Improve EPSS Importer and Getting EPSS history data #2435

EPSS is currently represented as a VulnerabilitySeverity associated with an AdvisoryDataV2.

The importer creates one advisory for each CVE and stores the EPSS score and percentile as a vulnerability severity.

The current importer fetches the latest EPSS dataset for a single day, for example: epss_scores-2026-09-14.csv.gz, The pipeline then processes the entire dataset and creates a large number of advisory records. A single EPSS import currently takes approximately 3–12 hours.

This becomes a major problem when trying to import the complete EPSS history. The EPSS project provides historical daily datasets covering multiple years.

With approximately 2,190 daily files for roughly 6 years of history:

~365 days × 6 years = ~2,190 files
At an average of ~3 hours per import:
2,190 × 3 = 6,570 hours
6,570 / 24 = 273.75 days

Therefore, importing the complete history through the existing pipeline is not practical without implementing some form of bulk creation.
see:

• Add Support for the EPSS History Pipeline Importer #2433

There is also a second problem:

• importing every historical file through the normal AdvisoryDataV2 workflow would generate a very large number of advisory records, even though EPSS is fundamentally time-series data where the score for a CVE changes over time.

I think we need a better way to handle EPSS data.