Changeset 0.12.0 (#63)

server/providers/anthropic.go

@@ -263,6 +263,40 @@ func (p *AnthropicProvider) doRequest(ctx context.Context, cfg ProviderConfig, r
 					Content:   m.Content,
 				},
 			}
+		} else if len(m.ContentParts) > 0 && m.Role == "user" {
+			// Multimodal user message: convert ContentParts to Anthropic blocks
+			for _, p := range m.ContentParts {
+				switch p.Type {
+				case "image_url":
+					if p.ImageURL != nil {
+						// Parse data URI: "data:image/jpeg;base64,{data}"
+						mediaType, b64Data := parseDataURI(p.ImageURL.URL)
+						if b64Data != "" {
+							antMsg.Content = append(antMsg.Content, anthropicContentBlock{
+								Type: "image",
+								Source: &anthropicImageSource{
+									Type:      "base64",
+									MediaType: mediaType,
+									Data:      b64Data,
+								},
+							})
+						}
+					}
+				default: // "text", "document"
+					if p.Text != "" {
+						antMsg.Content = append(antMsg.Content, anthropicContentBlock{
+							Type: "text",
+							Text: p.Text,
+						})
+					}
+				}
+			}
+			if len(antMsg.Content) == 0 {
+				// Fallback: shouldn't happen, but safety net
+				antMsg.Content = []anthropicContentBlock{
+					{Type: "text", Text: m.Content},
+				}
+			}
 		} else {
 			// Regular text message
 			antMsg.Content = []anthropicContentBlock{
@@ -354,6 +388,8 @@ type anthropicContentBlock struct {
 	Type string `json:"type"`
 	// text
 	Text string `json:"text,omitempty"`
+	// image (type="image")
+	Source *anthropicImageSource `json:"source,omitempty"`
 	// tool_use
 	ID    string          `json:"id,omitempty"`
 	Name  string          `json:"name,omitempty"`
@@ -363,6 +399,15 @@ type anthropicContentBlock struct {
 	Content   string `json:"content,omitempty"`
 }
 
+// anthropicImageSource is the Anthropic-native image format.
+// Anthropic uses {"type":"base64","media_type":"image/jpeg","data":"..."}
+// rather than OpenAI's data URI approach.
+type anthropicImageSource struct {
+	Type      string `json:"type"`       // "base64"
+	MediaType string `json:"media_type"` // "image/jpeg", "image/png", etc.
+	Data      string `json:"data"`       // raw base64 (no data: prefix)
+}
+
 type anthropicMessage struct {
 	Role    string                  `json:"role"`
 	Content []anthropicContentBlock `json:"content"`
@@ -425,3 +470,33 @@ type anthropicStreamEvent struct {
 		Name string `json:"name"`
 	} `json:"content_block,omitempty"`
 }
+
+// parseDataURI splits a data URI into media type and base64 data.
+// Input:  "data:image/jpeg;base64,/9j/4AAQ..."
+// Output: "image/jpeg", "/9j/4AAQ..."
+// Returns empty strings if the URI is malformed.
+func parseDataURI(uri string) (mediaType, data string) {
+	// Strip "data:" prefix
+	if !strings.HasPrefix(uri, "data:") {
+		return "", ""
+	}
+	rest := uri[5:]
+
+	// Split on comma (separates header from data)
+	commaIdx := strings.Index(rest, ",")
+	if commaIdx < 0 {
+		return "", ""
+	}
+
+	header := rest[:commaIdx]
+	data = rest[commaIdx+1:]
+
+	// Extract media type from header (before ";base64")
+	if semiIdx := strings.Index(header, ";"); semiIdx >= 0 {
+		mediaType = header[:semiIdx]
+	} else {
+		mediaType = header
+	}
+
+	return mediaType, data
+}