User repository on GitHub
Optional typed TypeSafe AI Jev decisions for OpenClaw, with SecretRef credentials and strict API validation.
- Forks
- 0
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Integrations & Adapters
Openclaw TypeSafe AI
An OpenClaw plugin that registers a single optional tool, typesafe_decide, forwarding caller-supplied text or JSON state plus typed noul, choice and score questions to the System One endpoint, with jev-latest as the default model.
- Stars
- 0
- Language
- TypeScript
- License
- MIT
- Published by
- Community
- Author
- Olli0103
- Last verified
- Oct 2026
About this project
openclaw-typesafe-ai exposes one optional agent tool, typesafedecide, labelled TypeSafe Decide. The plugin ID is typesafe-ai. It sends explicit text or JSON state and typed questions to POST with jev-latest` as the default model. Jev returns decisions over predefined answer spaces. It does not generate conversation, prose or code. This package therefore registers a tool and no model provider, model picker entry, hook or background service.
Quoted from the project's README. Read the full README.
Works forJev for Classification
Used inJev for Software & DevTools
- Languages
- TypeScript
Highlights
- One tool and nothing else: no model provider, model picker entry, hook or background service is registered.
- The request contract is a closed union - noul, choice with 1 to 255 options, score with 2 to 10 indexed descriptions.
- Compatibility is pinned to OpenClaw 2026.9.4 on Node 24.20.0 and macOS arm64; no other release is claimed.
- The key is stored as an environment reference through the CLI builder, not as a literal string in config.
Watch out
Compatibility is a point-in-time host test against two specific builds, not a range, so a different OpenClaw or Node release is untested by the author's own account. The example request and response payloads in the README are described as illustrative fixtures rather than live API observations. No lint tool is configured, and .npmrc pins npm's legacy peer resolver because the modern one crashed on the OpenClaw and Vitest dependency graph, which is a maintenance signal worth knowing about. The comparison row describing the other community package being rejected before load is explicitly framed as a host compatibility observation, not a security judgement about that project. Installation can restart a managed Gateway, and no live installation or Gateway restart was performed during development. Copy reviewed against the repository README on 2026-10-04, not from running it.
Questions about this entry
Short answers drawn from the linked project or page.
You cannot run both. Both packages use the plugin ID typesafe-ai, so a normal installation will not load them side by side; the README tells you to inspect the existing plugin inventory before installing and warns that --force also permits overwriting an existing target. The choice is about design: this one requires an agent to call the tool explicitly over data it supplies, while the community package uses runtime hooks for group triage, tool guardrails, model routing, input audit and compaction.
The tool is optional in both runtime registration and manifest metadata, and no profile adds it automatically, so enabling the plugin is only step one. You also add typesafe_decide to tools.allow, preserving existing entries since OpenClaw rejects allow and alsoAllow in the same scope, include typesafe-ai in plugins.allow if that list is configured, and run config validate plus plugins inspect with --runtime to confirm. Agent, provider and sandbox restrictions still apply on top.
No, if you use the supported path. openclaw config set plugins.entries.typesafe-ai.config.apiKey with --ref-provider default --ref-source env --ref-id TYPESAFE_API_KEY stores a reference reading source env, provider default, id TYPESAFE_API_KEY rather than the key itself. The schema also accepts store, file and exec references. The README warns that a terminal environment variable is not inherited by an already running launchd service, and says never to paste the key into chat, arguments or example config.
The README reports checks against the installed 6f80261 build copied into an isolated test environment and against the published npm build 3a9d69d after a clean dependency installation, on one OpenClaw release, one Node version and one architecture. The host test invokes the real loader and CLI against disposable configuration, state and cache directories with an obviously fake credential; it installs no plugin and starts no Gateway. No other release is claimed compatible.
Only state (a string, JSON object or array), a nonempty questions map, and an optional model defaulting to jev-latest. Each question needs a type and instructions, and the union is closed: noul, choice with 1-255 options, or score with 2-10 descriptions indexed from zero.
No. Both use the plugin ID typesafe-ai, so a normal installation cannot load them side by side. Inspect existing plugin inventory before installing and do not overwrite an unrelated plugin.
Use Node 24.20.0, pinned by .nvmrc, then npm ci --ignore-scripts and the typecheck, plugin:build, test, plugin:validate and test:host scripts. The .npmrc uses npm's legacy peer resolver because the modern one crashed on the OpenClaw and Vitest dependency graph.
More Jev resources like this
69GitHub stars
Provider-neutral TypeScript runtime for System One decisions: write choice, noul, and score code once, then run it on TypeSafe Jev, local Reflex, any compatible /v1/systemone endpoint, or a custom provider.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
iamaamir/system-oneTypeScript
iamaamir/system-oneTypeScript38GitHub stars
File-first project notebook where many agents share context in .context/ files while Jev only judges the next step and records it as JSON.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
pedrocivita/tocketTypeScript
pedrocivita/tocketTypeScript15GitHub stars
Uncertainty-aware semantic control flow and routable procedures for Effect, with type-safe patterns and policies over Effect Decision and DecisionModel where uncertain answers stay explicit branches.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
doeixd/discernTypeScript
doeixd/discernTypeScript12GitHub stars
Command-line checker that asks Jev whether each selected source file violates a plain-English convention, returning file-level probabilities for an agent or CI step to act on.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
iamtoomas/JevLintTypeScript
iamtoomas/JevLintTypeScript21GitHub stars
Language and VM for agent programs that wait for approval, survive restarts, and replay portable receipts without recalling the model.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
hraness/algalTypeScript
hraness/algalTypeScript8GitHub stars
Shadcn registry search block that shows lexical hits instantly then re-ranks the top candidates with Jev intent scoring in one request, with keyword fallback on failure.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
kylemclaren/jevsearchTypeScript
kylemclaren/jevsearchTypeScriptFrom the Jev community
Posts from builders shipping with Jev right now.