georgep4181/pydanticai-DeepAgents

Pydantic AI DeepAgents Training Dataset

A high-quality conversational dataset for fine-tuning LLMs on the pydantic-deepagents library.

Dataset Details

• Format: Conversational (JSONL with messages array)
• Entries: 1,125 training examples
• Language: Python (pydantic-ai framework)
• Purpose: Code generation and agent creation patterns

Dataset Structure

Each entry follows the conversational format:

{
  "messages": [
    {"role": "system", "content": "You are an expert programmer..."},
    {"role": "user", "content": "Create a function that..."},
    {"role": "assistant", "content": "def function(): ..."}
  ]
}

Content Categories

• Agent Factory Patterns (~150 entries): Creating and configuring deep agents
• Dependency Injection (~80 entries): DeepAgentDeps patterns
• Backend Operations (~120 entries): StateBackend, LocalBackend, DockerSandbox
• Toolset Usage (~100 entries): Todo, Console, SubAgent, Skills toolsets
• Type System (~200 entries): Generics, TypeVar, protocols, overloads
• Async Patterns (~90 entries): Async functions, context managers
• Testing (~270 entries): Fixtures, assertions, mock patterns
• Language Fundamentals (~115 entries): Conditionals, loops, decorators

Usage

from datasets import load_dataset

dataset = load_dataset("georgep4181/pydanticai-DeepAgents")

# Access training data
train_data = dataset["train"]

# Example entry
example = train_data[0]
print(example["messages"])

Source

Extracted from the pydantic-deepagents repository, a production-ready framework for building AI agents with pydantic-ai.

License

MIT License - Same as the source repository.