Hire me! I want to help an Elixir company deeply embed advanced AI Agents in ways that make sense and add real value to your product and value for your users. www.linkedin.com/posts/mark-e... #MyElixirStatus #ElixirLang #AI #LLM
Sagents Live Debugger is inspecting an agent and the status shows as interrupted with a hand icon and the text "Awaiting human decision"
Sagents Live Debugger is inspecting a sub-agent and shows it's status as "Interrupted" with a status badge.
The SagentsLiveDebugger was updated to raise visibility into the HITL state of your your agents and subagents. github.com/sagents-ai/s...
A sub-agent task was interrupted and is asking permission to delete a file.
The sub-agent was denied delete permission, the sub-agent completed it task and returned to the main agent.
Even sub-agents get the same treatment.
Shows agent conversation continuing after denied permission to delete a file.
The agent pauses, shows you what it wants to do (even multiple actions at once), and waits. You approve or reject each one. Then it picks up right where it left off.
Shows conversation interruption requesting user permission to delete a file.
Sagents v0.3.x is out! Lots to share!
AI agents doing things on your behalf is great... until they do the wrong thing. Sagents lets you define exactly which operations need human approval. github.com/sagents-ai/s... #MyElixirStatus #ElixirLang #LLM #AI #agents
🧵👇
News includes Hackney v3.1.0 with pure #Erlang HTTP/3 support, Hornbeam running #Python apps on the BEAM, the Easel Canvas 2D drawing library for #LiveView, Hologram v0.7.0 reaching 96% Erlang runtime coverage, and more! #ElixirLang @elixir-lang.org www.youtube.com/watch?v=DBuH...
I just published v0.2.0 of this Sagents library. The agents_demo was updated too, and is now cluster aware. When you shut down a node, it transfers a running agent to another node.
A new version is coming today or tomorrow that is fully cluster aware via Horde.
There's currently no linkage between Sagents and Claude Code. It could be done through one or more tools. If you create a Sagents middleware that adds ClaudeCode awareness and connection, I'd be like 🤩
Sagents is open source and ready to try:
Framework: github.com/sagents-ai/sagents
Live Debugger: github.com/sagents-ai/sagents_live_debugger
Demo App: github.com/sagents-ai/agents_demo
Full walkthrough: www.youtube.com/watch?v=r-14...
Debugging agents is hard. So we built a real-time LiveView debugger.
See every message, tool call, middleware action, and sub-agent event as it happens. Drop it into any Phoenix app with one line in your router.
github.com/sagents-ai/sagents_live_debugger
Need parallel execution? Sagents supports SubAgents. The parent agent spawns specialized child agents that work independently and report back.
Each SubAgent gets its own tools, system prompt, and isolated context. HITL approvals propagate up to the parent.
Agents that plan their work. The TodoList middleware lets agents break down complex tasks and track progress, visible to the user in real-time through the sidebar.
No more black-box execution.
AI agents shouldn't run unchecked. Sagents has built-in Human-in-the-Loop middleware.
Configure which tools need approval. The agent pauses, shows you exactly what it wants to do, and waits for your decision: approve, edit, or reject.
Sagents agents are built from composable middleware. TodoList, FileSystem, SubAgents, Summarization, and more. Snap them together like building blocks.
Here's the demo app with a virtual filesystem: create, read, edit, and search files, all in conversation.
Introducing Sagents -- an open-source #ElixirLang framework for building #AI agents with human oversight, composable middleware, and real-time Phoenix LiveView integration.
Built on OTP. Each agent is a supervised GenServer.
Watch the full demo: www.youtube.com/watch?v=r-14...
To me, it’s an entirely different thing to vary consciously design and develop with an LLM, but where it does all of the grunt work and boiler plate. But the decisions, the design, the architecture, that’s all mine. I don’t call that vibe coding.
To me the term “vibe coding” means I’ve handed over all thinking about how it should work to the LLM. That’s what works great for experiments and side projects. Let’s me build something like a little tool. I wouldn’t have created otherwise.
News includes @elixir-lang.org v1.20 achieving 2x compilation speedup, Telelang alternative #Erlang syntax, Membrane’s #AI-powered YOLO plugin, @josevalim.bsky.social PR Quiz tool for interactive code review learning, and the #ElixirConfEU speaker lineup! #ElixirLang www.youtube.com/watch?v=Makf...
News includes @elixir-lang.org v1.20.0-rc.1, 2 new Elixir-native Bash interpreters, zero-boilerplate component storybooks, #DurableObject library for persistent actors, @postgresql.org search extension, a new #RAG approach skips vector DBs, and more! #ElixirLang www.youtube.com/watch?v=BzOk...
News includes porting Erlang to QNX for automotive systems, ExDoc 0.40.0 with LLM features, Oban Pro for #Python, new #AI evaluation tools Tribunal and Beamlens, Tidewave hits 100k ARR, PhD thesis on typing Elixir, and more! #ElixirLang www.youtube.com/watch?v=a967...
Elixir’s 15th anniversary brings v1.20 RC with full type inference, Chris McCord launches Sprites(dot)dev, LiveVue v1.0 goes stable, Gust #workflow engine debuts, and more! #ElixirLang @elixir-lang.org #VueJS www.youtube.com/watch?v=dJA9...
News includes Arcana #RAG library for Phoenix, MquickjsEx embedding #JavaScript in #ElixirLang, LiveDebugger adds Streams support, Durable workflow engine, @josevalim.bsky.social teasing type improvements, Hologram receiving EEF support, and more! www.youtube.com/watch?v=cULP...
News includes #NextJS to Phoenix migration story, massive Credo performance improvements, Torchx on Apple Metal, DNSimple’s #Erlang case study, first #GleamLang conference announced, and more! Plus a 2025 retrospective. #ElixirLang www.youtube.com/watch?v=OW0R...
Wow! This is great! The recording quality is great too. Love the sound texture and feel. It's very chill without being loopy. It's beautiful. Love it. Nice work!
News includes running #LiveView in the browser with #WebAssembly, New Nerves book launches in beta, PinStripe for Stripe integration, Beam Bots for robotics, @sasajuric.bsky.social's #Goatmire talk, Tidewave updates, and more! #ElixirLang www.youtube.com/watch?v=8-f3...
Elixir v1.20 with full-type inference coming soon, @gleam.run v1.14.0-rc1, mjml_eex v0.13.0 for email templates, Dashbit’s nimble_zta library for zero trust auth, Björn Gustavsson’s #BEAM history talk, and more! #ElixirLang www.youtube.com/watch?v=rzjb...
Celebrating #Erlang’s 27th birthday as #OpenSource, Elixir Hub 2025 survey results, new data_migration LiveDashboard package, critical #React RCE vuln, and Zig leaves #GitHub, and more! #ElixirLang www.youtube.com/watch?v=0FWy...
Trying something new in the YouTube video of the episode! We’re showing the articles or projects as we talk about them. Is that helpful? Do you want me to continue doing it? Or do you prefer the audio only approach?
News includes cutting-edge type system research from @josevalim.bsky.social, Phoenix LiveView v1.1.18 with view transition support, LiveDebugger’s biggest release, Nx #Python-Elixir integration, and more! #ElixirLang @elixir-lang.org www.youtube.com/watch?v=IDvG...
