Article
CommunityJev Practices & Patterns
Jev in practice: typed decisions, scoped authority
Tenuo's engineering write-up of safe-upgrade, a dependency-upgrade agent built on Jev, LangGraph and Tenuo: Jev turns repository evidence into typed probabilistic choices, the graph routes on confidence, and each action gets a short-lived task-scoped warrant.
- Format
- Article
- Reading time
- 2 min
- Published by
- Community
- Author
- Tenuo Engineering
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Languages
- TypeScript
Highlights
- Design rule from the post: keep judgment, control flow and authority separate across Jev, LangGraph and Tenuo.
- Jev returns choice, confidence and a probability per option; low confidence or malformed output takes a deterministic fallback.
- The candidate action set is computed before the request and validated again after it, in ordinary TypeScript.
- Workers get narrowed warrants: smaller tool sets, run_check restricted to tests, 600-second TTL, terminal so they cannot delegate.
- Built with @typesafe-ai/sdk; a run shows authoring a regression test chosen at confidence 0.82.
Quickstart
ts
import { choice, TypeSafeClient } from "@typesafe-ai/sdk";
const client = new TypeSafeClient();
const criteria = Object.fromEntries(
input.eligibleActions.map(({ action, reason }) => [action, reason]),
);
const result = await client.systemOne({
state: JSON.parse(JSON.stringify(input)),
questions: {
next: choice("Choose the eligible action that most directly reduces the unresolved risk in this upgrade. Every option is already permitted; pick the most useful one.", criteria),
},
});Watch out
Grounded in the tenuo.ai engineering post captured on 2026-10-15; the article is truncated in this batch, and the example run is illustrative, with no published benchmarks for the upgrade agent's decision quality.
Questions about this entry
Short answers drawn from the linked project or page.
A dependency-upgrade agent built on Jev, LangGraph and Tenuo: it interprets release evidence against a repository, decides what work is needed, changes code, and wraps each resulting action in scoped authority.
Whether a release note affects how the repository uses the package, whether an existing test would catch the break, which eligible step reduces the most risk, and whether the final patch addresses each migration finding.
As a small typed decision: the Choice response carries the selected action, a probability distribution and confidence; routing code validates the label against the candidate list and falls back deterministically below a confidence threshold.
Low confidence, a malformed response or a transport failure all send the workflow down a deterministic route; the candidate set is fixed before the request and the returned choice is validated again afterward.
Task-scoped warrants: trusted routing maps Jev's chosen action to a worker, which gets a narrowed, short-lived, terminal session—for example read tools plus test-file writes, run_check restricted to tests, 600-second TTL.
Keep judgment, control flow and authority separate: Jev makes the typed judgment, LangGraph carries it through a durable workflow, and Tenuo constrains what the resulting action is allowed to do.
It is an engineering walkthrough with illustrative runs rather than a benchmark: the post was truncated in this batch, publishes no measured accuracy for the agent's decisions, and the example output is labeled as illustrating a run.
More Jev resources like this
An open lab for using Jev correctly next to Claude Code: a patched jev-router with per-turn model routing, a health-gated fallback with alerts, and write-ups of two silent bugs and of where tokens actually go.
Jev Practices & Patterns#community#catalog-import#logicrw
Community
pasblinn/jev-labTypeScript
pasblinn/jev-labTypeScriptA personal-assistant agent with 100 mocked tools that measures how many steps a task takes when the LLM picks the tool versus when Jev picks it before every model step.
Jev Practices & Patterns#community#typescript#agents
Community
vinilana/jev-eval-agentHTML
vinilana/jev-eval-agentHTMLA 34-page technical review and implementation framework by Av1dlive for using Jev's typed decisions inside agent workflows, separating observation, model judgment, deterministic policy, authorized action, and outcome verification.
Jev Practices & Patterns#community#catalog-import#logicrw
Open spike issue in the Orcha multi-agent orchestration platform: it proposes moving judgment out of expensive state-of-the-art models and onto Jev, which emits precise instructions for a cheaper executor to carry out, and requires findings, a cost model, and a recommendation as issue comments.
Jev Practices & Patterns#community#orchestration#cost-control
Community
open-orcha/orchaTypeScript
open-orcha/orchaTypeScriptPlugin for Claude Code and native Codex that layers Jev-assisted task routing, evidence search, context compaction and readable tool output on top of the host, where typed judgments inform policy but the host keeps ownership of permissions and execution.
Jev Practices & Patterns#community#jevbest#catalog-import
Community
MongLong0214/jev-gateTypeScript
MongLong0214/jev-gateTypeScriptA TypeScript starter for the layer around the model: decision contracts, confidence thresholds that route to auto, fallback or human review, deterministic mock providers, and an offline evaluation harness.
Jev Practices & Patterns#community#awesomejev#catalog-import
Community
hamakyo/jev-starterTypeScript
hamakyo/jev-starterTypeScriptFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.