Archives
42 posts published

A mini code agent with Docker Agent + Docker Model Runner - Part 1
Build a mini local code agent with Docker Agent and Docker Model Runner: a single shell tool on the small Mellum2 model, the agent loop, and running it safely in an sbx sandbox.
-docker model runnertiny language models
Docker Model Runner, standalone edition (`dmr`)
Install and use dmr, the standalone Docker Model Runner: one binary with the inference daemon and the full model CLI, no Docker Desktop needed.
-docker model runnerlocal models
Programmable guardrails for your local AI agent: NeMo Guardrails + Docker Agent, 100% on-device
Wire NVIDIA NeMo Guardrails into a local Docker Agent to catch jailbreaks, refuse out-of-scope requests, and mask PII on-device.
-docker agentsecurity
Preventing personal data (PII) from leaking to your LLMs β 100% local with Presidio, Docker Agent and Docker Model Runner
Stop PII from leaking to your LLMs with Microsoft Presidio, doing 100% on-device detection and redaction inside Docker Agent.
-docker agentsecurity
Controlling a local AI agent with Docker Agent hooks
Swap in Meta's Llama Guard 2 8B classifier to moderate a local agent's prompts through Docker Agent hooks, all on-device.
-docker agentmoderation
Controlling a local AI agent with Docker Agent hooks
Add a 100% on-device content-safety guardrail to Docker Agent by plugging Google's ShieldGemma 2B classifier into a hook.
-docker agentmoderation
`pk`, the TUI for building "coding helpers" - Part 3
Final tweaks to pk before learning Rust: context packing, new models, and finishing the local coding-helper setup.
-tiny language modelslocal models
`pk`, the TUI to build "coding helpers" - Part 2
Harden the pk coding helper: secure sandbox, a web IDE, RAG-based documentation search, and language server support.
-ragsecurity
`pk`, the TUI for creating "coding helpers" with very small local language models (<=`10b` parameters)
Build pk, a TUI for crafting coding helpers backed by very small local language models (<=10b parameters) for offline Rust work.
-tiny language modelstui
Running Mistral's Vibe CLI in an `sbx` sandbox
Run Mistral's Vibe coding agent CLI securely inside an sbx sandbox, with credential handling and network isolation.
-agentic composesecurity- D
Docker Agent meets Kronk safely with `sbx`
Run docker-agent against the Kronk local inference server, all wrapped safely inside an sbx sandbox with GGUF models.
-docker agentlocal models - I
Improving the Developer Experience with SBX Kits
Improve the developer experience with sbx kits: reusable sandbox customizations for Go, Node.js, and code-server extensions.
-docker model runnerdevcontainer - G
Giving capabilities to our mini `docker-agent` with "mini skills"
Give your mini docker-agent new capabilities with mini skills that fork context and delegate heavy lifting to scripts.
-skillstiny language models - H
How to cook a little coding agent with Docker Model Runner and Docker Agent (and `sbx`)
Cook a little coding agent from a 4B model with docker-agent and sbx, constraining tools so small local LLMs stay reliable.
-tiny language modelsdocker agent - E
Embedding a Web IDE and Docker Agent into `sbx`, a fully secured sandbox
Embed a web IDE and docker-agent into an sbx sandbox with Docker Model Runner for a ready-to-use secure dev environment.
-docker model runnerdocker agent - B
Building a form with π« Pablo
Go further with Pablo by building a real form: multiple fields, buttons, layout, event handling, and saving to a file.
-gobubbletea - P
Pablo, a small Golang library to make your life easier with BubbleTea
Pablo is a small declarative TUI framework for Go, built on BubbleTea, bringing HTML-like component composition to the terminal.
-gobubbletea - U
Using `docker-agent` with Docker Model Runner and `sbx`
Run docker-agent fully secured inside an sbx sandbox while it talks to Docker Model Runner for local inference.
-docker agentsecurity - H
Hello Estrela βοΈ
Estrela is a Lua 5.4 interpreter written entirely in GoloScript, pushing the revived Golo language into real language implementation.
-gogoloscript - "
"Micro" Vibe Coding with Docker Agent and Qwen3 Coder 30b
Micro vibe coding: use skills and precise templates to get moderately complex code out of a local model like Qwen3 Coder 30b.
-skillslocal models