Getting Started Guide
This guide walks you through setting up your first annotation project with Potato.
Installation
pip install potato-annotation
Or install from source:
git clone https://github.com/davidjurgens/potato.git
cd potato && pip install -r requirements.txt
For full installation details, see Installation & Usage.
Your First Annotation Task
The fastest way to get started:
pip install potato-annotation
potato start examples/classification/single-choice/config.yaml -p 8000
Open http://localhost:8000 and you're ready to annotate. Browse the examples/ directory for ready-to-use templates.
For the step-by-step walkthrough, see Quick Start.
Understanding Configuration
Every Potato project is defined by a YAML configuration file. Key sections:
| Section | Purpose |
|---|---|
data_files |
Input data paths |
annotation_schemes |
Define annotation types and labels |
item_properties |
Map data fields (id_key, text_key) |
task_dir |
Output directory for annotations |
See Configuration Reference for all options, and Data Format for input/output specifications.
Choosing an Annotation Type
Potato supports 30+ annotation types. Start with the decision guide:
- Choosing the Right Annotation Type - Interactive decision tree
- Schema Gallery - Visual gallery of all types with examples
Common starting points: - Radio buttons for single-choice classification - Checkboxes for multi-label tasks - Likert scales for rating tasks - Span annotation for NER and text highlighting
Setting Up Workflows
Potato supports multi-phase workflows: consent, instructions, training, annotation, and post-study surveys.
- Multi-Phase Workflows - Configure phase progression
- Training Phase - Practice annotations before the real task
Exporting Results
After annotation, export in multiple formats:
- Export Formats - JSON, CSV, COCO, YOLO, CoNLL, Parquet
- HuggingFace Export - Push directly to HuggingFace Hub
Next Steps
- Administrator Guide - Managing annotators and quality control
- Developer Guide - Extending Potato and API integration
- AI-Assisted Annotation Guide - Speed up annotation with LLMs
- Agent Evaluation Guide - Evaluate AI agent systems
- Crowdsourcing Guide - Run tasks on Prolific and MTurk