> ## Documentation Index
> Fetch the complete documentation index at: https://docs.prodexlabs.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Your First Project

> A hands-on quickstart: upload a data file, talk to Dexter, and walk away with your first artifact.

## Overview

This is a walkthrough for your first real session in ProDex — the kind of work that fits in one focused sitting. The spine is simple: **upload a file, talk to Dexter, get an artifact you can use.** Most users' first productive action isn't building a simulation — it's handing [Dexter](/product/ai-assistant) a spreadsheet and asking a real question about it. That's what this page walks through.

Building a simulation model is covered separately in [Your First Simulation](/getting-started/your-first-simulation) — come back here first.

## Before You Start

You need two things:

* You're **signed in** to ProDex ([First Login](/getting-started/first-login))
* A **Factory is selected** in the factory selector in the left sidebar. If you don't have one yet, create one from the factory selector — a name reflecting the operation you're modeling is enough.

You land on [**Chat**](/product/dexter/chat-and-tasks) by default. That's where the rest of this walkthrough happens.

## Step 1: Attach a File in Chat

Take a real file from your operation — a production log, a cycle-time spreadsheet, an MES export, an ERP demand report, a PDF work instruction, even a screenshot of a report.

<Tip>
  Anything close to your actual work is better than a toy example. Real column names, real status codes, and real quirks are exactly what Dexter is built to handle — and what makes the output immediately useful.
</Tip>

Two ways to attach it to a message:

* **Drag the file onto the Chat window** — the composer accepts the drop
* **Use the attach affordance in the message composer** (next to the text input) to browse and select a file

Supported inputs cover the formats manufacturing teams typically have on hand: **CSV, Excel (.xlsx / .xls), PDF, Word, PowerPoint, images, JSON**, and **CAD files (DWG, DXF, STEP)** — see the [CAD overview](/product/cad/overview) for what happens with drawings. If a format isn't accepted, Dexter will tell you in the reply.

Attach the file, add a short message describing what it is and what you want, and send:

> *"This is our last 90 days of cycle times from Line 3. Give me the distribution per station and flag anything that looks wrong."*

## Step 2: Let Dexter Profile It

On upload, Dexter will:

1. **Profile the file**: open it, understand its structure, identify columns, sample values, and get a sense of what each field means.
2. **Ask clarifying questions**: usually about column meanings, units, timezones, or the status codes and abbreviations only your team knows. Answer plainly and specifically — *"STN\_04 is the deburr station; times are in seconds; SCRAP means the part was scrapped downstream, not at that station"* is exactly the kind of answer that unlocks good work. Vague answers here produce vague analyses later.
3. **Show its work in the task list**: you'll see the steps Dexter is executing while it works. Profiling a typical spreadsheet takes seconds; larger files or PDFs may take a minute or two.

You don't need to wait passively — Dexter surfaces anything it needs from you as it goes.

<Note>
  Your answers aren't throwaway. What you teach Dexter about your data and your operation can be saved to the [Knowledge Base](/product/dexter/memories) — ask Dexter to record it, and the same questions won't come up again on the next upload.
</Note>

## Step 3: Ask for Something Valuable

Once the file is profiled, ask for something useful. Good first asks:

* *"Give me the cycle time distribution per station and flag any outliers."*
* *"Turn this into a reusable set of processing times I can use in a simulation later."*
* *"Show me demand by SKU by week and highlight the top 10."*
* *"Compare this month's throughput against the previous three."*

For work with real consequences, Dexter pauses at **review briefs** — structured summaries with an **Approve** / **Request Changes** decision. A brief typically shows the assumptions Dexter made, the numbers it derived, and what happens if you approve. If something is wrong, choose **Request Changes** and say what needs to change; Dexter reworks and re-presents.

<Warning>
  Read the review briefs. They are your leverage: catching a wrong assumption in a brief is far cheaper than catching it later in a downstream plan or model.
</Warning>

## Step 4: Find Your Artifact

Depending on what you asked for, the result lands in one of a few places — all inside your active Factory:

* A [**pipeline**](/product/templates) under the sidebar's [**Data**](/product/data-overview) surface: a reusable, auditable transformation from your uploaded file into derived inputs (cycle time distributions, processing times, demand profiles). Pipelines can be re-run when new data arrives.
* An [**insight**](/product/insights): a dashboard, chart, or KPI view for [exploratory analysis](/product/exploratory-analysis). Insights are factory-scoped and stay available for your whole team.
* A downloadable [**report**](/product/dexter/reports): PDF, PowerPoint, Word, or Excel, when you asked for something shareable.

Dexter tells you where the artifact lives and usually surfaces an **action card** in Chat with a button that opens it directly — the most reliable way to jump straight to it. Every artifact is yours to reuse — a pipeline feeds a simulation model, an insight anchors a team discussion, a report goes to your executives.

## Step 5: When to Graduate to a Project

A plain Chat conversation is the right container when the work fits in one sitting: profile a file, get a chart, draft a report, done.

Graduate to a [**Project**](/product/dexter/projects) when the work will span multiple conversations or needs staged sign-offs — a full simulation study, a multi-week model build, an ongoing planning effort. Projects give the work a durable home with **phases**, **gated sign-offs** at each phase boundary, and the ability for Dexter to **resume across sessions** without you restating everything.

A new conversation opens with quick-action cards grouped under category chips — **Projects**, **Quick tasks**, and **Chat**. The **New simulation project** card under **Projects** (*"Create a project and build a sim from your data"*) is the fast path to a scaffolded simulation Project from your uploaded data. Other cards under **Quick tasks** offer common starting shapes; browse them, or just tell Dexter what you're trying to do.

You can also ask Dexter mid-conversation to promote the work into a Project — Dexter scopes the phases and opens the Project with your approval.

## What's Next

* **Model your operation**: [Your First Simulation](/getting-started/your-first-simulation) builds a simulation model — the pipeline you just created can feed it directly
* **Plan production**: head to the sidebar's [Planning](/product/planning-overview) surface if you have demand orders, inventory, and [BOMs](/product/bom-overview) to work with
* **Build a shareable report**: ask Dexter in Chat to turn your insight into a PDF or PowerPoint deliverable
* **Teach Dexter your operation**: the more it knows about how your factory actually runs — told in Chat or edited directly in the [Knowledge Base](/product/dexter/memories) — the sharper every subsequent piece of work becomes

The pattern that just worked — **upload, talk, get an artifact, approve** — is the pattern the whole platform is built around. Every capability in the sidebar is a variation of it.