linkshell-cli 0.2.104 → 0.2.106

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "linkshell-cli",
-  "version": "0.2.104",
+  "version": "0.2.106",
   "description": "Remote terminal bridge for Claude Code / Codex / Gemini / Copilot — control from your phone",
   "type": "module",
   "main": "dist/index.js",
@@ -29,9 +29,16 @@
     "type": "git",
     "url": "git+https://github.com/LiuTianjie/LinkShell.git"
   },
+  "scripts": {
+    "build": "tsc -p tsconfig.json",
+    "postinstall": "node scripts/postinstall.mjs",
+    "typecheck": "tsc -p tsconfig.json --noEmit",
+    "dev": "tsx src/index.ts",
+    "test": "vitest run"
+  },
   "dependencies": {
-    "@linkshell/gateway": "^0.2.34",
-    "@linkshell/protocol": "^0.2.29",
+    "@linkshell/gateway": "workspace:^0.2.35",
+    "@linkshell/protocol": "workspace:^0.2.29",
     "commander": "^13.1.0",
     "node-pty": "^1.0.0",
     "qrcode-terminal": "^0.12.0",
@@ -46,12 +53,5 @@
     "@types/ws": "^8.5.14",
     "tsx": "^4.19.3",
     "vitest": "^3.2.1"
-  },
-  "scripts": {
-    "build": "tsc -p tsconfig.json",
-    "postinstall": "node scripts/postinstall.mjs",
-    "typecheck": "tsc -p tsconfig.json --noEmit",
-    "dev": "tsx src/index.ts",
-    "test": "vitest run"
   }
-}
+}

package/src/runtime/acp/agent-session.ts

@@ -11,6 +11,12 @@ import { resolveAgentCommand } from "./provider-resolver.js";
 
 type AgentStatus = "unavailable" | "idle" | "running" | "waiting_permission" | "error";
 
+function protocolSupportsImages(protocol: AgentProtocol | undefined): boolean {
+  return protocol === "codex-app-server" ||
+    protocol === "claude-agent-sdk" ||
+    protocol === "claude-stream-json";
+}
+
 interface AgentMessage {
   id: string;
   role: "user" | "assistant" | "system";
@@ -1032,7 +1038,7 @@ export class AgentSessionProxy {
         error: enabled ? undefined : this.error,
         supportsSessionList: enabled,
         supportsSessionLoad: enabled,
-        supportsImages: false,
+        supportsImages: enabled && protocolSupportsImages(this.activeProtocol),
         supportsAudio: false,
         supportsPermission,
         supportsPlan: enabled,

package/src/runtime/acp/agent-workspace.ts

@@ -502,6 +502,12 @@ function textFromBlocks(blocks: AgentContentBlock[]): string {
     .join("\n");
 }
 
+function protocolSupportsImages(protocol: AgentProtocol | undefined): boolean {
+  return protocol === "codex-app-server" ||
+    protocol === "claude-agent-sdk" ||
+    protocol === "claude-stream-json";
+}
+
 function isSubagentItemType(itemType: string | undefined): boolean {
   const normalized = normalizedIdentifier(itemType);
   return (
@@ -1056,7 +1062,7 @@ export class AgentWorkspaceProxy {
       const protocol = this.agentProtocols.get(provider);
       const runtimeCapabilities = this.providerCapabilities.get(provider);
       const enabled = Boolean(client);
-      const supportsImages = enabled && protocol === "codex-app-server";
+      const supportsImages = enabled && protocolSupportsImages(protocol);
       const isClaudeFallback = protocol === "claude-stream-json";
       const supportsPermission = enabled && !isClaudeFallback;
       const supportsReasoningEffort = enabled && !isClaudeFallback;
@@ -1260,7 +1266,7 @@ export class AgentWorkspaceProxy {
     if (!client) return;
 
     const protocol = this.protocolForProvider(conversation.provider);
-    if (payload.contentBlocks.some((block) => block.type === "image") && protocol !== "codex-app-server") {
+    if (payload.contentBlocks.some((block) => block.type === "image") && !protocolSupportsImages(protocol)) {
       conversation.status = "idle";
       conversation.lastActivityAt = Date.now();
       this.emitConversation(conversation);

package/src/runtime/acp/claude-sdk-client.ts

@@ -32,6 +32,13 @@ interface ClaudeSdkMessage {
 
 type ClaudeQuery = (input: Record<string, unknown>) => AsyncIterable<ClaudeSdkMessage>;
 
+type AgentInputContentBlock = {
+  type?: string;
+  text?: string;
+  data?: string;
+  mimeType?: string;
+};
+
 function id(prefix: string): string {
   return `${prefix}-${Date.now().toString(36)}-${Math.random().toString(36).slice(2, 8)}`;
 }
@@ -64,6 +71,48 @@ function extractToolResultText(content: unknown): string {
   return String(content ?? "");
 }
 
+function splitImageDataUrl(value: string, fallbackMimeType = "image/png"): { data: string; mimeType: string } {
+  const match = value.match(/^data:([^;,]+)?;base64,(.*)$/is);
+  if (!match) return { data: value, mimeType: fallbackMimeType };
+  return {
+    data: match[2] ?? "",
+    mimeType: match[1] || fallbackMimeType,
+  };
+}
+
+function toClaudeMessageContent(blocks: AgentInputContentBlock[]): Record<string, unknown>[] {
+  return blocks
+    .map((block) => {
+      if (block.type === "image" && block.data) {
+        const image = splitImageDataUrl(block.data, block.mimeType);
+        return {
+          type: "image",
+          source: {
+            type: "base64",
+            media_type: image.mimeType,
+            data: image.data,
+          },
+        };
+      }
+      return { type: "text", text: block.text ?? "" };
+    })
+    .filter((block) =>
+      block.type === "image" ||
+      (typeof block.text === "string" && block.text.trim().length > 0),
+    );
+}
+
+async function* singleUserMessage(content: Record<string, unknown>[]): AsyncIterable<Record<string, unknown>> {
+  yield {
+    type: "user",
+    message: {
+      role: "user",
+      content,
+    },
+    parent_tool_use_id: null,
+  };
+}
+
 function isRealClaudeSessionId(value: string | undefined): value is string {
   return Boolean(value && !value.startsWith("agent-session-"));
 }
@@ -168,7 +217,9 @@ export class ClaudeSdkClient {
     const abortController = new AbortController();
     this.abortController = abortController;
 
-    const prompt = (input.content as Array<{ type?: string; text?: string; data?: string; mimeType?: string }>)
+    const inputBlocks = input.content as AgentInputContentBlock[];
+    const hasImages = inputBlocks.some((block) => block.type === "image" && block.data);
+    const prompt = inputBlocks
       .map((block) => {
         if (block.type === "image") return `[${block.mimeType ?? "image"} attachment]`;
         return block.text ?? "";
@@ -221,7 +272,8 @@ export class ClaudeSdkClient {
     let currentMessageId: string | undefined;
 
     try {
-      for await (const message of this.query({ prompt, options: sdkOptions })) {
+      const queryPrompt = hasImages ? singleUserMessage(toClaudeMessageContent(inputBlocks)) : prompt;
+      for await (const message of this.query({ prompt: queryPrompt, options: sdkOptions })) {
         if (abortController.signal.aborted) break;
         this.handleSdkMessage(message, {
           cwd: input.cwd ?? this.input.cwd,

package/src/runtime/acp/claude-stream-json-client.ts

@@ -7,6 +7,13 @@ import type { AgentFraming, AgentProtocol } from "./provider-resolver.js";
 
 type AgentPermissionMode = "read_only" | "workspace_write" | "full_access";
 
+type AgentInputContentBlock = {
+  type?: string;
+  text?: string;
+  data?: string;
+  mimeType?: string;
+};
+
 interface ClaudeStreamEvent {
   type: string;
   subtype?: string;
@@ -65,6 +72,15 @@ function extractToolResultText(content: unknown): string {
   return String(content ?? "");
 }
 
+function splitImageDataUrl(value: string, fallbackMimeType = "image/png"): { data: string; mimeType: string } {
+  const match = value.match(/^data:([^;,]+)?;base64,(.*)$/is);
+  if (!match) return { data: value, mimeType: fallbackMimeType };
+  return {
+    data: match[2] ?? "",
+    mimeType: match[1] || fallbackMimeType,
+  };
+}
+
 export class ClaudeStreamJsonClient {
   private child: ChildProcessWithoutNullStreams | undefined;
   private claudeSessionId: string | undefined;
@@ -145,10 +161,11 @@ export class ClaudeStreamJsonClient {
     }
 
     // Build the user message
-    const contentBlocks = (input.content as Array<{ type: string; text?: string; data?: string; mimeType?: string }>).map(
+    const contentBlocks = (input.content as AgentInputContentBlock[]).map(
       (block) => {
         if (block.type === "image" && block.data) {
-          return { type: "image", source: { type: "base64", media_type: block.mimeType ?? "image/png", data: block.data } };
+          const image = splitImageDataUrl(block.data, block.mimeType);
+          return { type: "image", source: { type: "base64", media_type: image.mimeType, data: image.data } };
         }
         return { type: "text", text: block.text ?? "" };
       },