fix: switch augment prompt to plain text input, drop GLiNER seed entities

[asteier2026]

Summary

• Switches the entity augment prompt to use COL_TEXT (plain, untagged) instead of COL_INITIAL_TAGGED_TEXT, so the augment LLM sees a clean slate with no GLiNER annotations. Removes the seed entities line and the do-not-repeat rule; updates both strict and non-strict example blocks to match plain text input.
• Because the augmenter now sees plain text, it commonly re-detects spans GLiNER already found and validated. resolve_overlaps() previously broke ties on identical spans purely alphabetically by label, so an unvalidated augmenter relabeling (e.g. Seattle -> address) could silently overwrite an already-validated detector span (Seattle -> city). Adds a provenance tiebreak so entities sourced from the augmenter always lose to any other source on identical spans, matching the invariant resolve_overlaps() already documented but didn't enforce.
• NOTE: extensive tests were run and not showing augment the gliner entities and say avoid these, actually helps augment find more entities that gliner didn't find.

Test plan

• python -m pytest tests/engine/test_detection_postprocess.py — 70 passed, including new regression tests covering the augmenter-vs-detector provenance tiebreak
• make format-check

🤖 Generated with Claude Code

[greptile-apps]

Confidence Score: 4/5

The PR is not yet safe to merge because broader sensitive augmenter spans can still be discarded when they overlap narrower detector spans.

Findings

1.   Broader sensitive spans discarded ▶

Summary

• Replaces tagged text and seed-entity prompt inputs with plain COL_TEXT.
• Adds overlap filtering so augmenter spans cannot displace validated spans.
• Adds regression coverage for prompt construction and augmenter provenance precedence.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
    A[Plain input text] --> B[Augmenter extraction]
    C[Validated detector entities] --> D[Overlap resolution]
    B --> D
    D -->|Overlapping augmenter span dropped| E[Resolved entities]
    D -->|Non-overlapping augmenter span retained| E
    E --> F[Validation and anonymization]

Loading

_{Reviews (3) · Last reviewed commit: "fix: switch augment prompt to plain text..."}

[greptile-apps]

Broader sensitive spans discarded

When the augmenter identifies a broader sensitive value that overlaps a narrower detector span, this filter discards the entire augmenter entity instead of preserving its non-overlapping sensitive portion. For example, if the augmenter finds a full street address containing a detector-identified city, only the city reaches validation and replacement. The rest of the address can remain unanonymized.

How this was verified: Only entities retained by resolve_overlaps() are passed into validation and replacement, and no later stage reconstructs the discarded span or its residual intervals.

[asteier2026]

This is by design. If gliner finds Seattle and augment finds "near downtown Seattle", it's better to hold on to the validated gliner entity "Seattle" and throw out the rest. I think it's too messy to try and weed out parts of the augment span that should still be an entity. Augment is supposed to find minimal spans. If "Seattle, Washington" shows up, we want augment to tag "Seattle" and "Washington" individually.

[greptile-apps]

That makes sense. Given the augmenter’s minimal-span contract, dropping an overlapping augmenter result is intentional: the validated GLiNER span is preserved, while the augmenter is expected to emit separate minimal entities such as Seattle and Washington rather than a broader span containing Seattle. I’m withdrawing this concern.