wh000wh000/awesome-jev-live

What the repository is

Awesome Jev Live is a continuously‑updated “awesome‑list” that catalogs every public project that works with Jev, the System One decision‑making model released by TypeSafe AI. The list is refreshed automatically every two hours, pulling in new repositories, re‑grading existing entries, and showing the latest star counts, languages, and evidence level (official, observed, inferred, or unverified).

The repository contains:

• a pipeline that crawls GitHub, extracts metadata, and grades each entry.
• a static site (hosted at wh000wh000.github.io/awesome-jev-live) that presents the data in a searchable, multilingual UI.
• a curated “picks of the moment” table that highlights the top‑ranked project in each major category (official SDKs, community adapters, routing, evaluation, etc.).
• detailed sections grouping projects by function (SDKs, agent tooling, routing, benchmarks, applications, writing, etc.) and by implementation language.

The list is meant for developers who want to discover ready‑made Jev integrations, compare SDKs, or find research‑grade reproductions.

---

Who might find it useful

• Engineers building AI agents or workflow pipelines that need fast, typed decisions (e.g., routing, triage, scoring). The list points them to official SDKs (Python, TypeScript, Swift) and community adapters for many other languages.
• Researchers looking for open‑source reproductions, weights, or benchmark suites for System One models.
• Product teams evaluating Jev‑based guardrails, approval flows, or low‑latency gating mechanisms.
• Educators and students who want a single place to explore the ecosystem around a new class of LLM‑derived decision models.

---

How it works (high‑level)

1. Crawling – The repo’s CI workflow runs a script that queries the GitHub API for repositories mentioning Jev‑related keywords, topics, or containing the jev identifier.
2. Filtering & verification – Each candidate is examined for:
   • Presence of a public source tree (observed)
   • Explicit TypeSafe/​Jev branding (official)
   • Matching vocabulary without a manual check (inferred)
   • No confirmation (unverified).
3. Scoring – Stars, recent activity, and evidence grade are combined to produce a ranking per category.
4. Static site generation – The processed data is rendered into a searchable HTML page with multilingual README links and image/video thumbnails.
5. Automatic refresh – A scheduled GitHub Action triggers the whole pipeline every two hours, ensuring the index stays current.

---

Example entry (from the README)

• Project: typesafe-ai/skills
• Category: Official SDKs and developer tools
• Stars: 2,696 (official)
• Description: Agent‑skill definitions for building with TypeSafe’s System One API. It serves as a de‑facto specification of how the vendor expects Jev to be used.
• Evidence level: official (published by TypeSafe AI itself).

---

Limitations / caveats

• The index only reflects public GitHub repositories; private or non‑GitHub projects are not captured.
• “Observed” and “inferred” grades rely on automated heuristics and may include false positives.
• The list does not host any of the underlying code; it merely links to the original repositories.
• Because Jev is a relatively new, proprietary model, many entries are thin wrappers or experimental demos rather than production‑ready libraries.

---

Quick start for contributors

If you want to add or improve an entry:

1. Fork the repo and edit data/entries.json (or run the provided script to re‑run the crawler).
2. Submit a pull request; the CI will re‑run the grading pipeline and update the live site.
3. Follow the contribution guidelines in CONTRIBUTING.md for naming, tagging, and providing screenshots or videos.

---

Bottom line: Awesome Jev Live is a well‑maintained, automatically refreshed directory of tools and research around the Jev System One decision model, useful for anyone building or studying typed‑decision AI systems.