wave-agent-sdk 0.18.7 → 0.19.0

package/src/utils/convertMessagesForAPI.ts

@@ -39,15 +39,26 @@ function safeToolArguments(args: string): string {
   }
 }
 
+/**
+ * Options for converting messages to API format.
+ */
+export interface ConvertMessagesOptions {
+  /** When false, images are stripped and replaced with text placeholders. */
+  supportsVision?: boolean;
+}
+
 /**
  * Convert message format to API call format, stopping when a compacted message is encountered.
  * Messages with no meaningful content or tool calls are filtered out.
  * @param messages Message list
+ * @param options Optional conversion options (e.g. supportsVision)
  * @returns Converted API message format list
  */
 export function convertMessagesForAPI(
   messages: Message[],
+  options?: ConvertMessagesOptions,
 ): ChatCompletionMessageParam[] {
+  const supportsVision = options?.supportsVision !== false;
   const recentMessages: ChatCompletionMessageParam[] = [];
 
   const startIndex = messages.length - 1;
@@ -86,53 +97,64 @@ export function convertMessagesForAPI(
       const completedToolIds = new Set<string>(); // Record completed tool IDs
 
       if (toolBlocks.length > 0) {
+        // Collect image user messages to place after all tool messages
+        const imageUserMessages: ChatCompletionMessageParam[] = [];
+
         toolBlocks.forEach((toolBlock) => {
           // Only add completed tool blocks (i.e., stage is 'end')
           if (toolBlock.id && toolBlock.stage === "end") {
             completedToolIds.add(toolBlock.id);
 
-            // Check for image data
+            // Always add a proper tool message to satisfy tool_call_id pairing
+            recentMessages.unshift({
+              tool_call_id: toolBlock.id,
+              role: "tool",
+              content: stripAnsiColors(toolBlock.result || ""),
+            });
+
+            // If there are images, also prepare a user message with image content
             if (toolBlock.images && toolBlock.images.length > 0) {
-              // If there is an image, create a user message instead of a tool message
-              const contentParts: ChatCompletionContentPart[] = [];
+              if (supportsVision) {
+                const contentParts: ChatCompletionContentPart[] = [];
 
-              // Add tool result as text
-              const toolResultText = `Tool result for ${toolBlock.name || "unknown tool"}:\n${stripAnsiColors(toolBlock.result || "")}`;
-              contentParts.push({
-                type: "text",
-                text: toolResultText,
-              });
+                toolBlock.images.forEach((image) => {
+                  const imageUrl = image.data.startsWith("data:")
+                    ? image.data
+                    : `data:${image.mediaType || "image/png"};base64,${image.data}`;
 
-              // Add image
-              toolBlock.images.forEach((image) => {
-                const imageUrl = image.data.startsWith("data:")
-                  ? image.data
-                  : `data:${image.mediaType || "image/png"};base64,${image.data}`;
-
-                contentParts.push({
-                  type: "image_url",
-                  image_url: {
-                    url: imageUrl,
-                    detail: "auto",
-                  },
+                  contentParts.push({
+                    type: "image_url",
+                    image_url: {
+                      url: imageUrl,
+                      detail: "auto",
+                    },
+                  });
                 });
-              });
 
-              // Add user message
-              recentMessages.unshift({
-                role: "user",
-                content: contentParts,
-              });
-            } else {
-              // Normal tool message
-              recentMessages.unshift({
-                tool_call_id: toolBlock.id,
-                role: "tool",
-                content: stripAnsiColors(toolBlock.result || ""),
-              });
+                imageUserMessages.push({
+                  role: "user",
+                  content: contentParts,
+                });
+              } else {
+                // Non-vision model: replace images with a text placeholder
+                imageUserMessages.push({
+                  role: "user",
+                  content:
+                    "[Tool returned an image, but the current model does not support image recognition]",
+                });
+              }
             }
           }
         });
+
+        // Insert image user messages after all tool messages but before the
+        // assistant message (which will be unshifted next). Since tool messages
+        // were unshifted to the front, we splice images right after them.
+        if (imageUserMessages.length > 0) {
+          // Tool messages are at indices 0..N-1, existing messages start at N
+          const toolMsgCount = completedToolIds.size;
+          recentMessages.splice(toolMsgCount, 0, ...imageUserMessages);
+        }
       }
 
       // Construct the content of the assistant message
@@ -227,32 +249,40 @@ export function convertMessagesForAPI(
           block.imageUrls &&
           block.imageUrls.length > 0
         ) {
-          block.imageUrls.forEach((imageUrl: string) => {
-            // Check if it's already base64, convert if not
-            let finalImageUrl = imageUrl;
-            if (!imageUrl.startsWith("data:image/")) {
-              // If it's a file path, it needs to be converted to base64
-              try {
-                finalImageUrl = convertImageToBase64(imageUrl);
-              } catch (error) {
-                logger.error(
-                  "Failed to convert image path to base64:",
-                  imageUrl,
-                  error,
-                );
-                // Skip this image, do not add to content
-                return;
+          if (!supportsVision) {
+            // Non-vision model: replace images with a text placeholder
+            contentParts.push({
+              type: "text",
+              text: "[User shared an image, but the current model does not support image recognition]",
+            });
+          } else {
+            block.imageUrls.forEach((imageUrl: string) => {
+              // Check if it's already base64, convert if not
+              let finalImageUrl = imageUrl;
+              if (!imageUrl.startsWith("data:image/")) {
+                // If it's a file path, it needs to be converted to base64
+                try {
+                  finalImageUrl = convertImageToBase64(imageUrl);
+                } catch (error) {
+                  logger.error(
+                    "Failed to convert image path to base64:",
+                    imageUrl,
+                    error,
+                  );
+                  // Skip this image, do not add to content
+                  return;
+                }
               }
-            }
 
-            contentParts.push({
-              type: "image_url",
-              image_url: {
-                url: finalImageUrl,
-                detail: "auto",
-              },
+              contentParts.push({
+                type: "image_url",
+                image_url: {
+                  url: finalImageUrl,
+                  detail: "auto",
+                },
+              });
             });
-          });
+          }
         }
 
         // If there is a tool block in user message, add its result

package/src/utils/mcpUtils.ts

@@ -132,6 +132,7 @@ export function findToolServer(
 ): McpServerStatus | undefined {
   return servers.find(
     (s) =>
-      s.status === "connected" && s.tools?.some((t) => t.name === toolName),
+      (s.status === "connected" || s.status === "reconnecting") &&
+      s.tools?.some((t) => t.name === toolName),
   );
 }

package/src/utils/modelCapabilities.ts

@@ -0,0 +1,30 @@
+/**
+ * Model Capability Detection
+ *
+ * Pure declarative capability accessors. Capabilities are configured per-model
+ * via the `capabilities` field in `ModelConfig` (settings.json), not via regex
+ * or environment variables.
+ */
+
+import type { ModelCapabilities } from "../types/config.js";
+
+/**
+ * Determines if a model supports prompt caching.
+ * @param capabilities - Declarative model capabilities
+ * @returns True if promptCaching is explicitly enabled (default: false)
+ */
+export function supportsPromptCaching(
+  capabilities?: ModelCapabilities,
+): boolean {
+  return capabilities?.promptCaching ?? false;
+}
+
+/**
+ * Determines if a model supports image/vision recognition.
+ * Opt-out semantics: by default all models are assumed to support vision.
+ * @param capabilities - Declarative model capabilities
+ * @returns True unless vision is explicitly disabled (default: true)
+ */
+export function supportsVision(capabilities?: ModelCapabilities): boolean {
+  return capabilities?.vision ?? true;
+}

package/src/utils/skillParser.ts

@@ -57,7 +57,9 @@ export function parseSkillFile(
       skillPath.includes("/.wave/skills") ||
       skillPath.includes("\\.wave\\skills") ||
       skillPath.includes("/.claude/skills") ||
-      skillPath.includes("\\.claude\\skills")
+      skillPath.includes("\\.claude\\skills") ||
+      skillPath.includes("/.agents/skills") ||
+      skillPath.includes("\\.agents\\skills")
         ? "project"
         : "personal";