Python reference implementation

Architecture illustration: source, evidence, scoring, and findings.

Use case

Qualitative source material needs traceable evidence extraction and repeatable grouping into candidate findings.

How it works

A deterministic pipeline for normalization, signal detection, evidence extraction, scoring, grouping, and candidate findings. Optional local analysis is stored separately.

Inspect the implementation

The repository provides a canonical pipeline entry point, fixtures, a golden result, and pytest instructions.

Scope

The published reference is a Python pipeline. Optional analysis requires a configured local endpoint and model.

Evidence and interpretation

The canonical run_pipeline entry point normalizes sources, detects signals, creates evidence, scores and groups it into candidate findings. Optional local analysis receives only the evidence supporting a finding and is validated separately.

The reference includes fixtures and a golden result. Analysis cannot change evidence, scores, ranking, provenance, or finding membership.

Review procedure

python -m pytest

Content reviewed against repository documentation on November 5, 2026.