before and after screenshots
A simple CLI tool to make Excalidraw SVG exports theme-aware (supporting automatic light/dark mode) and transparent github.com/chgeuer/excalidraw_themify
@christian.geuer-pollmann.de
Principal Engineer working on @Azure, security & cloud architect, Apache Committer, DG5EAW, http://genserver.social/chgeuer. Private account. #ElixirLang. Archive https://bsky.app/profile/chgeuer.bsky.social
before and after screenshots
A simple CLI tool to make Excalidraw SVG exports theme-aware (supporting automatic light/dark mode) and transparent github.com/chgeuer/excalidraw_themify
Things would be spicy if flamingos would fly in Kaliningrad ๐ฆฉ
Taking Chris McCord's idea one step further. What if #elixirlang accepts a TCP connection, chats with the client, then hands the live socket to a Rust process? One unbroken TCP stream, two languages. SCM_RIGHTS is wild. ๐ฆ๐ฎ github.com/chgeuer/hand_off_to_rust Really hard /cgi-bin vibes..
Inspired by Chris McCord's tweets yesterday โ built a blue/green hot deployment demo in Elixir. Accept a TCP connection on v1 on BEAM node 1, then live-upgrade to v2 mid-session using SCM_RIGHTS FD passing between peer nodes. The client never disconnects. ๐ต๐ข github.com/chgeuer/blue_green
translate this to english and read: steady.page/de/u-m/posts...
Merci vielmals ๐. Der Tweet von @ftes.de hielt mir vor Augen, wie oft ich selber gerade PORT=4000irgendwas tippte, und wie sehr das nervt.
Just put "PORT=$(phx-port)" into your run scripts and never thing about the friggin port again
Your Elixir Phoenix projects are fighting over port 4000 on your machine? ๐ Fix it forever: github.com/chgeuer/phx-... #myelixirstatus #elixirlang
That was the one I was looking for yesterday. I remember seeing it but then I totally forgot about it. Will check it out. Looks super interesting!
I donโt know yet what cool #elixirlang stuff that is but it clearly needs more visibility #elixir #erlang #beam
Yeah, Iโm waiting for Mike Hostetler to catch some breath and put some missing Jido libraries on hex.pm, currently my project only builds when you locally check out everything
Check this out. You get all the data of tool calls, inputs, outputs, usage (in case of Copilot)
bsky.app/profile/chri...
My neighbour did that (with a Python-based web harness) and after I showed him what I can get via ACP or GitHub copilots even richer CLI server protocol, he ripped out all the fragile PTY stuff immediately, no looking back
Iโm all in for that. But if I may say: try to avoid all that PTY nonsense, and adopt stuff like ACP which gives you many more details. Having the animated ANSI sequence stuff in your xterm.js feels badass, but you miss out on the full data set, tool call details and stuff
Nx?
Screenshot of bash sessions of the LLM using the BEAM introspection skill to figure out which Erlang processes have most reductions.
This is a better visualization how the beam-introspection skill worked
That visualization looks nice. How do you do this, in the frontend, or do you have some sort of LiveView SVG renderer?
How did they do the lookup from IP address to email? Did they check DNS contacts? Or ask the hoster?
And ... fixed
"My live view process is stuck. Use beam introspection to figure out what problem I'm having here in this application."
This indeed figures out the 15.7 billion reductions itself and makes my laptop fan cool down again.
Itโs a bit like tidewave without all the UI stuff, just a bunch of shell scripts
Your AI coding agent can iex --remsh into your running BEAM node. Agents use a single shell script + a skill definition. Now they inspect GenServer state, trace processes, hot-reload fixes (Soul of Erlang&Elixir), but the operator is an AI. ๐คฏ cookbook.geuer-pollmann.de/elixir-beam/... #ElixirLang
๐ฏ%. Mรถgen Israelis und Perser die ganze Nummer gut durchstehen und die iIslamofaschistischen Mullah endlich loswerden
I was just discussing with GitHub Co-Pilot that Credo has no autofix capability So your post here came right in handy
Yes please
Yeah, when you have Tailscale, then regular Erlang clustering is definitely the best option. I was just trying to push my Azure messaging bits a bit to the edge with something unusual
The only issue is that it needs real supervised applications running, so I cannot do it early in the BEAM startup process (like the pure TCP or SSL/TLS carriers do), cause all the AMQP 1.0 production-grade stuff is a shit-ton of work ๐ง
Having a distributed Erlang/Elixir cluster with Erlang distribution running over Azure Relay Hybrid Connections, not too shabby. Allows me to do cross-site clustering without Wireguard etc.
#ElixirLang #Elixir #Erlang