MCP Chat Analysis Server

import pytest import numpy as np from mcp_chat_analysis.embeddings import EmbeddingGenerator @pytest.fixture def embedding_generator(): return EmbeddingGenerator( model_name="sentence-transformers/all-MiniLM-L6-v2", device="cpu", batch_size=2 ) @pytest.mark.asyncio async def test_single_embedding_generation(embedding_generator): text = "This is a test message" embedding = await embedding_generator.generate_single(text) assert isinstance(embedding, list) assert len(embedding) == embedding_generator.embedding_dim assert all(isinstance(x, float) for x in embedding) @pytest.mark.asyncio async def test_batch_embedding_generation(embedding_generator): texts = [ "First test message", "Second test message", "Third test message" ] embeddings = await embedding_generator.generate(texts) assert isinstance(embeddings, list) assert len(embeddings) == len(texts) assert all(len(emb) == embedding_generator.embedding_dim for emb in embeddings) @pytest.mark.asyncio async def test_similarity_computation(embedding_generator): text1 = "This is about artificial intelligence" text2 = "AI and machine learning are related topics" text3 = "The weather is nice today" # Similar texts should have higher similarity sim1 = await embedding_generator.compute_similarity(text1, text2) sim2 = await embedding_generator.compute_similarity(text1, text3) assert 0 <= sim1 <= 1 assert 0 <= sim2 <= 1 assert sim1 > sim2 # Related topics should be more similar @pytest.mark.asyncio async def test_similar_chunks_finding(embedding_generator): query = "artificial intelligence" texts = [ "AI and machine learning", "Natural language processing", "The weather forecast", "Today's temperature", "Deep learning algorithms" ] results = await embedding_generator.find_similar_chunks( query, texts, threshold=0.3, top_k=3 ) assert isinstance(results, list) assert len(results) <= 3 assert all("score" in r and "text" in r for r in results) assert all(0 <= r["score"] <= 1 for r in results) # Check ordering scores = [r["score"] for r in results] assert scores == sorted(scores, reverse=True) @pytest.mark.asyncio async def test_long_text_handling(embedding_generator): long_text = "a" * 1000 # Text longer than max_length embedding = await embedding_generator.generate_single(long_text) assert isinstance(embedding, list) assert len(embedding) == embedding_generator.embedding_dim @pytest.mark.asyncio async def test_empty_text_handling(embedding_generator): empty_text = "" embedding = await embedding_generator.generate_single(empty_text) assert isinstance(embedding, list) assert len(embedding) == embedding_generator.embedding_dim @pytest.mark.asyncio async def test_different_similarity_methods(embedding_generator): text1 = "This is a test" text2 = "Another test message" cosine_sim = await embedding_generator.compute_similarity( text1, text2, method="cosine" ) euclidean_sim = await embedding_generator.compute_similarity( text1, text2, method="euclidean" ) assert 0 <= cosine_sim <= 1 assert 0 <= euclidean_sim <= 1 @pytest.mark.asyncio async def test_invalid_similarity_method(embedding_generator): with pytest.raises(ValueError): await embedding_generator.compute_similarity( "text1", "text2", method="invalid" )