AI-Assisted Annotation Guide

This guide covers all the ways Potato uses AI and machine learning to speed up annotation.

AI Label Suggestions

Integrate any LLM provider to pre-annotate instances with suggested labels. Annotators review and correct rather than labeling from scratch.

Supported providers: OpenAI, Anthropic, Ollama, vLLM, Gemini, HuggingFace, OpenRouter

ai_support:
  enabled: true
  endpoint_type: openai
  ai_config:
    model: gpt-4o-mini
    api_key: ${OPENAI_API_KEY}

See AI Support for full configuration.

Visual AI Support

Use YOLO object detection and vision LLMs for image and video annotation tasks:

See Visual AI Support.

Chat Assistant

An LLM-powered sidebar where annotators ask questions about difficult instances. The AI provides guidance informed by your task description without auto-labeling:

See Chat Support.

Active Learning

Automatically reorder the annotation queue based on model uncertainty, so annotators label the most informative instances first:

active_learning:
  enabled: true
  schema_names: ["sentiment"]
  query_strategy: "hybrid"
  hybrid_weights:
    uncertainty: 0.7
    diversity: 0.3

In-Context Learning

Use few-shot examples from existing annotations to improve LLM labeling accuracy:

See ICL Labeling.

Option Highlighting

AI-assisted highlighting of likely annotation options to draw annotator attention:

See Option Highlighting.

Solo Mode (Human-LLM Collaboration)

A workflow where the system learns from annotator feedback and progressively transitions to autonomous LLM labeling as agreement improves:

Embedding Visualization

UMAP-based dashboard for visualizing instance similarity and exploring annotation patterns:

See Embedding Visualization.

AI Architecture

For developers extending the AI integration:

See AI Integration Internals.