mirror of
https://github.com/bytedance/deer-flow.git
synced 2026-07-10 00:18:36 +00:00
* feat(community): add Browserless web_capture screenshot tool Add a web_capture tool that renders a page via Browserless /screenshot and presents it through the artifact system, alongside the existing Browserless web_fetch provider. Hardening: - SSRF guard: reject URLs resolving to private/loopback/link-local (incl. the 169.254.169.254 cloud-metadata endpoint)/reserved/multicast/unspecified addresses; opt out via allow_private_addresses for internal targets. - Surface a warning when Browserless renders a target page that itself responded with a non-2xx/3xx status (X-Response-Code), so an error/anti-bot page is not mistaken for valid visual evidence. - Dedupe colliding output filenames instead of silently overwriting prior captures. Docs: comment out token: $BROWSERLESS_TOKEN in tool examples (an unset $VAR fails AppConfig startup) and document allow_private_addresses. * fix(community): format web_capture guard + document local Browserless startup Address PR #3881 review: fix the lint-backend failure (ruff format on browserless/tools.py) and add local Browserless startup instructions to CONFIGURATION.md so reviewers can run the service to try web_fetch/web_capture. |
||
|---|---|---|
| .. | ||
| wizard | ||
| _detector_cli.py | ||
| check.py | ||
| check.sh | ||
| cleanup-containers.sh | ||
| config-upgrade.sh | ||
| configure.py | ||
| deploy.sh | ||
| detect_blocking_io_static.py | ||
| detect_thread_boundaries.py | ||
| detect_uv_extras.py | ||
| docker.sh | ||
| doctor.py | ||
| export_claude_code_oauth.py | ||
| load_memory_sample.py | ||
| run-with-git-bash.cmd | ||
| sandbox_memory_profile.py | ||
| scan_changed_blocking_io.py | ||
| serve.sh | ||
| setup-sandbox.sh | ||
| setup_wizard.py | ||
| start-daemon.sh | ||
| support_bundle.py | ||
| sync_labels.py | ||
| tool-error-degradation-detection.sh | ||
| wait-for-port.sh | ||