difftextcontent
April 30, 2026 · View on GitHub
diff_text_content is the in-memory raw-text comparison endpoint for unified diffs between caller-supplied text pairs.
Use this endpoint when
- you already have both comparison sources in memory,
- you want unified diff output for one or more raw-text pairs,
- you want the bounded raw-text comparison surface of the comparison family.
Do not use this endpoint when
- your comparison sources already exist on disk,
- you need filesystem path validation and on-disk reads,
- you want the file-backed comparison role that belongs to
diff_files.
For those cases, use diff_files.
Public role
- Accepts
pairswithleftContentandrightContent. - Supports optional
leftLabelandrightLabelwith stable defaults. - Applies stricter cumulative raw-text budgeting before diff generation.
- Returns unified diff output bounded by the text-diff family cap.
- Keeps the raw-text comparison role separate from file-backed diffing.
Local documentation
CONVENTIONS.md— endpoint-local conventions and guardrails.DESCRIPTION.md— endpoint-local architectural explanation for LLM agents.