Pydantic AI Stash (prerelease)

Efficient serialization and storage for Pydantic AI conversations with binary content.

Overview

Pydantic AI Stash provides a clean, maintainable library for handling binary content within Pydantic AI message structures. It enables efficient storage and retrieval of binary data (images, audio, documents, etc.) while preserving metadata and ensuring round-trip fidelity.

Currently, we just have a simplistic file store adapter wired up. An adapter is designed specifically to hold your custom business logic (deduping, expiration times, db, s3, id, etc.)

Key Features

• Flexible Adapter Layer: A basic adapter class handles the identification, stashing, and loading of content types. Plug in your DB, S3, etc. tools here.
• Type Safety: Full type hints and compatibility with Pydantic AI message formats
• Metadata Preservation: Maintains all original BinaryContent metadata during stash/load cycles
• Nested Content Support: Handles BinaryContent in nested message part content lists
• Storage Abstraction: Pluggable storage adapters (filesystem included, extensible for cloud storage)
• Error Handling: Robust error handling for storage operations

Quick Start

from pydantic_ai_stash import BinaryStash, FSAdapter

# Set up storage
storage = FSAdapter("./binary_storage")
stash = BinaryStash(storage)

# Stash binary content in messages
stashed_messages = stash.stash_binaries_in_messages(messages)

# Later, load binary content back
loaded_messages = stash.load_binaries_in_messages(stashed_messages)

Installation

pip install pydantic-ai-stash

License

MIT License - see LICENSE file for details.