[[2025-19]]> ## Live LLAVA on Jetson Orin #til #jetpack #embedded I got my NVIDIA AGX Orin flashed to latest JetPack 6.2, which is based on Ubuntu 22.04. My goal for today was to get the jetson-containers examples working. Specifically, I was hoping to get the [Live LLAVA](https://www.jetson-ai-lab.com/tutorial_live-llava.html) example working. I kept hitting an issue though that was based on some `gstreamer` breakage that started to manifest in JetPack 6.1, attempting to open a WebRTC stream would result in an error like ``` window** ERROR:/opt/jetson-utils/codec/gstEncoder.cpp:876:static void gstEncoder::onWebsocketMessage(WebRTCPeer*, const char*, size_t, void*): 'sinkpad' should not be nullptr Bail out! ERROR:/opt/jetson-utils/codec/gstEncoder.cpp:876:static void gstEncoder::onWebsocketMessage(WebRTCPeer*, const char*, size_t, void*): 'sinkpad' should not be nullptr Fatal Python error: Aborted ``` This apparently stopped working after JetPack 6.1 for some reason. Someone on GH Issues said that [installing gstreamer1.0-nice fixes the issue](https://github.com/dusty-nv/jetson-utils/issues/185#issuecomment-1832735084). Deets are captured in this gist: https://gist.github.com/a10y/eff6b611b2c9ff7c6f18e1d1794a0ea2 ## Vortex on Ice Post #vortex #iceberg I wrote a post at work about Apache Iceberg integration for [Vortex](vortex.dev). You can find the post here: https://spiraldb.com/post/vortex-on-ice ---- ## GeoArrow #arrow #geo #vortex #webgl I've been thinking about how to improve the extension types in Vortex recently, and it's helpful to look at community extension types in Arrow/Parquet land to understand where they're being used to deliver value to people. To that end, I started with GeoArrow/GeoParquet, both because geospatial data is very close to my heart, and because its data that tends to be extremely messy to work with and I'm curious how Arrow is applicable to it. This series of Notebooks by Kyle Barron seems to summarize the usage model very well: - https://observablehq.com/@kylebarron/geoparquet-on-the-web - https://observablehq.com/@kylebarron/geoarrow-and-geoparquet-in-deck-gl - https://observablehq.com/@kylebarron/prototyping-georust-geoarrow-in-webassembly The notebook uses [deck.gl](https://deck.gl/), a GPU accelerated graphics layer for data visualization. Barron write a Deck.gl plugin that allows zero-copy rendering of vector data in `GeoArrow` format. GDAL added support for reading vector layers out as Arrow arrays using the `GeoArrow` syntax. By allowing this, the whole point is that we have This observable notebook does a great job of describing the usage of GeoArrow and what it enables: efficient cross-platform memory layout that can be shared between JS/Rust to perform calculations in the browser.