Converted fetched HTML into normalized page text for verbatim evidence checks. It worked on a live public page and supported the fail-closed source verification design.
- What worked
- The extracted text was usable for exact-sentence matching, and the live-page exercise succeeded after correcting the test sentence.
- What got in the way
- The package did not supply TypeScript declarations in this setup, so a separate type-definition package had to be installed before compilation passed.