Is this a tool I install, or a pile of write-ups?

A pile of write-ups, by design. The repository describes itself as small, independent projects for experiments, research and investigations, each one a dated directory with its own README and tags. There is nothing to install and no shared runtime; you read the lab you care about.

How do I get one lab without cloning everything?

The README gives the exact recipe: git clone with --depth 1 --filter=blob:none --sparse, then git sparse-checkout set for the shared root files plus the one lab directory you want. You still get the shared mise tasks and Makefile, so the lab's task runner works without the rest of the tree.

Which lab should I read if I care about Jev?

There are two. One is an evaluation lab covering API behavior, input limits and multilingual evaluation, tagged with typesafe, jev, tokenizer, long-context, multilingual and api. The other is a safety lab on moderation and pre-execution command risk checks, tagged with typesafe, jev, moderation, safety, classification, calibration and agent.

What do these labs assume about my machine?

The shared tooling is mise tasks driven from the repo root, and the labs that are listed lean heavily on Apple Silicon, MLX and Impeller targets, so the environment is not generic Linux. Check the individual lab README before assuming a task will run, since each one pins its own setup.

What license is the labs repository under?

MIT, per the license badge in the README.

How are the labs organized?

Into date-named directories under YYYY/MM/DD/{slug}, each with its own README and shared tag pages, with mise tasks and a Makefile at the repository root that every lab inherits.

What broad topics do the labs cover besides Jev?

Flutter and Impeller bug reproductions, MLX and quantization benchmarks, an MCP 1 to MCP 2 migration, WebRTC orchestration experiments, and video model ports, alongside the two Jev labs.

CommunityJev Practices & Patterns

Labs

A sparse-checkout research workspace of independent lab write-ups, two of which cover Jev: one on API behaviour, input limits and multilingual evaluation, another on moderation and pre-execution command risk checks.

See highlights

Open github.com

Stars: 0
Language: Dart
License: MIT
Published by: Community
Author: kiarina
Last verified: Oct 2026

kiarina/labs

User repository on GitHub

View on GitHub

Small, independent projects for experiments, research, and investigations.

Forks: 0
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

Small, independent projects for experiments, research, and investigations. Use a shallow, partial clone with sparse checkout to fetch only one lab and the shared tasks: Connecting N orchestrator apps over WebRTC: one brain drives workers on many bodies, every console shows the same state - flutter, dart, webrtc, flutter-webrtc, data-channel, codex, codex-app-server, claude, claude-agent-sdk, kiapi, multi-agent, orchestration, distributed, macos.

Quoted from the project's README. Read the full README.

Tagscommunity jevbest catalog-import

Works forJev for Classification Jev for Moderation Jev for Evaluation

Used inJev for Software & DevTools

Languages: Python

Highlights

Clone one lab instead of the repo with a shallow, blob-filtered clone plus git sparse-checkout set for the shared tasks and a single lab directory.
The Jev evaluation lab covers API behaviour, input limits and multilingual evaluation, tagged typesafe, jev, tokenizer, long-context and multilingual.
The Jev safety lab covers moderation and pre-execution command risk checks, tagged typesafe, jev, moderation, safety, calibration and agent.
Shared mise tasks and a Makefile sit at the root alongside the labs, so each lab inherits the same task runner.

Watch out

This is a workspace of unrelated experiments rather than an installable tool, and the two Jev labs are two write-ups among dozens of others spanning Flutter and Impeller bug reproductions, MLX and quantization benchmarks, an MCP 1 to MCP 2 migration, and video model ports. The labs are partitioned into date-named directories with per-lab READMEs and shared tag pages, so the entry point is an index rather than a command you run. To read a single lab you are expected to use the shallow sparse-checkout flow rather than pulling the whole tree. The listing here is the repository index, so the substance of each lab body lives in its own README rather than at the root. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

jev-on-a-laptop

Unofficial laptop study of parallel constrained decoding: prefill once, share the KV cache across fields, and decide all schema fields in one batched forward pass without writing JSON.