Why use this toolkit instead of calling the endpoint directly?

Direct calls leave you to implement retries, typed errors, provider dispatch, batching, and per-call timeouts. This wraps POST /v1/systemone with retry policy, typed error mapping, a models listing, and header overrides.

Are the batched-call efficiency claims believable?

The digest reports live benchmarks with batched calls running much faster than sequential single-question calls while using fewer tokens. Treat that as directional and measure it on your own question sets.

What happens when some states fail in a large evaluation?

The Evaluator keeps going: per-state failures are captured in EvaluationRecord.error while summary aggregates per-question means plus latency and token usage. A usage ledger and circuit breaker cover accounting and resilience.

What should I check before wiring it into my own pipeline?

Plan for Python 3.10 or newer with httpx and pyyaml, plus matplotlib extras only for figures. Keys resolve from the environment, providers include hosted and self-hosted backends plus a proxy, and structuring decisions stays provider-agnostic.

Which providers can it talk to?

One wire contract across hosted jev (TypeSafe), self-hosted jeff, kev, localjev, and openthai-systemone, plus the openrouter proxy, selected with a global --provider flag or for_provider() and open_client() constructors.

What does the MCP server expose?

daf-jev serve exposes nine MCP tools (jev_ask, jev_evaluate, jev_models, jev_composite_score, jev_confidence_gate, jev_tiered_gate, jev_docs_verify, jev_posteriors_load, jev_reask_plan) plus a jev://docs/snapshot resource, over stdio.

Can Jev drive a Bayesian network?

Yes, as a factor source: elicit_cpts elicits every CPT in one batched ask and propose_structure proposes a topology via pairwise choices, while BayesNet runs exact local variable elimination plus MPE, ancestral sampling, and what-if sweeps.

CommunityJev Practices & Patterns

Daf Jev

Python client and decision toolkit for the Jev System One endpoint with noul, choice, and score primitives, batched asks, composition helpers, concurrent evaluation, and an MCP server.

See highlights

Open github.com

Stars: 7
Language: Python
License: MIT
Published by: Community
Author: docxology
Last verified: Oct 2026

docxology/daf-jev

User repository on GitHub

View on GitHub

daf-jev: composable Python toolkit for TypeSafe's Jev (System One) decision API — question builders, confidence gates, evaluator, calibration, CLI, MCP server, agent skill

Forks: 1
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

Modular, composable Python client and decision toolkit for the TypeSafe Jev (System One) API. One HTTP endpoint, three question primitives, and a set of pure-logic composition patterns built on top of the answers — plus a concurrent batch evaluation harness, an MCP server, a figure registry, and a reproducible manuscript pipeline. An explicit benchmark backend interface adds local priors/classifiers, local model servers and OpenRouter decision candidates, with prepared datasets, durable attempt receipts and…

Quoted from the project's README. Read the full README.

Tagscommunity jevbest catalog-import awesomejev

Works forJev for Classification Jev for Evaluation Jev for Tool calling

Used inJev for Software & DevTools

Languages: Python

Highlights

JevClient and AsyncJevClient wrap one endpoint with retries, typed errors, and per-call timeout and header overrides.
Pure composition helpers include composite scoring, confidence gating, routing, and tiered gating with deterministic fallbacks.
Evaluator runs a fixed question set over many states concurrently without aborting on per-state failures.
Also ships Bayes-net elicitation with exact inference, usage ledger, circuit breaker, CLI, and MCP tools.

Watch out

The public README is large and the digest is truncated, so this covers the client, composition, evaluation, and MCP surface visible in the digest. 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.

More Jev resources like this

shade-arena-jev-monitor

A fork of the SHADE-Arena public split that evaluates TypeSafe Jev as a monitor for covert agent sabotage, against Gemini 2.5 Flash and Pro. Thirteen hand-built transcripts across three tasks, 353 tool calls, and a full LaTeX report.