opa-archive

A Python library for creating and signing OPA (Open Prompt Archive) files.

OPA is a ZIP-based archive format for packaging AI prompts together with data assets, session history, and metadata -- similar to how JAR files package Java applications.

Installation

pip install opa-archive

Quick Start

from opa import OpaArchive, Manifest, Prompt, ExecutionMode

manifest = Manifest(
    title="My Prompt",
    description="A packaged AI prompt with data",
    execution_mode=ExecutionMode.BATCH,
)

prompt = Prompt("Summarise the attached data file.")

archive = OpaArchive(manifest=manifest, prompt=prompt)
archive.add_data_file("data/report.csv", "local/report.csv")
archive.write("my_prompt.opa")

Features

• Manifest -- JAR-style MANIFEST.MF with title, description, execution mode, and custom fields.
• Prompt -- Markdown prompt files with {{variable}} template support.
• Data assets -- Bundle arbitrary files and directories with a typed INDEX.json catalog.
• Session history -- Include multi-turn conversation context (text, images, tool use).
• Signing -- JAR-compatible PKCS#7 digital signatures using the cryptography library or the openssl CLI.

Signing an Archive

from opa import OpaArchive, Manifest, Prompt, Signer
from opa import generate_signing_key, generate_self_signed_cert

# Build the archive
archive = OpaArchive(
    manifest=Manifest(title="Signed Example"),
    prompt=Prompt("Hello, world!"),
)
archive.write("example.opa")

# Generate a key pair and sign
private_key = generate_signing_key(key_type="rsa", key_size=2048)
certificate = generate_self_signed_cert(private_key, common_name="My Signer")

signer = Signer(private_key=private_key, certificate=certificate)
signer.sign("example.opa")

You can also load existing PEM keys:

from opa import load_private_key, load_certificate, Signer

key = load_private_key("key.pem")
cert = load_certificate("cert.pem")
signer = Signer(private_key=key, certificate=cert)
signer.sign("example.opa")

Template Variables

Prompts support {{variable}} placeholders:

prompt = Prompt("Translate the following to {{language}}:\n\n{{text}}")
print(prompt.variables())  # {'language', 'text'}
print(prompt.render({"language": "French", "text": "Hello"}))

Data Assets

Bundle files with a typed index:

from opa import DataIndex

index = DataIndex()
index.add("data/feed.xml", description="RSS feed", content_type="application/rss+xml")
index.add("data/config.json", description="Config file", content_type="application/json")

archive.set_data_index(index)
archive.add_data_file("data/feed.xml", "local/feed.xml")
archive.add_data_dir("local/data_folder/", archive_prefix="data/")

Session History

Include conversation context:

from opa import SessionHistory

session = SessionHistory()
session.add_user("What is the capital of France?")
session.add_assistant("The capital of France is Paris.")

archive.set_session(session)

Dependencies

The core library has no required dependencies. Signing works out of the box if openssl is on your PATH. Install the optional cryptography package for native Python signing:

pip install opa-archive[crypto]

License

MIT