Skip to content

quick-start.md

Latest commit

 

History

History
135 lines (87 loc) · 3.7 KB

quick-start.md

File metadata and controls

135 lines (87 loc) · 3.7 KB

Quick Start Guide

Get your first dataset ready in about 10 minutes! This guide walks you through installing FineFoundry and collecting your first batch of training data.

What You'll Need

  • A computer running Windows, macOS, or Linux
  • Python 3.10 or newer — Download Python here if you don't have it
  • An internet connection for downloading and collecting data

Optional (for sharing your work online):

Step 1: Download and Install

Option A: The Easy Way (Recommended)

Open your terminal (Command Prompt on Windows, Terminal on Mac/Linux) and run these commands:

git clone https://github.com/SourceBox-LLC/FineFoundry.git FineFoundry-Core
cd FineFoundry-Core
pip install uv

Then start the app:

On Mac/Linux:

chmod +x run_finefoundry.sh
./run_finefoundry.sh

On Windows:

uv run src/main.py

Option B: Traditional Installation

If the above doesn't work, try this instead:

git clone https://github.com/SourceBox-LLC/FineFoundry.git FineFoundry-Core
cd FineFoundry-Core
python -m venv venv

Activate your virtual environment:

  • Mac/Linux: source venv/bin/activate
  • Windows: .\venv\Scripts\Activate.ps1

Then install and run:

pip install -e .
python src/main.py

Step 2: Tour the App

When FineFoundry opens, you'll see a window with tabs across the top:

  • Data Sources — Where you collect data from websites or documents
  • Publish — Prepare and share your datasets
  • Training — Teach AI models with your data
  • Inference — Test your trained models by chatting with them
  • Merge Datasets — Combine multiple data collections
  • Analysis — Check your data quality
  • Settings — Set up accounts and preferences

Step 3: Collect Your First Data

Let's grab some data to work with:

  1. Click the "Data Sources" tab
  2. Choose a source — For this example, select "4chan"
  3. Pick some boards — Click a few board chips like b, pol, or x
  4. Set your limits:
    • Max Threads: 50
    • Max Pairs: 500
    • Delay: 0.5
    • Min Length: 10
  5. Click "Start"

Watch the progress bar and logs as data flows in. This usually takes 1-3 minutes.

Step 4: Preview Your Data

When the collection finishes:

  1. Click "Preview Dataset"
  2. You'll see a two-column view showing conversation pairs

Each row shows an "input" (like a question or prompt) and an "output" (the response). This is what the AI will learn from!

Your data is automatically saved, so you won't lose it if you close the app.

What's Next?

You've just collected your first dataset! Here's what you can do now:

Want to train a model?

Go to the Training Tab Guide to learn how to teach an AI using your data.

Want to share your dataset?

Go to the Publish Tab to upload it to Hugging Face.

Want more data?

  • Try different sources (Reddit, Stack Exchange)
  • Collect from multiple boards
  • Generate synthetic data from your own documents

Want to combine datasets?

Use the Merge Datasets Tab to mix data from different collections.

Having Problems?

  • App won't start? Make sure Python 3.10+ is installed
  • No data collected? Check your internet connection and try different boards
  • Other issues? See the Troubleshooting Guide

Still stuck? Ask for help in GitHub Discussions.

Next: Data Sources Tab | Back to Documentation Index