Rossum Transaction Scripts

The Rossum platform for automating document-based business transactions can evaluate snippets of Python code that can manipulate such transactions: Transaction Scripts (or TxScripts). The principal use of these TxScript snippets is to automatically fill in computed values of formula type fields. The code can be also evaluated in serverless function based extensions.

This is a user-friendly Python "extended runtime" that allows for code snippets in the Rossum platform (formula fields or serverless functions) that are very short, easy to read and follow, and easy to write (possibly in a recipe-like way for non-programmers, or by LLMs with low risk of mistakes).

See Rossum Transaction Scripts in Rossum API Docs for reference documentation.

Python 3.8

Note that the module code must be compatible with Python 3.8, since that's the serverless runtime. This mainly means that | None and type[...] constructs are not available.

Benchmarking

curl -H "Authorization: Bearer ..." -H "Content-Type: application/json" -XPOST "https://...rossum.app/api/v1/hooks/.../generate_payload" -d '{"annotation":"https://...rossum.app/api/v1/annotations/...", "event":"annotation_content", "action":"initialize", "status": "to_review"}' >payload.json
time ./benchmark.py payload.json

This is also an easy way to run txscript locally e.g. for debugging purposes.

Fields data model

We have three levels of data:

• attrs stored in FlatData._content dict indexed by schema_id.
• Field (and descendants) instances that provide a typed interface to attrs as well as the logic to work with complex data types and generating values and tracking updates.
• FieldValueBase (and descendants) instances that provide the user-facing interface to Field objects and emulate native Python types backed ultimately by attrs content.

User changes are tracked as such:

• When changing fields, the field is added to the Fields._updates journal, together with the information about which attributes changed. The field's .attrs itself (also stored in the FlatData) is updated with the new value. Removed fields are removed from all_objects.
• Newly added fields can be identified by the missing id attribute.

Publishing updates to PyPI

In the txscript package root directory, run:

pip install build twine
python3 -m build

Next, inspect the dist/txscript*.tar.gz tarball of source distribution to make sure it looks sane and doesn't contain any surprising extra files.

Finally, run:

python3 -m twine upload dist/*