What is the R package coal?

An R client for Jev, TypeSafe's classification model. It evaluates a text state against typed questions and returns probabilities and confidence rather than generated text, built directly against the HTTP API.

Why build an R client at all?

The README explains that TypeSafe publishes Python and JavaScript SDKs only, so coal targets the HTTP API directly; it is a thin wrapper whose argument names match the HTTP field names.

From the package directory in R, run devtools::install_local('.'); while working on the package itself the README suggests devtools::load_all() instead.

How does it get the API key?

Put TYPESAFE_API_KEY in a gitignored .env file at the project root and load it with readRenviron('.env'). TYPESAFE_BASE_URL and TYPESAFE_DEFAULT_MODEL are also read when set.

What does the API feel like?

You write the questions once in a single spec object using helpers like jev_choice and jev_score, so wording and thresholds can be reviewed together, and the main function is vectorized over states, with raw responses staying reachable.

What example does the README show?

A support spec that asks which team should handle a text, with billing, technical and sales criteria, plus a frustration score question; the captured excerpt cuts off in the middle of that example.

What are the limits of this entry?

README and GitHub metadata reviewed 2026-10-17; the excerpt truncates mid-example, so the rest of the helpers, error handling and tests are not visible, GitHub lists no stars, and the license is marked NOASSERTION.

Jev directory

CommunityJev Repos & SDKs

coal (dhicks)

An R client for Jev built directly against the HTTP API, because TypeSafe publishes Python and JavaScript SDKs only: declare typed questions once with helpers like jev_choice and jev_score, run them vectorized over states, and keep the raw responses.

See highlights Open github.com

Stars: 0
Language: R
Reading time: 2 min
Published by: Community
Author: dhicks
Last verified: Oct 2026

dhicks/coal

User repository on GitHub

View on GitHub

Forks: 0
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.

Tagscommunity github r typesafe

Works forJev for Classification

Used inJev for Software & DevTools

Platforms: API
Capabilities: Open sourceRequires API key

Highlights

An R client for Jev built directly against the HTTP API, since TypeSafe publishes Python and JavaScript SDKs only.
Questions are declared once with helpers like jev_choice and jev_score; the main function is vectorized over states.
Install with devtools::install_local('.') or load a working copy with devtools::load_all().
Reads TYPESAFE_API_KEY from a gitignored .env, plus TYPESAFE_BASE_URL and TYPESAFE_DEFAULT_MODEL if set.
A thin wrapper: argument names are the HTTP field names and raw responses stay reachable.

Watch out

README and GitHub metadata reviewed 2026-10-17; the captured excerpt cuts off mid-example in the first spec, so the remaining helpers, error handling and tests are not visible; GitHub shows 0 stars and lists the license as NOASSERTION.

Questions about this entry

Short answers drawn from the linked project or page.

From the Jev guides

Original write-ups on what this entry is for.

The Open Reimplementations: Local Models That Speak System One

Sep 22, 2026

Closed weights have not stopped the ecosystem from rebuilding the shape in the open. API-compatible servers, retrained heads on small models, and inference-time tricks that skip generation entirely.

Jev Repos & SDKs#open-models#research#local

Read article

From the Jev AI model community

Posts from builders shipping with Jev right now.

Follow @typesafeai