feat(sdk/go): add support for image inputs in ai calls

sdk/go/agent/agent_test.go

@@ -512,7 +512,9 @@ func TestAI(t *testing.T) {
 			Choices: []ai.Choice{
 				{
 					Message: ai.Message{
-						Content: "AI response",
+						Content: []ai.ContentPart{
+							{Type: "text", Text: "AI response"},
+						},
 					},
 				},
 			},

sdk/go/ai/README.md

@@ -146,6 +146,32 @@ Functional options for customizing AI requests:
 - `ai.WithStream()` - Enable streaming
 - `ai.WithJSONMode()` - Enable JSON object mode
 - `ai.WithSchema(schema interface{})` - Enable structured outputs with schema
+##### Multimodal
+- `ai.WithImageFile(path string)` - Attach an image from a local file
+- `ai.WithImageURL(url string)` - Attach an image from a remote URL
+- `ai.WithImageBytes(data []byte, mimeType string)` - Add an image from raw bytes (SDK encodes automatically)
+
+### Multimodal Inputs (Images)
+
+You can attach images files to AI requests.
+
+```go
+// Image from file
+response, _ := agent.AI(ctx, "Describe this image",
+    ai.WithImageFile("./photo.jpg"),
+)
+
+// Image from URL
+response, _ = agent.AI(ctx, "Describe this image",
+    ai.WithImageURL("https://example.com/image.jpg"),
+)
+
+// Image from bytes
+data, _ := os.ReadFile("image.png")
+response, _ = agent.AI(ctx, "What's in this image?",
+    ai.WithImageBytes(data, "image/png"),
+)
+```
 
 ### Response Methods
 

sdk/go/ai/client.go

@@ -39,7 +39,12 @@ func (c *Client) Complete(ctx context.Context, prompt string, opts ...Option) (*
 	// Build base request
 	req := &Request{
 		Messages: []Message{
-			{Role: "user", Content: prompt},
+			{
+				Role: "user",
+				Content: []ContentPart{
+					{Type: "text", Text: prompt},
+				},
+			},
 		},
 		Model:       c.config.Model,
 		Temperature: &c.config.Temperature,
@@ -78,7 +83,17 @@ func (c *Client) CompleteWithMessages(ctx context.Context, messages []Message, o
 
 func (c *Client) doRequest(ctx context.Context, req *Request) (*Response, error) {
 	// Marshal request
-	body, err := json.Marshal(req)
+
+	var body []byte
+	var err error
+
+	if c.config.IsOpenRouter() {
+		payload := transformForOpenRouter(req)
+		body, err = json.Marshal(payload)
+		fmt.Printf("[DEBUG] OpenRouter JSON:\n%s\n", string(body))
+	} else {
+		body, err = json.Marshal(req)
+	}
 	if err != nil {
 		return nil, fmt.Errorf("marshal request: %w", err)
 	}
@@ -155,7 +170,12 @@ func (c *Client) StreamComplete(ctx context.Context, prompt string, opts ...Opti
 		opts = append(opts, WithStream())
 		req := &Request{
 			Messages: []Message{
-				{Role: "user", Content: prompt},
+				{
+					Role: "user",
+					Content: []ContentPart{
+						{Type: "text", Text: prompt},
+					},
+				},
 			},
 			Model:       c.config.Model,
 			Temperature: &c.config.Temperature,

sdk/go/ai/client_test.go

@@ -73,7 +73,9 @@ func TestComplete(t *testing.T) {
 		assert.NoError(t, err)
 		assert.Len(t, req.Messages, 1)
 		assert.Equal(t, "user", req.Messages[0].Role)
-		assert.Equal(t, "Hello", req.Messages[0].Content)
+		assert.Len(t, req.Messages[0].Content, 1)
+		assert.Equal(t, "text", req.Messages[0].Content[0].Type)
+		assert.Equal(t, "Hello", req.Messages[0].Content[0].Text)
 
 		// Send response
 		resp := Response{
@@ -85,8 +87,13 @@ func TestComplete(t *testing.T) {
 				{
 					Index: 0,
 					Message: Message{
-						Role:    "assistant",
-						Content: "Hello! How can I help you?",
+						Role: "assistant",
+						Content: []ContentPart{
+							{
+								Type: "text",
+								Text: "Hello! How can I help you?",
+							},
+						},
 					},
 					FinishReason: "stop",
 				},
@@ -97,6 +104,7 @@ func TestComplete(t *testing.T) {
 				TotalTokens:      15,
 			},
 		}
+
 		w.WriteHeader(http.StatusOK)
 		json.NewEncoder(w).Encode(resp)
 	}))
@@ -124,7 +132,17 @@ func TestComplete_WithAPIKeyOverride(t *testing.T) {
 
 		resp := Response{
 			Choices: []Choice{
-				{Message: Message{Content: "ok"}},
+				{
+					Message: Message{
+						Role: "assistant",
+						Content: []ContentPart{
+							{
+								Type: "text",
+								Text: "ok",
+							},
+						},
+					},
+				},
 			},
 		}
 		w.WriteHeader(http.StatusOK)
@@ -162,9 +180,20 @@ func TestComplete_WithOptions(t *testing.T) {
 
 		resp := Response{
 			Choices: []Choice{
-				{Message: Message{Content: "Response"}},
+				{
+					Message: Message{
+						Role: "assistant",
+						Content: []ContentPart{
+							{
+								Type: "text",
+								Text: "Response",
+							},
+						},
+					},
+				},
 			},
 		}
+
 		w.WriteHeader(http.StatusOK)
 		json.NewEncoder(w).Encode(resp)
 	}))
@@ -197,9 +226,20 @@ func TestComplete_WithOpenRouterHeaders(t *testing.T) {
 		receivedHeaders = r.Header
 		resp := Response{
 			Choices: []Choice{
-				{Message: Message{Content: "Response"}},
+				{
+					Message: Message{
+						Role: "assistant",
+						Content: []ContentPart{
+							{
+								Type: "text",
+								Text: "Reponse",
+							},
+						},
+					},
+				},
 			},
 		}
+
 		w.WriteHeader(http.StatusOK)
 		json.NewEncoder(w).Encode(resp)
 	}))
@@ -325,9 +365,20 @@ func TestCompleteWithMessages(t *testing.T) {
 
 		resp := Response{
 			Choices: []Choice{
-				{Message: Message{Content: "Response"}},
+				{
+					Message: Message{
+						Role: "assistant", // optional but recommended
+						Content: []ContentPart{
+							{
+								Type: "text",
+								Text: "Response",
+							},
+						},
+					},
+				},
 			},
 		}
+
 		w.WriteHeader(http.StatusOK)
 		json.NewEncoder(w).Encode(resp)
 	}))
@@ -343,8 +394,18 @@ func TestCompleteWithMessages(t *testing.T) {
 	require.NoError(t, err)
 
 	messages := []Message{
-		{Role: "system", Content: "You are helpful"},
-		{Role: "user", Content: "Hello"},
+		{
+			Role: "system",
+			Content: []ContentPart{
+				{Type: "text", Text: "You are helpful"},
+			},
+		},
+		{
+			Role: "user",
+			Content: []ContentPart{
+				{Type: "text", Text: "Hello"},
+			},
+		},
 	}
 
 	resp, err := client.CompleteWithMessages(context.Background(), messages)

sdk/go/ai/config.go

@@ -89,3 +89,58 @@ func (c *Config) IsOpenRouter() bool {
 	return c.BaseURL == "https://openrouter.ai/api/v1" ||
 		c.BaseURL == "https://openrouter.ai/api/v1/"
 }
+
+type OpenRouterRequest struct {
+	Messages []OpenRouterMessage `json:"messages"`
+	Model    string              `json:"model,omitempty"`
+}
+
+type OpenRouterMessage struct {
+	Role    string                  `json:"role"`
+	Content []OpenRouterContentPart `json:"content"`
+}
+
+type OpenRouterContentPart struct {
+	Type     string     `json:"type"`
+	Text     string     `json:"text,omitempty"`
+	ImageURL *ImageData `json:"image_url,omitempty"`
+}
+
+type ImageData struct {
+	URL    string `json:"url"`
+	Detail string `json:"detail,omitempty"`
+}
+
+func transformForOpenRouter(req *Request) *OpenRouterRequest {
+	var messages []OpenRouterMessage
+
+	for _, m := range req.Messages {
+		msg := OpenRouterMessage{
+			Role: m.Role,
+		}
+
+		for _, c := range m.Content {
+			part := OpenRouterContentPart{
+				Type: c.Type,
+				Text: c.Text,
+			}
+
+			// Map Go struct's input_image to OpenRouter's image_url
+			if c.Type == "input_image" && c.ImageURL != "" {
+				part.Type = "image_url"
+				part.ImageURL = &ImageData{
+					URL: c.ImageURL,
+				}
+			}
+
+			msg.Content = append(msg.Content, part)
+		}
+
+		messages = append(messages, msg)
+	}
+
+	return &OpenRouterRequest{
+		Messages: messages,
+		Model:    req.Model,
+	}
+}

sdk/go/ai/multimodal.go

@@ -0,0 +1,18 @@
+package ai
+
+import "strings"
+
+func detectMIMEType(path string) string {
+	switch {
+	case strings.HasSuffix(path, ".png"):
+		return "image/png"
+	case strings.HasSuffix(path, ".jpg"), strings.HasSuffix(path, ".jpeg"):
+		return "image/jpeg"
+	case strings.HasSuffix(path, ".gif"):
+		return "image/gif"
+	case strings.HasSuffix(path, ".webp"):
+		return "image/webp"
+	default:
+		return "application/octet-stream"
+	}
+}