jevbooks

← All projects

jarviscore-framework

Prescott-Data/jarviscore-framework · Homepage

Peer-to-peer multi-agent runtime with zero-trust credentials and durable state.

Agent Decisions80%Runner-up: Classification & Routing

What it is

JarvisCore is a Python framework for building AI agent fleets that plan, reason, execute code, browse the web, and connect to external services. It targets developers who need crash recovery, shared memory, and peer-to-peer discovery without a central orchestrator.

How it uses Jev

From 1.12, agents ask typed Choice, Score, and Noul questions through a decision client separate from the text model. The Kernel picks a specialist subagent by Choice, and each retrieved RAG passage is withheld from the generating model when its prompt-injection Noul exceeds 0.70.

Primitives:choicescorenoul

Technique worth stealing

Use a separate typed decision client for routing and safety gates, keeping the generative model out of those judgments.

Try it

pip install "jarviscore-framework[typesafe]", set TYPESAFE_API_KEY, then call self.decisions.evaluate.

View on GitHub

judged by Jevjev-1.13.0

Evidence

Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.

  • Jev-centricunclear0.41
  • Shows a System One patternyes0.80
  • Handles uncertaintyno0.38
  • Measuredno0.26
  • Runnableyes0.97
  • Worth recommendingunclear0.43
  • Model replicano0.21
  • Problem scopescore on a 0–2 scale1.73
  • About Jevyes0.77

Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.