Lightpanda 1.0.5 cuts memory, adds grid
Lightpanda released version 1.0.5 of its Zig-based headless browser for AI agents, cutting peak memory by storing page HTML once and adding CSS grid support. The update introduces a breaking change with asynchronous page.evaluate() in PandaScript alongside Gemini integration for browser decisions.
Lightpanda is rapidly closing the gap between a barebones scraper and a full headless Chrome alternative, giving AI agent workflows standard web fidelity without the multi-gigabyte RAM overhead.
- –Asynchronous page.evaluate() in PandaScript enables parallel DOM querying with Promise.all, substantially accelerating batch scraping and multi-element extractions.
- –Major memory optimizations—holding page HTML once instead of three times and shrinking the DOM tree—make running hundreds of concurrent agent browser contexts on low-cost VPS instances viable.
- –Layout additions like display: grid, calc(), and computed text wrapping improve hit testing accuracy, reducing misplaced element clicks by vision and automation agents.
- –First-class Gemini support for autonomous actions (classify, acceptCookies, login) expands LLM-driven web navigation beyond TypeSafe defaults.
- –Tightened CDP and Playwright compatibility ensures existing browser automation pipelines can swap out Chromium without rewriting session management logic.
DISCOVERED
3h ago
2026-11-20
PUBLISHED
4h ago
2026-11-20
RELEVANCE
AUTHOR
lightpanda_io