Official repositories for working with Jev
The official documentation links to the typesafe-ai GitHub organization. These repositories serve different jobs. Start with the SDK for your language; an evaluation adapter or agent skill is optional.
| Repository | What it provides | Use it when |
|---|---|---|
| typesafe-sdk-python | Python API client | Your application uses Python |
| typesafe-sdk-js | JavaScript / TypeScript API client | Your application uses Node.js |
| skills | Coding-agent guidance for TypeSafe | You want an agent to consult the API workflow |
| system-one-adapter-python | A System One-style interface backed by other LLM APIs | You want to compare model approaches |
Each link goes to the upstream project. Jev Fieldguide does not maintain these repositories or distribute modified packages.
Install the client for your language
The Python reference documents typesafe-sdk; the JavaScript reference documents @typesafe-ai/sdk. Choose one for the project you are building:
python -m pip install typesafe-sdknpm install @typesafe-ai/sdkThe quick-start requirements are Python 3.10 or later and Node.js 20 or later respectively. Supply TYPESAFE_API_KEY through the environment. Installing the client alone does not create a TypeSafe account or API key.
For your first connection, the HTTP walkthrough exposes the request fields without adding a language-specific abstraction. Move to an SDK when its typed answers and error handling are useful to your app.
An agent skill and an adapter solve different problems
The official agent-skill guide explains how coding agents can use TypeSafe-specific guidance. It is documentation and workflow context, not a local copy of Jev or an alternative account system.
The comparison adapter calls other model providers through a TypeSafe-like interface. Seeing an OpenAI provider in that adapter does not mean Jev itself is an OpenAI model. Use our benchmark checklist to keep provider, model, and evaluation settings visible.
Does GitHub mean Jev is open source?
A public client repository exposes the code used to call an API. That fact alone does not establish that the hosted model’s weights, training code, or inference server are public.
We verified the repositories above, but did not verify an official downloadable Jev model or a supported local inference package in the reviewed materials. Use the access and download guide for the distinction. Do not assume a third-party project with “Jev” in its name is an official model release.
A practical repository checklist
Before adding a dependency, compare the package name with the official SDK documentation, read its current requirements, and check the license for the repository you intend to use. Save the installed version in your project’s lockfile so another developer can recreate it.
For an evaluation, also record the model returned by the API. A pinned client library and a moving model alias are different sources of change. Keep your API key outside committed source files.
Sources checked September 23, 2026. This is an independent guide; access, packages, and provider terms can change.
Sources & verification
- Python SDK documentation ↗
- Python repository ↗
- JavaScript SDK documentation ↗
- JavaScript repository ↗
- Official agent skill ↗
- Agent skill repository ↗
- System One comparison adapter ↗
Checked September 23, 2026. This guide summarizes documentation; it is not an independent benchmark. Provider details can change.