David Gageot

Now it is... That's so me

Tried watching a soccer game with my son on a Brazilian channel, with French radio commentary.

Why? We don’t speak Portuguese; and the charm of a foreign language disappears when you hear about a goal 12 seconds before you see it.

No solution on Google.
So I coded this: github.com/dgageot/dela...

🍳 New blog post! Using @docker.com #cagent with Tiny Language Models & #Skills.

I turned a 4b param local LLM into a cooking mentor using Docker Model Runner + #cagent skills. No cloud, no big model, just a small LLM with the right prompts & recipes.
k33g.org/GRP-Coding%2...

AI agents don’t always understand containers.
Gordon does.

Now in beta, Gordon is Docker’s AI agent that works with your actual Docker environment - inspect logs, fix issues, generate Dockerfiles.

Just run `docker ai` in Docker Desktop 4.61+ to get started.

Learn more → https://bit.ly/46XgGAv

Can we do without Claude Code?
I tested #cagent from @docker.com with gpt-oss-120b on OVH AI endpoints.
First try on WebAssembly: big fail.
Second try with "knowledge skills": it worked! 🎉
The key: feed your coding agent with examples & context.
k33g.org/GRP-Coding%2...

🎥 Replay dispo !
Mon talk à la @cloudnativedays.fr est en ligne 🚀

Salle comble
🐳 @docker.com Bake, builds en parallèle + présentation des features
⚙️ Bake dans les outils de CI CD, App Python + CI as Code avec #dagger

👉 Lien en commentaire
Merci à l’orga et à la communauté ❤️

#Docker #DevOps #GCP

It's been my new go to model for the past 2 days. I do feel it's bother faster and as good.

When I switched from opus 4.5 to opus 4.6, I didn't see much difference for coding and it felt that it was consuming more tokens.

Des parents qui ont leur enfant à l'ESILV ? De mon point de vue, cette école est catastrophique et je cherche des avis concordants ou contraires.

github.com/docker/cagent is Docker's open source Agent Builder and Runtime.

Writing an agent's is great. But how do you make sure it's working well? 🤔

Use evals! cagent comes with built-in evals:💡
- Record a session with `/eval`
- Replay your sessions, with an updated prompt, with `cagent eval`

github.com/docker/cagent is Docker's open source Agent Builder and Runtime.

🐳 Here's a daily tip just for you!

Want to breeze through those Anthropic outages like it’s nothing? ⚠️

cagent supports fallback models for when a provider fails to respond. 🎉

github.com/docker/cagent is Docker's open source Agent Builder and Runtime.

What's an AI Agent without Skills?

🐳 Here's a daily tip just for you!

Your agent only needs this to support Skills. Whether it's Codex style Skills, Claude style or Agent Skills standard style, we got you covered!

github.com/docker/cagent is Docker's open source Agent Builder and Runtime.

It's very customisable. Use it to build and share any kind of AI Agents. Not just Coding Agents!

🐳 Here's a daily tip just for you!

An agent can have its own super fast local RAG in less than 10 lines of yaml.

I find myself comparing LLMs a lot and models.dev is a GREAT source of reference data but their UI is frustrating.

So I built modles.dev

It's more searchable, allow comparing models, works on a phone, allows copying ids and details with just a click.

And it's fast!

cagent run --model=anthropic/claude-opus-4-6

github.com/docker/cagent is Docker's open source Agent Builder and Runtime. It's very customisable. That make's it my coding agent of choice!

🐳 Here's a daily tip just for you!

Use /cost to see the detailed cost of a session.

Sexy primes are not sexy for everyone!
Happy 59!

Breaking Free From AI Vendor Lock-in: Integrating GitHub Models with Docker cagent

Breaking Free From AI Vendor Lock-in: Integrating GitHub Models with Docker cagent/#docker #container - The landscape of AI development is rapidly evolving, and one of the most exciting developments in 2025 from Docker is the release of Docker... https://tinyurl.com/2a3dw28v

Break Free From AI Lock-In: GitHub Models + cagent | Docker See how Docker cagent integrates with GitHub Models to build and ship multi-agent apps without vendor lock-in.

Want true AI vendor freedom? Docker Captain Naga Santhosh Reddy Vootukuri shows how Docker cagent + GitHub Models lets you spin up multi‑agent workflows with just a YAML file.

No Python soup, no SDK mess.
Build, run & share AI agents - with one GitHub token.
https://bit.ly/3L1pZrs

Dynamic MCPs in your editor?? Yep.
With cagent + ACP, you can run dynamic MCP agents in Neovim or Zed - with tool discovery, code-mode, and full sandboxing.
Read more: www.docker.com/blog/dynamic...

Building AI Agents Using Open-Source Docker cagent and GitHub Models Discover how Docker’s open-source cagent framework and GitHub Models simplify AI agent orchestration. Learn to build, pack...

#Contributed #Content #Docker #Social #- #Facebook #Social […]

[Original post on cloudnativenow.com]

I love those wide events. Similar technics have worked like a charm for me

Demande à Claude de simplifier ton code / le rendre plus simple à lire et maintenir, tout en gardant 100% des fonctionnalités.

To read: the @rumpl.bsky.social vibe-coding adventures with cagent
rumpl.dev/posts/vibe-y...

Le mot de passe n’a aucune valeur. Il ne sert qu’à protéger toutes les autres données

Gemini 3 just dropped.
The buzz is real.
And yes, cagent already works with it. 🔥

Built by engineers, for engineers.
Have you tried it yet?

#Docker #cagent #Gemini3 #AI #DevTools

Well, the less time it spends in Python space, the better !

Ah ah. I spent the whole day with Apache Arrow in python

GitHub - docker/cagent: Agent Builder and Runtime by Docker Engineering Agent Builder and Runtime by Docker Engineering. Contribute to docker/cagent development by creating an account on GitHub.

Did I mention cagent works well with Gemini 3? Well, it does! github.com/docker/cagent

You can share agents as OCI artifacts or docker images.

It uses OpenAI, Mistral, Anthropic, Gemini, Docker Model Runner and any OpenAI endpoint.

What else?

You can plug mcp tools (local, remote, MCP Gateway). There's a lot of built-ins (filesystem, shell, todo, memory...), or any http api.

You have sub-agents, /commands, aliases, a TUI...

Agents can run interactively or be exposed with our http API. It also supports exposing over MCP, A2A and ACP.