nixfredFLOXFIELD GUIDE · BY NIX UNOFFICIAL
LEARN / LANGUAGES AND TOOLS

Languages and tools

Flox does not need per-language plugins: any toolchain in the Flox Catalog can be pinned in the manifest. The docs add per-language guides, a VS Code extension, a secrets pattern with plugins for four secret stores, and a run of blog walk-throughs for AWS, Ollama, GPU model serving, ComfyUI and Slurm.

This page is drawn from https://flox.dev/docs/languages/python, https://flox.dev/docs/languages/rust, https://flox.dev/docs/install-flox/ide-extensions, https://flox.dev/docs/concepts/secrets-management, https://flox.dev/docs/tutorials/cuda, https://flox.dev/docs/concepts/activation and the Flox blog posts linked in the tables below. Blog posts are dated by the source and can be older than the current CLI, so check a recipe against the current docs before promising it.

Languages

The docs carry one guide per ecosystem under docs/languages (https://flox.dev/docs/languages/python).

LanguageWhat the source saysURL
PythonCovers a new project and an existing project, with a hook that sets up a venv and installs from pyproject.toml, then a “Build with Flox” sectionhttps://flox.dev/docs/languages/python
Node.jsA “Build with Flox” guide, including what to do when an app has no bundler and needs node_modules at runtimehttps://flox.dev/docs/languages/nodejs
GoA “Build with Flox” guidehttps://flox.dev/docs/languages/go
RustBasic environment, target on PATH, cargo aliases, nightly compilers, then builds; the example environment lives in the floxenvs repositoryhttps://flox.dev/docs/languages/rust
C/C++A “Build with Flox” guidehttps://flox.dev/docs/languages/c
JVMA “Build with Flox” guidehttps://flox.dev/docs/languages/jvm
RubyA “Build with Flox” guidehttps://flox.dev/docs/languages/ruby

Each page points back to the builds concept so the same environment that develops the code also builds it (https://flox.dev/docs/concepts/builds). Migration guides exist for Homebrew and nvm (https://flox.dev/docs/tutorials/migrations/homebrew, https://flox.dev/docs/tutorials/migrations/nvm).

Say: “Whatever your stack, the toolchain is a pkg-path in the manifest; the language page shows the hook and build pattern.” Ask: “Which language and which package manager does the team use today?”

Editors, agents and shells

ToolWhat the source saysURL
VS CodeThe Flox extension needs VS Code 1.87.0 or later and works with compatible editorshttps://flox.dev/docs/install-flox/ide-extensions
VS Code (blog, 2026-01-30)The extension “hooks into VS Code’s workspace and integrated terminal lifecycle”, so environments start pre-activated, with services and logs in the sidebarhttps://flox.dev/blog/reproducible-dev-environments-in-your-workflow-with-flox-and-vs-code/
Flox Skills for agentsflox, floxify and flox-debug skills teach AI coding agents to create environments, onboard a repo and diagnose resolution errorshttps://flox.dev/docs/install-flox/ide-extensions
direnvWith use flox in .envrc the activation is in-place and runs hook.on-activate and [profile] scriptshttps://flox.dev/docs/concepts/activation

Flox also ships native auto-activation, which the 2026-07-29 post says replaces “hacking direnv” (https://flox.dev/blog/auto-activation-is-now-native-in-flox/).

Secrets managers

The docs say secrets “should never be hardcoded in a manifest.toml file” and recommend retrieving them just in time during flox activate, exported as environment variables that live only while the environment is active (https://flox.dev/docs/concepts/secrets-management). Plugins exist for 1Password, HashiCorp Vault, OpenBao and Infisical (https://flox.dev/docs/concepts/secrets-management).

ToolWhat the source saysURL
Secrets patternAuthenticate once to a store, retrieve in an activation hook or plugin, secret gone when the shell exitshttps://flox.dev/docs/concepts/secrets-management
1Password pluginInstall flox/plugin-1password and declare references in [plugins.1password]https://flox.dev/docs/concepts/secrets-management
Portable secrets CLI (blog, 2025-03-18)Fetch credentials at runtime with a third-party manager across local dev, GitHub and AWShttps://flox.dev/blog/get-your-preferred-secrets-manager-in-a-portable-cross-platform-cli-toolkit/
AWS secrets (blog, 2024-12-05)Reproducible environments for managing AWS secrets and authenticating to AWS serviceshttps://flox.dev/blog/flox-and-aws-taking-the-chaos-out-of-secrets-management/

AWS

ToolWhat the source saysURL
AWS SAM and Lambda (2024-12-02)Install AWS CLI, AWS SAM and GitHub CLI from the Catalog, build and test Lambda functions locally, and containerize them for deployment; the walk-through lists aws-sam-cli (1.119.0)https://flox.dev/blog/flox-and-aws-building-testing-and-deploying-lambda-functions-with-flox-and-aws-sam/
AWS secrets (2024-12-05)See the secrets table abovehttps://flox.dev/blog/flox-and-aws-taking-the-chaos-out-of-secrets-management/

The 1.119.0 figure is the version in a 2024 post, not a current claim [verify] against flox show aws-sam-cli.

LLMs and local inference

ToolWhat the source saysURL
Ollama (blog, 2025-02-04)A pre-built flox/ollama environment: flox activate -s -r flox/ollama starts its services, including a Next.js chat UI at http://localhost:3000https://flox.dev/blog/get-deepseek-r1-on-your-laptop-with-ollama-and-flox/
ComfyUI (blog, 2025-06-10)Uses an environment from FloxHub to get ComfyUI running across systems, because setup instructions “vary across Linux, Mac, and Windows”https://flox.dev/blog/run-comfyui-anywhere-with-flox/
llama.cpp serviceThe auto-activation post shows a manifest that runs a llama.cpp proxy as a service with auto-start = truehttps://flox.dev/blog/auto-activation-is-now-native-in-flox/

GPU, CUDA and model serving

ToolWhat the source saysURL
CUDA in the CatalogPre-built CUDA packages sit under the flox-cuda namespace; the CUDA page is the tutorialhttps://flox.dev/docs/concepts/cuda-catalog-and-cache, https://flox.dev/docs/tutorials/cuda
GPU-optimized PyTorch (2026-03-16)Build and publish targeted GPU or CPU builds; the post reports a CUDA PyTorch runtime “about 60% the size” of the nixpkgs onehttps://flox.dev/blog/gpu-optimized-pytorch-builds-made-easy-with-flox-and-nix/
Triton, vLLM, llama.cpp (2026-03-16)Declarative model-serving environments for NVIDIA Triton, TensorRT, vLLM and llama.cpp that run on laptops, DGX Spark, Slurm or Kubernetes GPU nodeshttps://flox.dev/blog/production-model-serving-using-nvidia-triton-vllm-llamacpp-with-flox/
Slurm (2026-06-15)A nanoGPT training environment pinned with Nix and carried from a MacBook to Slurm; “Platform teams maintain just one environment for Slurm (training) and Kubernetes (prod)”https://flox.dev/blog/training-nanogpt-on-slurm-with-a-nix-pinned-environment/
AI stacks as code (2026-07-15)Reference patterns, including four modular AI environments composed into a declarative RAG stack, and a llama.cpp serving environment that answers questions about Flox docs and blog postshttps://flox.dev/blog/ai-stack-as-code-define-it-once-run-it-anywhere/
Kubernetes without containersFlox packages mounted into pods from the node cachehttps://flox.dev/kubernetes/

The CUDA tutorial frames environments as “reproducible collections of tools, environment variables, services, and setup scripts” that give the same setup in local development, CI, a container or a deployed service (https://flox.dev/docs/tutorials/cuda). When a customer asks about ML hand-offs, lead with that portability claim and then ask which hand-off hurts most: laptop to cluster, cluster to production, or one team to another.

Say: “The same environment runs on the laptop, on Slurm and on a GPU node.” Ask: “Which CUDA version and GPU generation do your clusters run, and where does PyTorch come from today?”

Not settled by these sources: GPU support on any given customer driver stack. The CUDA page covers redistribution and cache use, not driver compatibility [verify] (https://flox.dev/docs/concepts/cuda-catalog-and-cache).

← CLI, services, CI and tokens · Containers and Kubernetes →