Scraper Audit

This is a utility for running audits on scraper output for legislative data entities (bill or event) using SQLMesh. The script processes JSON data into a DuckDB database and runs a SQLMesh plan, returning any audit errors.

Features

• Merges entity-level JSON files into one dataset.
• Initializes a DuckDB database with merged data.
• Runs sqlmesh plan on the staged models (staged.bill or staged.event).
• Extracts and prints any audit-related warnings or errors.

Requirements

• Python 3.9+
• poetry

Installation

Clone the repository and install dependencies using Poetry:

git clone [email protected]:openstates/scraper-audit.git
cd scraper-audit
poetry install

Usage

Ensure that the main directory contains a data folder with the JSON output files to audit. These files are typically generated by the OpenStates Scraper and should follow the naming pattern: */*/<entity>*.json For example:

_data/or/bill_0a3faf9c-1969-11f0-aaa5-4ef1b5972379.json
_data/or/bill_0a21cf0c-196b-11f0-aaa5-4ef1b5972379.json

Run poetry run python main.py --entity <entity name> for example poetry run python main.py --entity bill. This should produce output similar to:

INFO:openstates:Initializing data with arguments: entity=bill, jurisdiction=None
INFO:openstates:Merging JSON files matching pattern: ./*/*/bill*.json
INFO:openstates:Merged 1179 records into merged_entities.json
INFO:openstates:Creating DuckDB schema and loading data...
INFO:openstates:bill: initialized successfully
INFO:openstates:Running SQLMesh plan via subprocess...
INFO:openstates:SQLMesh plan output:

`prod` environment will be initialized

Models:
└── Added:
    └── staged.bill
Models needing backfill:
└── staged.bill: [2025-04-24 - 2025-05-05]

staged.bill  created
Updating physical layer ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100.0% • 1/1 • 0:00:00

✔ Physical layer updated


[WARNING] staged.bill: 'assert_bills_have_sponsor' audit error: 1179 rows failed. Learn more in logs: ~/scraper-audit/logs/sqlmesh_2025_05_06_20_58_30.log

[1/1] staged.bill  [insert/update rows, audits ❌1] 0.03s   
Executing model batches ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100.0% • 1/1 • 0:00:00                                                                                                                                                                                                                          
                                                                                                                                                                                                                                                                                                                 ✔ Model batches executed

staged.bill  created
Updating virtual layer  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100.0% • 1/1 • 0:00:00

✔ Virtual layer updated


Audit failed:
 [WARNING] staged.bill: 'assert_bills_have_sponsor' audit error: 1179 rows failed. Learn more in logs: ~/scraper-audit/logs/sqlmesh_2025_05_06_20_58_30.log

Docker Usage

To run the scraper-audit using Docker, you can build the image locally and execute it. Ensure your data directory is accessible to Docker. For example, if your JSON files are in a local _data folder:

Build the Docker Image

From the root of the project run

docker build -t scraper-audit .

Run the Container

Assuming your JSON data is in a local _data directory run

docker run --rm -v ./_data:/app/_data scraper-audit --entity "bill"

Note: The --entity flag is optional. If provided, it can be set to bill, event, or vote_event to audit a specific entity. If omitted, audits will run for all entities.