🧠 Plato Neural

PLATO neural inference engine — tile scoring, Q&A, and knowledge gap detection

Fine-tuned Qwen2.5-0.5B model for PLATO operations. Ask questions, score tile quality via perplexity, generate new tiles, and detect knowledge gaps.

Install

pip install plato-neural

Quick Start

Ask Questions

from plato_neural import PlatoBrain

brain = PlatoBrain(model_path="path/to/fine-tuned-model")
result = brain.ask("What is the flywheel?", temperature=0.7)
print(f"A: {result['answer']}")
print(f"Confidence: {result['confidence']}")
print(f"Latency: {result['latency_ms']}ms")

Score Tile Quality

result = brain.score("What is X?", "X is Y", content="additional context")
print(f"Perplexity: {result['perplexity']}")
print(f"Quality: {result['quality']}")  # excellent/good/fair/weak/poor

Generate Tiles

tile = brain.generate_tile(domain="architecture", topic="flywheel")
print(f"Q: {tile['question']}")
print(f"A: {tile['answer']}")

Batch Scoring

from plato_neural import PlatoScorer

scorer = PlatoScorer(model_path="path/to/model", plato_api="http://localhost:8847")
tiles = [{"question": "Q1", "answer": "A1"}, {"question": "Q2", "answer": "A2"}]
results = scorer.score_tiles(tiles)
gaps = scorer.find_gaps(tiles, threshold_ppl=100)

Part of Cocapn · Agent Infrastructure