agentmark-prompt-core 0.12.0

Python prompt-core package for AgentMark - high-level runtime for working with AgentMark prompts

AgentMark Prompt Core (Python)

Python implementation of the AgentMark prompt-core package. This package provides the high-level runtime for working with AgentMark prompts.

Installation

pip install agentmark-prompt-core

Usage

This package transforms pre-parsed MDX AST trees. The AST is typically obtained by:

• Parsing MDX with the TypeScript @agentmark-ai/templatedx package
• Loading a pre-parsed AST from a JSON file
• Receiving an AST from the AgentMark runtime

import asyncio
import json
from agentmark.prompt_core import create_agentmark, DefaultAdapter

async def main():
    # Create an AgentMark instance with the default adapter
    agentmark = create_agentmark(adapter=DefaultAdapter())

    # Load a pre-parsed MDX AST (from TypeScript parser or JSON file)
    with open("math.prompt.mdx.json") as f:
        ast = json.load(f)

    # Load and format a text prompt
    prompt = await agentmark.load_text_prompt(ast)
    result = await prompt.format(props={"userMessage": "What is 2+2?"})

    print(result)

asyncio.run(main())

Features

• Prompt Types: Text, Object, Image, and Speech prompts
• Message Extraction: System, User, and Assistant message roles
• Attachments: Image and file attachments in User messages
• Schema Validation: Pydantic-based validation matching TypeScript Zod schemas
• Adapters: Extensible adapter interface for different LLM providers
• Eval Registry: Registry for evaluation functions

Development

# Install dev dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Run linting
ruff check src tests

# Run type checking
mypy src/agentmark --strict

License

MIT

TypeScript parity roadmap

Known gaps vs @agentmark-ai/prompt-core, tracked as roadmap items:

• built_in_models validation — TS createAgentMark({ builtInModels }) has no Python equivalent yet.