Jev Fastapi

A self-hosted System One API billed as drop-in compatible with TypeSafe Jev: a bidirectional encoder with direct decision heads returns calibrated probabilities and structured choices in a single forward pass of about 33 ms.

An open-source, non-autoregressive System One decision engine inspired by TypeSafe AI: it evaluates state context against typed questions in a single forward pass, returning probability distributions and calibrated confidence without generating text.

A small FastAPI demo that triages customer support tickets with Jev as the decision layer: one request returns typed answers for department, urgency, refund intent and human-review need, while Python rules route and LangChain can draft a reply.

A technical knowledge assistant that runs adaptive RAG over 11 technical books (2,810 chunks) with hybrid dense and BM25 search, reciprocal rank fusion and a JEV System-1 decision engine over a local vector database.

Scaffold project for implementing Jev as a gateway integration layer, demonstrated by a FastAPI service that answers weather questions for a City/State: an SLM works out intent, location and date, Open-Meteo supplies facts, and an LLM writes the answer.

Drop-in guardrails for any AI app, powered by Jev: it asks narrow yes/no questions about text, such as whether something is a prompt injection, gets a probability, and lets the app allow, review or block in milliseconds, with adapters for Gemini, Anthropic, OpenAI and FastAPI.

A GitHub App that routes pull-request review work through Jev: it ingests the webhook, normalizes the PR context, gets a LOW, NORMAL or SPECIALIST decision, and a deterministic policy layer maps it to auto-approval, a review request or a check-run update.

Generative painting instrument where Jev picks paint materials and colour palettes for 105 procedural sketches: you select part of a sketch, and the chosen material flies in and paints itself. FastAPI + React + TypeScript.

Two runnable demo apps that send identical typed questions to OpenAI Decisions (gpt-6-luna) and TypeSafe Jev (jev-latest) in parallel: a support ticket router over 20 labelled tickets and a cloud alert triage over 15 labelled alerts, with thresholds in plain Python making the final calls.