Useful for bigger jobs or when you want to sanity-check the approach.

You can switch between modes from the chat input.

[SCREENSHOT: Plan Mode toggle in the chat input area]

Example Workflow

You: [Upload monthly_sales.csv  ] "Can you tell me what this data shows?"

Assist 2.0: (sandbox activity visible in the thinking panel)

"Your sales data contains 12 months of information with revenue trending upward. Here's what I found…"

• Chart showing trends
• Statistical summary
• Optional: a downloadable .xlsx   or deck if you ask for one

Best Practices

✅ Do this:

• Upload clear, well-structured data files
• Ask specific questions about what you want to know
• Use natural language — no special commands needed
• Tell Assist 2.0 what format you want the output in (chart, spreadsheet, deck, doc)

❌ Avoid this:

• Uploading image files expecting data analysis (use OCR-style requests instead)
• Asking for analysis without uploading the data
• Using overly technical language when a plain question works better

A Note on Workspaces and Knowledge

Assist 2.0 is workspace-scoped. A file you upload in one workspace stays in that workspace — it won't leak into another. If you're working across multiple projects, upload the file into the workspace where you want the analysis to live.

Key Benefits

• No coding required: You don't need to know Python or any programming
• Automatic processing: Smart detection means less work for you
• Visual results: Charts and graphs are produced automatically
• Comprehensive analysis: Statistical insights, trends, and patterns identified
• Reusable outputs: Generate spreadsheets, decks, or docs from the same dataset in parallel
• Secure environment: Your data is processed safely inside an isolated sandbox