ask-jev
Zero-dependency Python CLI that hands small semantic judgments and verbatim passage selection to Jev; the calling agent keeps authority.
Overview
ask-jev is a standard-library Python CLI for AI coding agents and shell pipelines that delegates small semantic judgments to TypeSafe's System One model. The calling agent keeps reasoning, verification and action authority; ask-jev only advises, and it is designed to fail open.
Its commands map to the official primitives: choose (one outcome from exclusive alternatives, with full distribution and confidence), check (probability one proposition is true given evidence, not a severity or relevance measure), score (rating across 2 to 10 ordered descriptions, explained as a probability-weighted mean of level positions with a normalized score), batch (1 to 32 independent questions over one state) and purify (selects original passage spans with local evidence, no generated summary). It requires Python 3.10+ on POSIX, TYPESAFE_API_KEY and an explicit HARNESS_JEV_ALLOW_REMOTE=1 opt-in before any remote call. Docs are in English and Chinese, with code-triage patterns and an output contract.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which change category fits? | — |
| Yes / no | Is this proposition true given the supplied evidence? | yes · no |
| Score | How severe is this defect? | — |
| Extract | Which original passages answer the question? (purify) | — |
How it works
- Agent or script calls choose, check, score, batch or purify
- CLI requires explicit remote opt-in and a key
- Jev returns typed probabilities (or selected spans)
- Agent decides what to do; failures fail open
Strengths
- No third-party dependencies
- Clear guidance on which primitive to use
- Explicit opt-in before data leaves the machine
- Verbatim extraction instead of generated summaries
Caveats
- Advisory only by design
- GPL-3.0 licence
- POSIX only
Getting started
export TYPESAFE_API_KEY=... and HARNESS_JEV_ALLOW_REMOTE=1, then python3 scripts/ask_jev.py choose --question 'Which change category fits?' ...At a glance
- Type
- CLI
- Status
- beta
- Language
- Python
- Runtime
- CLI / agent skill
- Has code
- yes
- License
- GPL-3.0
Use cases and integrations
How curated lists describe it
Python ecosystem: zero-dependency CLI that routes small semantic judgments — Choice, Noul, Score, batch questions, and verbatim passage extraction — to Jev for AI agents and CLI pipelines.
yibie/awesome-jevUltra-fast, fail-open advisory decisions and verbatim extractive reading view for AI coding agents and CLI pipelines
logicrw/awesome-jev-projects
Frequently asked questions
When should I use check versus score?
check is the probability a proposition is true; score rates on an ordered scale and is not a probability of truth.
Does it act on its own?
No, it only advises; the calling agent keeps authority.
Why the HARNESS_JEV_ALLOW_REMOTE flag?
To require explicit permission before any content is sent to TypeSafe.
Related projects
Go MCP server giving any client typed Jev decisions via OpenRouter or TypeSafe, with validated answers and no paid retries.
MCP serverproductionGoUnofficial hosted or local Jev MCP server giving agents a quick second opinion: it picks yes/no, scale or choice and returns probabilities.
MCP serverbetaTypeScriptRust CLI for Jev: ask yes/no, choice and rubric questions about text and branch on exit codes, JSON or MCP tools.
MCP serverbetaRustpip-installable Jev kit: a CLI for verify, screen, classify, extract, route and rerank, plus an MCP server for the primitives.
MCP serverbetaPythonPython library that converts a JSON Schema into Jev questions and returns JSON that validates against the original schema.
ServicebetaPythonAgent Client Protocol agent that exposes Jev Choice, Score and Noul decisions in IDEs with guided input and probability displays.
ProjectbetaTypeScript