0.36.0 (2026-09-19)
Features
- scrape: cache the fetch so a repeat scrape skips the network (601d5f5)
Bug Fixes
- crawl: do not fail a job because robots.txt could not be read (e8eea02)
- crawl: keep the last robots rule of a file that lands on the size cap (1e41fb0)
- extract: accumulate usage across legs, and report it on /v2/search (6623dce)
- extract: forward reasoning_effort from the structured builder (dc161c1)
- extract: stop corrupting markdown for non-HTML sources (0ae7c40)
- extract: stop the rescue candidate from reintroducing page chrome (8272600)
- harden consolidated crawl reliability changes (5691492)
- mcp: let CRW_LOCAL actually keep the run local (abc05e2), closes #557
- preserve crawl and renderer reliability invariants (65f385a)
- renderer: connect CDP over every resolved address, not just the first (77c0c74)
- scrape: scope the page cache per tenant (e0a583b)
- sdk: return the map sitemaps and keep local mode local (76c59c0), closes #557
- v2: carry llmUsage through to_v2_document (85fe020)