Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

decido

Python library for typed decisions with Jev or your own provider, adding validated distributions, batching, budgets, ranking and records.

LibrarybetaMIT
View on GitHub ↗

Overview

Decido lets you ask typed questions about evidence, keep every probability, and use the results in ordinary Python. It wraps a small provider contract, starting with TypeSafe's Jev, and adds validated distributions, batching within provider limits, request budgets, ranking and exportable decision records. Optional extras include Playwright web crawling and local MCP tools, plus a CLI that takes JSON input.

The README is careful about semantics: keep missing evidence separate from neutral evidence (an unknown option), model probabilities are not automatically calibrated, and independent questions do not form a joint distribution. A bundled lexical provider lets you exercise the API offline, with scores that are heuristics rather than semantic judgments. The CLI accepts JSON files or stdin for its decide and rank commands.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich team should handle this? Use unknown if the evidence is unclear.billing · technical · unknown

How it works

  1. Start a session with a provider (TypeSafe or custom)
  2. Call decide with state and typed questions
  3. Read typed answers and full distributions
  4. Rank, budget and export decision records

Strengths

  • Provider-agnostic contract
  • Request budgets and batching
  • Offline lexical provider for testing
  • Explicit guidance on unknown vs neutral evidence

Caveats

  • Probabilities are not automatically calibrated
  • Jev requests incur charges on your account

Getting started

pip install "decido[typesafe]", export TYPESAFE_API_KEY, or try offline with pip install decido and decido rank.

At a glance

Type
Library
Status
beta
Language
Python
Runtime
library, CLI, MCP tools
Has code
yes
License
MIT

Use cases and integrations

How curated lists describe it

  • Probabilistic decisions for Python. Use Jev or bring your own provider; crawl with Playwright.

    hellogumbo/awesome-jev

Frequently asked questions

Can I use it without Jev?

Yes, supply your own provider or use the offline lexical provider for testing.

Are probabilities calibrated?

The README notes they are not automatically calibrated.

What extras are included?

Ranking, decision records, Playwright crawling and local MCP tools.

Related projects

  • Jev Studio

    pip-installable Jev kit: a CLI for verify, screen, classify, extract, route and rerank, plus an MCP server for the primitives.

    MCP serverbetaPython
  • ruby_llm-providers-typesafe

    RubyLLM provider gem for TypeSafe Jev: block-style typed evaluations with noul, choice and score, plus RubyLLM.rerank support.

    ProjectbetaRuby
  • Jev Recipes

    248 small installable TypeScript recipes that each make one typed Jev decision: route, rerank, gate, grade, compare or label.

    LibraryproductionTypeScript
  • jev-starter

    TypeScript library of decision contracts, confidence thresholds, fallbacks, human review and evaluation patterns on top of the TypeSafe SDK.

    LibrarybetaTypeScript
  • spring-ai-typesafe

    Java SDK for TypeSafe's System One API plus Spring AI integrations using Jev as judge, guardrail, RAG filter/reranker and tool index.

    LibrarybetaJava
  • ego-jev

    Agent skill where Ego Lite drives the browser and Jev resolves semantic choices: targets, filtering, ranking and evidence checks.

    ProjectbetaMarkdown + TypeScript