Skip to content

fix: retain fetched page text as citation evidence - #9

Open
rudycelekli wants to merge 1 commit into
firecrawl:mainfrom
rudycelekli:fix/fetched-source-content
Open

rudycelekli wants to merge 1 commit into
firecrawl:mainfrom
rudycelekli:fix/fetched-source-content

Conversation

@rudycelekli

Copy link
Copy Markdown

Successful fetches currently keep an existing search preview in url_content because record_fetch uses setdefault. A docs answer can therefore save the short preview as its cited evidence even after the agent reads the full page. Replace the stored preview with the successfully fetched document text while keeping reader references out of engine_pool.

Validation:

  • Independent synthetic contract oracle: Actual Controller search→fetch call path: short preview is replaced by full fetched page; failed fetch preserves prior evidence and search accounting stays unchanged.
  • Exact upstream baseline: 1 failing / 1 passing focused tests. Fixed branch: full repository suite 134 passing.
  • PYTEST_DISABLE_PLUGIN_AUTOLOAD=1 /opt/homebrew/bin/python3.11 -m pytest -q and git diff --check pass.

No model/provider calls or private data. Agent loop, tool gate, depth cap, scoring rules, and denominator are unchanged.

Signed-off-by: Rudy Celekli <rudy@gradiahq.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant