ollama_ollama/server/images_test.go

package server

import (
	"strings"
	"testing"

	"github.com/ollama/ollama/fs/ggml"
	"github.com/ollama/ollama/template"
	"github.com/ollama/ollama/types/model"
)

func TestModelCapabilities(t *testing.T) {
	// Create completion model (llama architecture without vision)
	completionModelPath, _ := createBinFile(t, ggml.KV{
		"general.architecture": "llama",
	}, []*ggml.Tensor{})

	// Create vision model (llama architecture with vision block count)
	visionModelPath, _ := createBinFile(t, ggml.KV{
		"general.architecture":     "llama",
		"llama.vision.block_count": uint32(1),
	}, []*ggml.Tensor{})

	// Create embedding model (bert architecture with pooling type)
	embeddingModelPath, _ := createBinFile(t, ggml.KV{
		"general.architecture": "bert",
		"bert.pooling_type":    uint32(1),
	}, []*ggml.Tensor{})

	toolsInsertTemplate, err := template.Parse("{{ .prompt }}{{ if .tools }}{{ .tools }}{{ end }}{{ if .suffix }}{{ .suffix }}{{ end }}")
	if err != nil {
		t.Fatalf("Failed to parse template: %v", err)
	}

	chatTemplate, err := template.Parse("{{ .prompt }}")
	if err != nil {
		t.Fatalf("Failed to parse template: %v", err)
	}

	toolsTemplate, err := template.Parse("{{ .prompt }}{{ if .tools }}{{ .tools }}{{ end }}")
	if err != nil {
		t.Fatalf("Failed to parse template: %v", err)
	}

	testModels := []struct {
		name         string
		model        Model
		expectedCaps []model.Capability
	}{
		{
			name: "model with image generation capability via config",
			model: Model{
				Config: model.ConfigV2{
					Capabilities: []string{"image"},
				},
			},
			expectedCaps: []model.Capability{model.CapabilityImageGeneration},
		},
		{
			name: "model with completion capability",
			model: Model{
				ModelPath: completionModelPath,
				Template:  chatTemplate,
			},
			expectedCaps: []model.Capability{model.CapabilityCompletion},
		},

		{
			name: "model with completion, tools, and insert capability",
			model: Model{
				ModelPath: completionModelPath,
				Template:  toolsInsertTemplate,
			},
			expectedCaps: []model.Capability{model.CapabilityCompletion, model.CapabilityTools, model.CapabilityInsert},
		},
		{
			name: "model with tools capability",
			model: Model{
				ModelPath: completionModelPath,
				Template:  toolsTemplate,
			},
			expectedCaps: []model.Capability{model.CapabilityCompletion, model.CapabilityTools},
		},
		{
			name: "model with vision capability",
			model: Model{
				ModelPath: visionModelPath,
				Template:  chatTemplate,
			},
			expectedCaps: []model.Capability{model.CapabilityCompletion, model.CapabilityVision},
		},
		{
			name: "model with vision, tools, and insert capability",
			model: Model{
				ModelPath: visionModelPath,
				Template:  toolsInsertTemplate,
			},
			expectedCaps: []model.Capability{model.CapabilityCompletion, model.CapabilityVision, model.CapabilityTools, model.CapabilityInsert},
		},
		{
			name: "model with embedding capability",
			model: Model{
				ModelPath: embeddingModelPath,
				Template:  chatTemplate,
			},
			expectedCaps: []model.Capability{model.CapabilityEmbedding},
		},
	}

	// compare two slices of model.Capability regardless of order
	compareCapabilities := func(a, b []model.Capability) bool {
		if len(a) != len(b) {
			return false
		}

		aCount := make(map[model.Capability]int)
		for _, cap := range a {
			aCount[cap]++
		}

		bCount := make(map[model.Capability]int)
		for _, cap := range b {
			bCount[cap]++
		}

		for cap, count := range aCount {
			if bCount[cap] != count {
				return false
			}
		}

		return true
	}

	for _, tt := range testModels {
		t.Run(tt.name, func(t *testing.T) {
			// Test Capabilities method
			caps := tt.model.Capabilities()
			if !compareCapabilities(caps, tt.expectedCaps) {
				t.Errorf("Expected capabilities %v, got %v", tt.expectedCaps, caps)
			}
		})
	}
}

func TestModelCheckCapabilities(t *testing.T) {
	// Create simple model file for tests that don't depend on GGUF content
	completionModelPath, _ := createBinFile(t, ggml.KV{
		"general.architecture": "llama",
	}, []*ggml.Tensor{})

	// Create vision model (llama architecture with vision block count)
	visionModelPath, _ := createBinFile(t, ggml.KV{
		"general.architecture":     "llama",
		"llama.vision.block_count": uint32(1),
	}, []*ggml.Tensor{})

	// Create embedding model (bert architecture with pooling type)
	embeddingModelPath, _ := createBinFile(t, ggml.KV{
		"general.architecture": "bert",
		"bert.pooling_type":    uint32(1),
	}, []*ggml.Tensor{})

	toolsInsertTemplate, err := template.Parse("{{ .prompt }}{{ if .tools }}{{ .tools }}{{ end }}{{ if .suffix }}{{ .suffix }}{{ end }}")
	if err != nil {
		t.Fatalf("Failed to parse template: %v", err)
	}

	chatTemplate, err := template.Parse("{{ .prompt }}")
	if err != nil {
		t.Fatalf("Failed to parse template: %v", err)
	}

	toolsTemplate, err := template.Parse("{{ .prompt }}{{ if .tools }}{{ .tools }}{{ end }}")
	if err != nil {
		t.Fatalf("Failed to parse template: %v", err)
	}

	tests := []struct {
		name           string
		model          Model
		checkCaps      []model.Capability
		expectedErrMsg string
	}{
		{
			name: "completion model without tools capability",
			model: Model{
				ModelPath: completionModelPath,
				Template:  chatTemplate,
			},
			checkCaps:      []model.Capability{model.CapabilityTools},
			expectedErrMsg: "does not support tools",
		},
		{
			name: "model with all needed capabilities",
			model: Model{
				ModelPath: completionModelPath,
				Template:  toolsInsertTemplate,
			},
			checkCaps: []model.Capability{model.CapabilityTools, model.CapabilityInsert},
		},
		{
			name: "model missing insert capability",
			model: Model{
				ModelPath: completionModelPath,
				Template:  toolsTemplate,
			},
			checkCaps:      []model.Capability{model.CapabilityInsert},
			expectedErrMsg: "does not support insert",
		},
		{
			name: "model missing vision capability",
			model: Model{
				ModelPath: completionModelPath,
				Template:  toolsTemplate,
			},
			checkCaps:      []model.Capability{model.CapabilityVision},
			expectedErrMsg: "does not support vision",
		},
		{
			name: "model with vision capability",
			model: Model{
				ModelPath: visionModelPath,
				Template:  chatTemplate,
			},
			checkCaps: []model.Capability{model.CapabilityVision},
		},
		{
			name: "model with embedding capability",
			model: Model{
				ModelPath: embeddingModelPath,
				Template:  chatTemplate,
			},
			checkCaps: []model.Capability{model.CapabilityEmbedding},
		},
		{
			name: "unknown capability",
			model: Model{
				ModelPath: completionModelPath,
				Template:  chatTemplate,
			},
			checkCaps:      []model.Capability{"unknown"},
			expectedErrMsg: "unknown capability",
		},
	}

	for _, tt := range tests {
		t.Run(tt.name, func(t *testing.T) {
			// Test CheckCapabilities method
			err := tt.model.CheckCapabilities(tt.checkCaps...)
			if tt.expectedErrMsg == "" {
				if err != nil {
					t.Errorf("Expected no error, got: %v", err)
				}
			} else {
				if err == nil {
					t.Errorf("Expected error containing %q, got nil", tt.expectedErrMsg)
				} else if !strings.Contains(err.Error(), tt.expectedErrMsg) {
					t.Errorf("Expected error containing %q, got: %v", tt.expectedErrMsg, err)
				}
			}
		})
	}
}