typesafe-ai GitHub organisation and repos

TL;DR Four repos matter for building with Jev: typesafe-sdk-python (pip install typesafe-sdk, v0.6.0), typesafe-sdk-js (npm install @typesafe-ai/sdk, v0.6.0), skills (the agent skill / Claude Code plugin), and system-one-adapter-python (system-one-adapter, v0.1.4 — run the same API against an LLM for comparison). Everything else in the org is infrastructure or a fork.

Org listing

Org: https://github.com/typesafe-ai. Listing observed during ingestion on 2026-09-17; "Last push" is from that listing.

Only the four high-relevance repos were mirrored into raw/github/; pinned commits are recorded in raw/MANIFEST.json.

typesafe-sdk-python

Install and first call, verbatim from the repo README:

uv add typesafe-sdk

from typesafe_sdk import Choice, TypeSafeClient

with TypeSafeClient() as client:
    response = client.system_one(
        state={"document": "I was charged twice. Please fix this ASAP."},
        questions={
            "category": Choice(
                instructions="What is this ticket about?",
                criteria={"billing": None, "technical": None, "other": None},
            ),
        },
    )

print(response.choices["category"].choice)

Set TYPESAFE_API_KEY in the environment first. Docs: https://docs.typesafe.ai/sdk/python/. Note there is also a typesafe-ai 0.1.0 package on PyPI that is only a redirect shim depending on typesafe-sdk; the unrelated typesafe 0.9.1 package is not TypeSafe AI's.

typesafe-sdk-js

npm install @typesafe-ai/sdk

import { choice, TypeSafeClient } from "@typesafe-ai/sdk";

const client = new TypeSafeClient();
const response = await client.systemOne({
  state: { document: "I was charged twice. Please fix this ASAP." },
  questions: {
    category: choice("What is this ticket about?", {
      billing: null,
      technical: null,
      other: null,
    }),
  },
});

console.log(response.answers.category.choice);

"Answer types are inferred from your questions." Docs: https://docs.typesafe.ai/sdk/javascript/. The repo also ships a jsr.json, so a JSR publication target exists (contents not ingested).

skills

Install, verbatim:

claude plugin marketplace add typesafe-ai/skills
claude plugin install typesafe@typesafe-ai

npx skills add typesafe-ai/skills --skill typesafe-ai

"Select your agent when prompted. Installation is project-local by default; add -g to install globally." In Claude Code the skill can be invoked explicitly with /typesafe:typesafe-ai. Details in The typesafe-ai agent skill and Claude Code plugin.

system-one-adapter-python

"A drop-in replacement for typesafe_sdk's system_one evaluation API, backed by LLM APIs instead of TypeSafe. Useful for comparing TypeSafe against an LLM on cost/speed/intelligence." Provider SDKs are optional extras:

pip install 'system-one-adapter[openai]'      # OpenAI-compatible providers
pip install 'system-one-adapter[anthropic]'   # native Anthropic

from system_one_adapter import SystemOneAdapterClient, Noul, Score, Choice

client = SystemOneAdapterClient(
    structured_outputs=True,
    llm_answer_mode="probabilities",
    normalize_probabilities=True,
)

response = client.system_one(
    state="This book was a delight to read.",
    questions={"positive": Noul(instructions="The book review is positive.")},
    provider="openai",
    model="gpt-4o-mini",
)

This is the same wrapper the launch blog calls the "System One LLM" wrapper used to constrain LLMs in TypeSafe's published evals. Full option table and response extensions in system-one-adapter: LLM-backed drop-in for TypeSafeClient.

Forks and infrastructure repos

• LLaDA (fork, last push 2025-06-17) and vllm (fork, 2025-05-23) — forks of third-party projects, predating the public launch. Nothing in the sources says what TypeSafe changed in either, so any link to Jev's architecture or serving stack would be speculation.
• daggerverse (Python, 2026-09-09) — Dagger modules; build/CI tooling.
• pulumi-clickhouse (Go, 2026-07-08) — infrastructure-as-code for ClickHouse.
• Overwatch (Python, 2026-09-03) — no description published; purpose unknown.
• typesafe-ai.github.io (HTML, 2026-06-04) — GitHub Pages site for the org.

Related

• Community repos: what people built and how they use Jev — third-party repos built on Jev (community tier)
• Python SDK: install, clients, system_one() / JavaScript/TypeScript SDK: install, client, choice/score/noul — the SDK contracts
• The typesafe-ai agent skill and Claude Code plugin — the skill in detail
• system-one-adapter: LLM-backed drop-in for TypeSafeClient — the adapter's full API
• TYPESAFE_* environment variables across SDKs — TYPESAFE_* variables the SDKs read
• TypeSafe AI (company) — the company

Sources

• raw/github/typesafe-sdk-python/README.md, pyproject.toml (https://github.com/typesafe-ai/typesafe-sdk-python)
• raw/github/typesafe-sdk-js/README.md, package.json (https://github.com/typesafe-ai/typesafe-sdk-js)
• raw/github/skills/README.md, .claude-plugin/plugin.json, .claude-plugin/marketplace.json (https://github.com/typesafe-ai/skills)
• raw/github/system-one-adapter-python/README.md, pyproject.toml (https://github.com/typesafe-ai/system-one-adapter-python)
• Org listing and package registry facts recorded in CLAUDE.md "Facts collected during ingestion" (2026-09-17)