Agent-Workflow 0.11 has a built-in provider, so why keep this?

Keep this package for the standalone adapter and skill-evaluation surface. The host built-in covers routing questions only; the three skill behavior, completeness, and actionability questions stay package-specific.

Can this adapter override routing or approve work by itself?

It cannot. Direct commands return no_action receipts and the provider only offers evidence; thresholds, routing, lifecycle, review, and acceptance stay with the host, with a shadow comparative mode available.

Which install extra gives live calls versus offline checks?

Install the base package for discovery and receipts, add the typesafe extra only for live calls, and the eval extra for comparative evaluation. TYPESAFE_API_KEY is read only by the live adapter.

What happens when credentials or the service fail mid-decision?

Missing credentials, unavailable SDK, uncertain answers, and service failures each map to distinct statuses with no-action fallback, and keys never enter projections, receipts, or logs.

What does the adapter return instead of routing decisions?

Versioned, secret-free semantic receipts wrapping Choice, Noul, and Score evidence. Direct commands return advisory no_action receipts, and the host decides whether to apply them or keep them in shadow mode.

What are the two question sets the adapter defines?

routing/v1alpha1 asks a task-class Choice, an interaction-required Noul, and a semantic-risk Score. skill-evaluation/v1alpha1 asks a behavior-satisfied Noul plus completeness and actionability Scores.

How do I enable TypeSafe API call logging?

Set TYPESAFE_API_CALL_LOG to a JSONL file path. Logging is disabled by default; each entry records the bounded projected input, question set, normalized output or exception class, request hash, model, status, and duration, with secrets redacted.

CommunityJev Agents & MCP

Agent Workflow TypeSafe AI

Standalone Agent-Workflow adapter that projects bounded redacted evidence into typed Jev choice, noul, and score questions and returns secret-free advisory receipts.

See highlights

Open github.com

Stars: 1
Language: Python
License: APACHE-2.0
Published by: Community
Author: ngallodev-software
Last verified: Oct 2026

ngallodev-software/agent-workflow-typesafe-ai

Organization repository on GitHub

View on GitHub

Standalone TypeSafe AI/Jev semantic-evidence adapter for bounded Agent-Workflow decisions, explicit uncertainty, compatibility evaluation, and host-owned policy.

Forks: 0
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.Homepage

About this project

Standalone TypeSafe AI/Jev semantic-evidence adapter for bounded decisions, explicit uncertainty, and host-owned policy. Portfolio case study · JevHunt listing · Implementation skill. What it is: a standalone TypeSafe AI/Jev compatibility adapter that turns bounded task/skill evidence into typed Choice, Noul, and Score receipts. Why it matters: semantic judgment can be added without handing workflow authority to a model.

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

Tagscommunity github

Works forJev for Classification

Used inJev for Software & DevTools

Languages: Python

Highlights

Advisory routing and skill-evaluation question sets with versioned secret-free semantic receipts.
Exact-version compatibility check before live requests, plus compatibility and doctor diagnostics.
Opt-in JSONL call logging with redaction; disabled by default and treated as sensitive when enabled.
Two advertised modes: eligible policy consumption versus shadow comparative evidence capture.

Watch out

Version 0.1.2 targets Agent-Workflow 0.10.0-0.10.2 only; 0.11.x has its own built-in provider, so check the compat contract first. 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

wald-4b

Open-weight 4B decision model on Qwen3.5-4B-Base serving a Jev-compatible /v1/systemone API, returning per-option probabilities with adjustable thinking.