ClassOne: System 1 Decision Model Architecture

ClassOne is an open-source System 1 decision model architecture built on Google's Gemma 4 E2B (Effective 2B). It provides rapid, non-autoregressive, schema-constrained decision execution for classification, routing, verification, and scoring in a single forward pass.

Inspired by Daniel Kahneman's cognitive framework in Thinking, Fast and Slow and modern proper scoring rules (Brier, 1950; Gneiting & Raftery, 2007).

Key Features

• System 1 Architecture: Non-autoregressive decision model. Skips open-ended token generation entirely.
• Single-Pass Parallel Evaluation: Evaluates multiple typed questions (Noul, Choice, Score) simultaneously in a single forward pass over unstructured state (<10ms on MPS, <50ms on edge GPU).
• Zero Structural Hallucinations: Constrained by design to valid schema outputs.
• Calibrated Probabilities (RLCD): Trained and scored using Strictly Proper Scoring Rules (normalized Brier score + Negative Log-Likelihood) to penalize overconfidence.
• Dual API Endpoints: Exposes POST /v1/decide and POST /v1/classone for schema-out decision queries.
• Python Client SDK: Idiomatic synchronous and asynchronous client library (from classone import ClassOneClient).

Quickstart

1. Installation

git clone https://github.com/devops-thiago/class-one.git
cd class-one
python3 -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

2. Run Test Suite

pytest tests/

3. Start the API Server

uvicorn classone.server.app:app --host 0.0.0.0 --port 8000

4. Query via HTTP

curl -X POST http://localhost:8000/v1/classone \
  -H "Content-Type: application/json" \
  -d '{
    "model": "class-one-gemma-4-e2b",
    "state": {
      "message": "I was double billed for my subscription."
    },
    "questions": {
      "refund": {
        "type": "noul",
        "instructions": "Is customer asking for a refund?"
      },
      "routing": {
        "type": "choice",
        "instructions": "Which department?",
        "criteria": {
          "billing": "Charges and payouts",
          "technical": "App bugs and crashes"
        }
      },
      "urgency": {
        "type": "score",
        "instructions": "Ticket urgency",
        "criteria": ["low", "normal", "critical"]
      }
    }
  }'

5. Python SDK Usage

from classone import ClassOneClient, Choice, Noul, Score

with ClassOneClient(base_url="http://localhost:8000") as client:
    response = client.decide(
        state="Customer account was locked after three incorrect attempts.",
        questions={
            "urgent": Noul(instructions="Is this an urgent security event?"),
            "action": Choice(
                instructions="Action required:",
                criteria={"unlock": "Send unlock link", "escalate": "Escalate to SecOps"}
            ),
            "severity": Score(
                instructions="Assess severity:",
                criteria=["low", "medium", "critical"]
            ),
        }
    )

    print("Urgent P(true):", response.nouls["urgent"].noul)
    print("Action Choice:", response.choices["action"].choice)
    print("Severity Score:", response.scores["severity"].score)

CLI Utilities

# Run single-pass decision inference on Gemma:
python scripts/run_classone.py --model google/gemma-2-2b-it --device auto

# Benchmark latency vs traditional autoregressive LLMs:
python scripts/benchmark_latency.py --iterations 30

# Publish a model checkpoint to the Hugging Face Hub:
python scripts/push_to_hub.py --checkpoint-dir ./checkpoints/classone_v1 --repo-id your-org/classone-gemma-4-e2b

Reference

• Architecture: ARCHITECTURE.md

Legal & Attribution Notices

• Gemma is a trademark of Google LLC.
• Gemma is provided under and subject to the Gemma Terms of Use found at ai.google.dev/gemma/terms.