fix(lifecycle): ground company research in substantive evidence - #1023
Merged
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
Contributor
Contributor
Contributor
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Company capture could fill its six-snippet budget with navigation/footer labels. In dogfooding, an isolated “AI” label became a claim about product capabilities, while a company-site ranking was repeated without attribution.
This change excludes navigation/menu/footer and header-list subtrees, prefers main content with an empty-main fallback, and removes empty pages before research. The existing single-call enrichment prompt now requires substantive support and first-party attribution for retained promotional claims. The evidence schema and email flow stay the same.
Validation: lifecycle tests/check/lint/build and growth-research tests/check passed; 74 focused tests passed after expected failures for the new behavior. Independent code review passed. Six public-company cases completed through the production scraper and local changed generator; replaying the original Threadplane/Resend evidence omitted the observed misleading claims, and an empty-page control returned null fields with no claims. Final Resend capture/generation verified removal of duplicate header/footer labels and retained substantive API, SDK, simulation, and webhook descriptions. These are bounded smoke observations, not a general accuracy guarantee. No contacts changed or emails sent.