v0.13.2

Auto-released from merge of PR #52.

Added

• --browser-cdp-endpoint=<url> (env DEEPDIVE_BROWSER_CDP_ENDPOINT) — attach to an existing CDP browser instead of launching a local Chromium. BrowserSession.start() calls connectOverCDP when set, runs the fetch in an isolated context on the remote browser, and never downloads or launches Chromium — so containerized/fleet installs can share one hardened browser (e.g. a browser bridge) and skip the Playwright browser download. Launch args/stealth are owned by the remote browser; deepdive still sets the per-run context (UA, viewport, locale). deepdive doctor reports the CDP connection + remote Chrome version when the endpoint is set. Default (unset) behavior is unchanged.

Fixed

• Fetch batch over-allocated when the source pool was nearly full (src/agent.ts). The per-round batch was sliced to Math.max(headroom, candidatesFoundThisRound) — the larger of the two — so a round with 1 slot left but 5 candidates dispatched 5 Playwright fetches, discarding 4 once the keep-loop hit maxSources. Now capped at Math.min: the batch never exceeds the remaining headroom. Adds an agent-loop regression test (maxSources=2, 5 candidates → 2 fetches).
• robots.txt re-fetched once per URL instead of once per origin (src/cli.ts). runResearch built the agent config but never supplied a robotsCache, so canFetch's cache-miss path issued a fresh GET <origin>/robots.txt for every URL even when many shared a host. Now passes a per-run in-memory createRobotsCache().

---

Built + tested + npm-published inline by auto-release.yml on this run. See the workflow log for the provenance-attested publish output. (Body backfilled from the CHANGELOG — the extraction step shipped broken and left every auto-release body empty; fixed in #136.)