OPEN SOURCE · CURATED REPOSITORIES
Curated Jev SDKs, tools, demos, integrations, benchmarks, and research projects. Repository metrics come from GitHub; titles, summaries, and categories are editorial.
Ranked by GitHub Stars
Ranked by recent Star growth and update activity
allebeeallebee/pytest-jevpytest-jev is a pytest plugin that adds semantic assertions for LLM outputs, using TypeSafe's Jev model to return calibrated probabilities. It supports holds/lacks/expect/choice/score checks, caching, configurable thresholds, and can run via OpenRouter.
eriestraeriestra/blockly-jevA Blockly extension that exposes TypeSafe Jev / System One judgments (Noul, Choice, Score) as visual blocks, with a live demo, tutorials, and automated tests against the real model.
shitianfangshitianfang/jev-useA Claude Code / Codex / pi plugin that routes agent steps needing no text output to TypeSafe's Jev system-one model, with measured latency and typed escalation.
buberlobuberlo/dsh-jevA DeepSeek Harness plugin that uses TypeSafe Jev/System One for dynamic tool selection, pre-execution call assessment, model routing, and skill routing. Includes runnable examples, 120 tests, 25 evaluation fixtures, and a with/without-Jev benchmark.
prasanthjprasanthj/duckdb-jevA high-throughput native DuckDB extension for TypeSafe/Jev, enabling batched and streaming classification, scoring, and semantic predicate evaluation from SQL.
ajensenwaudajensenwaud/hermes-jev-pluginExposes TypeSafe Jev System One decision APIs (jev_check, jev_route, jev_score, jev_evaluate) as Hermes Agent tools. Includes schemas, a thin API client, usage skill, and integration tests.
LXBWOWLXBWOW/dsh-completion-supervisorA DSH plugin that verifies coding agents' 'done' claims by combining deterministic repository evidence, one batched TypeSafe Jev/System One assessment, and a pure policy. Fail-open; includes analysis tools, determinism probes, and real-fixture tests.
CrowdLinkerCrowdLinker/JevPromptCoachClaude Code plugin that scores coding-agent prompts with TypeSafe's Jev model and tracks whether your prompting improves. Default mode adds zero latency.
Wang-auspiciousWang-auspicious/codex-jev-compactionA Codex plugin that calls the TypeSafe Jev systemone API (model jev-latest) to decide which verified read-only tool evidence should remain in a handoff. Includes a skill, MCP tool, CLI, offline tests, and measured local fixture results.
seahskyseahsky/kelpieThis Claude Code plugin uses TypeSafe AI's Jev API to decide when to delegate work to cheaper subagents and to route model/effort per spawn. It includes a Jev-powered prompt triage, spawn gate, and a benchmark report.
hoshinodishoshinodis/opencode-intent-gateTypeSafe Jev-powered OpenCode plugin that detects ambiguous user requests and asks clarifying questions before the agent acts.
Wang-auspiciousWang-auspicious/pi-jev-compactionA Pi extension that uses TypeSafe AI's Jev model via the System One API to decide which read-only tool pairs to drop during context compaction, preserving recent messages and falling back to native compaction when Jev is unavailable.
