M.I.M.I.R - Multi-agent Intelligent Memory & Insight Repository

//go:build integration // +build integration package embed import ( "context" "os" "testing" "time" ) // Run with: go test -tags=integration -v ./pkg/embed/... // Requires llama.cpp server running on localhost:11434 func TestLlamaCppEmbeddings(t *testing.T) { // Skip if not running integration tests if os.Getenv("INTEGRATION_TEST") == "" { t.Skip("Set INTEGRATION_TEST=1 to run") } // Configure for llama.cpp (uses OpenAI-compatible format) config := &Config{ Provider: "openai", // llama.cpp uses OpenAI format APIURL: "http://localhost:11434", APIPath: "/v1/embeddings", Model: "bge-m3", Dimensions: 1024, Timeout: 30 * time.Second, } embedder := NewOpenAI(config) ctx := context.Background() // Test single embedding t.Run("SingleEmbed", func(t *testing.T) { vec, err := embedder.Embed(ctx, "hello world") if err != nil { t.Fatalf("Embed failed: %v", err) } if len(vec) != 1024 { t.Errorf("Expected 1024 dimensions, got %d", len(vec)) } t.Logf("✓ Single embedding: %d dimensions", len(vec)) }) // Test batch embedding t.Run("BatchEmbed", func(t *testing.T) { texts := []string{ "graph database stores relationships", "vector search finds similar content", "machine learning algorithms", } vecs, err := embedder.EmbedBatch(ctx, texts) if err != nil { t.Fatalf("EmbedBatch failed: %v", err) } if len(vecs) != 3 { t.Errorf("Expected 3 embeddings, got %d", len(vecs)) } for i, vec := range vecs { if len(vec) != 1024 { t.Errorf("Embedding %d: expected 1024 dims, got %d", i, len(vec)) } } t.Logf("✓ Batch embedding: %d texts", len(vecs)) }) // Test similarity (sanity check) t.Run("Similarity", func(t *testing.T) { vec1, _ := embedder.Embed(ctx, "cat") vec2, _ := embedder.Embed(ctx, "kitten") vec3, _ := embedder.Embed(ctx, "automobile") sim12 := vector.CosineSimilarity(vec1, vec2) sim13 := vector.CosineSimilarity(vec1, vec3) t.Logf("cat vs kitten: %.4f", sim12) t.Logf("cat vs automobile: %.4f", sim13) if sim12 <= sim13 { t.Errorf("Expected cat-kitten (%.4f) > cat-automobile (%.4f)", sim12, sim13) } t.Logf("✓ Similarity ordering correct") }) }