plato-semantic-sim 0.1.1

PLATO semantic similarity — embedding-based tile comparison and deduplication

🔬 Plato Semantic Sim

Embedding-based tile comparison and deduplication for PLATO rooms

Computes similarity between knowledge tiles using cosine similarity, Jaccard index, and edit distance. Finds duplicates via content hashing (exact) and vector comparison (semantic).

Install

pip install plato-semantic-sim

Quick Start

from plato_semantic_sim import SimilarityEngine, DedupEngine

engine = SimilarityEngine()
scores = engine.find_similar([0.1, 0.2, 0.3], {"a": [0.1, 0.2, 0.3], "b": [0.9, 0.8, 0.7]})
print(scores)  # [("a", 1.0)]

dedup = DedupEngine(similarity_threshold=0.85)
result = dedup.add("tile-1", "knowledge content", [0.1, 0.2, 0.3])
print(result)  # {"status": "unique"}

API

Class	Purpose
SimilarityEngine	Compare vectors, find similar, pairwise matrix
CosineSimilarity	Dot product / magnitude
JaccardSimilarity	Token overlap ratio
DedupEngine	Exact + semantic dedup in one pass

Part of Cocapn · Agent Infrastructure