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

decidealot

Docker service exposing local Laya, Von, a CLM head or hosted Jev through one TypeSafe-compatible HTTP and MCP endpoint.

MCP serverbeta
View on GitHub ↗

Overview

Decidealot runs decision models on your hardware or forwards to TypeSafe's hosted Jev, behind one TypeSafe-compatible /v1/systemone HTTP API and an MCP Streamable HTTP endpoint from the same container. Local providers include Laya, Von and a CLM projection head that sits over a configured Qwen3-8B embeddings endpoint.

At startup it downloads and verifies every enabled local bundle, then loads only the provider a request selects and unloads it after an idle period, with resource-aware batching. Answers are typed choice, score and noul with model probabilities. The quick start runs a locked-down CPU container (read-only filesystem, no capabilities, loopback only), and CUDA images are available. It suits developers who want one endpoint for agents and apps while switching freely between local and hosted models.

What the decision model decides

TypeQuestionAnswers
ChoiceTyped choice, score or noul questions forwarded to the selected model(inferred)—

How it works

  1. Run the Docker container with a model directory
  2. Send a /v1/systemone request naming a provider (laya, von, clm, jev)
  3. Decidealot loads the provider on demand and answers
  4. Idle providers are unloaded to free resources

Strengths

  • Local and hosted models behind one API and MCP
  • Lazy loading and idle unloading
  • Security-hardened container defaults

Caveats

  • CLM needs a separate embeddings endpoint
  • Model downloads at startup
  • Local model quality differs from hosted Jev

Getting started

docker run ... --publish 127.0.0.1:8080:8080 psyb0t/decidealot:latest, then POST to http://127.0.0.1:8080/v1/systemone with model laya.

At a glance

Type
MCP server
Status
beta
Language
Python
Runtime
Docker service (HTTP + MCP)
Has code
yes
Has tests
yes

Use cases and integrations

How curated lists describe it

  • Make the call. Run local Laya, Von, CLM or hosted Jev for typed classification and scoring through one HTTP and MCP API, with resource-aware batches.

    hellogumbo/awesome-jev

Frequently asked questions

Which models can it run?

Laya, Von and a CLM projection head locally, or hosted Jev.

Is it compatible with the TypeSafe SDK?

Yes, it exposes the TypeSafe /v1/systemone HTTP API.

Does it support GPUs?

Yes, CUDA images are documented alongside the CPU image.

Related projects

  • jev-mcp

    MCP server exposing Jev-style decision models served locally by Ollama's /v1/systemone as classify, yes/no and score tools.

    MCP serverbetaTypeScript
  • noulxp

    NoulXP: open standard and reference engine for packaging calibrated System One decision models so one runtime can run them all.

    ProjectbetaPython
  • jev-rs

    Rust engine that answers noul, choice and score from any LLM in one prefill via logprobs, wire-compatible with /v1/systemone and MCP.

    ModelbetaRust
  • opendecider

    Open calibrated decision models from 400M to 80B with Jev-compatible server, MCP server and clients, benchmarked against Jev.

    ModelbetaPython
  • edgejev

    Converts open Jev reproductions (Laya, Kev, NanoJev, PlayJev) to INT8 ONNX for offline CPU decisions at about 15.6 ms per question.

    ServicebetaPython
  • go-system-one

    Native Go runtime serving Gemma 4 12B as a Jev-like decision model with hand-tuned PTX kernels and no CGo or CUDA toolkit.

    ProjectbetaGo