core/server/providers/multimodal_test.go

package providers

import (
	"encoding/json"
	"testing"
)

// ── ContentPart Tests ──────────────────────

func TestContentPart_TextOnly(t *testing.T) {
	msg := Message{
		Role:    "user",
		Content: "hello",
		ContentParts: []ContentPart{
			{Type: "text", Text: "hello"},
		},
	}
	if len(msg.ContentParts) != 1 {
		t.Fatalf("expected 1 part, got %d", len(msg.ContentParts))
	}
	if msg.ContentParts[0].Type != "text" {
		t.Errorf("type = %q, want text", msg.ContentParts[0].Type)
	}
}

func TestContentPart_WithImage(t *testing.T) {
	msg := Message{
		Role:    "user",
		Content: "describe this",
		ContentParts: []ContentPart{
			{Type: "text", Text: "describe this"},
			{Type: "image_url", ImageURL: &ImageURL{URL: "data:image/png;base64,abc123", Detail: "auto"}},
		},
	}
	if len(msg.ContentParts) != 2 {
		t.Fatalf("expected 2 parts, got %d", len(msg.ContentParts))
	}
	if msg.ContentParts[1].ImageURL == nil {
		t.Fatal("expected non-nil ImageURL")
	}
	if msg.ContentParts[1].ImageURL.URL != "data:image/png;base64,abc123" {
		t.Errorf("URL = %q", msg.ContentParts[1].ImageURL.URL)
	}
}

// ── OpenAI Multimodal Serialization ────────

func TestOpenAI_TextOnlyContent_String(t *testing.T) {
	// When Content is a string, it should serialize as a JSON string
	msg := openaiMessage{
		Role:    "user",
		Content: "hello world",
	}
	data, err := json.Marshal(msg)
	if err != nil {
		t.Fatalf("Marshal: %v", err)
	}

	var raw map[string]interface{}
	json.Unmarshal(data, &raw)

	content := raw["content"]
	if s, ok := content.(string); !ok || s != "hello world" {
		t.Errorf("content = %v (%T), want string 'hello world'", content, content)
	}
}

func TestOpenAI_MultimodalContent_Array(t *testing.T) {
	// When Content is an array, it should serialize as a JSON array
	msg := openaiMessage{
		Role: "user",
		Content: []openaiContentPart{
			{Type: "text", Text: "describe this"},
			{Type: "image_url", ImageURL: &openaiImageURL{URL: "data:image/png;base64,abc", Detail: "auto"}},
		},
	}
	data, err := json.Marshal(msg)
	if err != nil {
		t.Fatalf("Marshal: %v", err)
	}

	var raw map[string]interface{}
	json.Unmarshal(data, &raw)

	content := raw["content"]
	arr, ok := content.([]interface{})
	if !ok {
		t.Fatalf("content is %T, want array", content)
	}
	if len(arr) != 2 {
		t.Fatalf("content has %d elements, want 2", len(arr))
	}

	// Verify first part is text
	part0 := arr[0].(map[string]interface{})
	if part0["type"] != "text" {
		t.Errorf("part[0].type = %v, want text", part0["type"])
	}

	// Verify second part is image_url
	part1 := arr[1].(map[string]interface{})
	if part1["type"] != "image_url" {
		t.Errorf("part[1].type = %v, want image_url", part1["type"])
	}
}

// ── Anthropic Image Source ─────────────────

func TestAnthropic_ImageSourceBlock(t *testing.T) {
	block := anthropicContentBlock{
		Type: "image",
		Source: &anthropicImageSource{
			Type:      "base64",
			MediaType: "image/jpeg",
			Data:      "abc123",
		},
	}
	data, err := json.Marshal(block)
	if err != nil {
		t.Fatalf("Marshal: %v", err)
	}

	var raw map[string]interface{}
	json.Unmarshal(data, &raw)

	if raw["type"] != "image" {
		t.Errorf("type = %v", raw["type"])
	}
	source := raw["source"].(map[string]interface{})
	if source["type"] != "base64" {
		t.Errorf("source.type = %v", source["type"])
	}
	if source["media_type"] != "image/jpeg" {
		t.Errorf("source.media_type = %v", source["media_type"])
	}
	if source["data"] != "abc123" {
		t.Errorf("source.data = %v", source["data"])
	}
}

func TestAnthropic_TextBlockUnchanged(t *testing.T) {
	block := anthropicContentBlock{
		Type: "text",
		Text: "hello",
	}
	data, err := json.Marshal(block)
	if err != nil {
		t.Fatalf("Marshal: %v", err)
	}

	var raw map[string]interface{}
	json.Unmarshal(data, &raw)

	if raw["type"] != "text" {
		t.Errorf("type = %v", raw["type"])
	}
	if raw["text"] != "hello" {
		t.Errorf("text = %v", raw["text"])
	}
	// Should NOT have source field
	if _, ok := raw["source"]; ok {
		t.Error("text block should not have source field")
	}
}

// ── parseDataURI ───────────────────────────

func TestParseDataURI_Valid(t *testing.T) {
	tests := []struct {
		uri       string
		wantType  string
		wantData  string
	}{
		{"data:image/jpeg;base64,/9j/4AAQ", "image/jpeg", "/9j/4AAQ"},
		{"data:image/png;base64,abc123", "image/png", "abc123"},
		{"data:image/webp;base64,RIFF", "image/webp", "RIFF"},
	}
	for _, tc := range tests {
		mediaType, data := parseDataURI(tc.uri)
		if mediaType != tc.wantType {
			t.Errorf("parseDataURI(%q) mediaType = %q, want %q", tc.uri, mediaType, tc.wantType)
		}
		if data != tc.wantData {
			t.Errorf("parseDataURI(%q) data = %q, want %q", tc.uri, data, tc.wantData)
		}
	}
}

func TestParseDataURI_Invalid(t *testing.T) {
	tests := []string{
		"",
		"not-a-data-uri",
		"data:nocomma",
		"https://example.com/image.png",
	}
	for _, uri := range tests {
		mediaType, data := parseDataURI(uri)
		if mediaType != "" || data != "" {
			t.Errorf("parseDataURI(%q) = (%q, %q), want empty", uri, mediaType, data)
		}
	}
}