Soham Kukreti d709d670f9 fix: expose crawler_config on all MCP scrape tools (#1963) ...

MCP tools (md, html, screenshot, pdf, execute_js) hardcoded
  CrawlerRunConfig() with no user input, so wait_until,
  delay_before_return_html, cache_mode, and all other
  CrawlerRunConfig fields were silently ignored. /crawl already
  had full passthrough; this brings the remaining tools to parity.

  - schemas.py: add crawler_config: Optional[Dict] to all five
    request schemas so mcp_bridge.py exposes the field in MCP
    tool inputSchemas automatically
  - server.py: handlers now load via CrawlerRunConfig.load() then
    stamp endpoint-required fields on top (screenshot, pdf, js_code);
    fix screenshot_wait_for/wait_for_images defaults from 2/False to
    None so they only override crawler_config when explicitly passed
  - api.py: handle_markdown_request accepts crawler_config kwarg;
    cache_mode precedence uses key-presence check instead of falsy
    check so crawler_config.cache_mode correctly wins over legacy c

  Tests: tests/mcp/test_mcp_crawler_config.py — 7 MCP SSE tests
  proving delay_before_return_html is honoured server-side on all tools

2026-05-12 21:13:17 +05:30

..

docker

fix: expose crawler_config on all MCP scrape tools (#1963)

2026-05-12 21:13:17 +05:30