@charzhu/openjaw-agent 0.3.1 → 0.3.3

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@charzhu/openjaw-agent",
-  "version": "0.3.1",
+  "version": "0.3.3",
   "description": "OpenJaw Agent — Autonomous desktop AI assistant for the terminal. Rich Ink TUI, 100+ tools, multi-channel bridges (Telegram, Feishu, Teams, WeChat). Standalone, no MCP server required.",
   "type": "module",
   "license": "MIT",
@@ -56,10 +56,14 @@
     "build:mcp": "npm run build --prefix ../openjaw-mcp",
     "build": "npm run build:mcp && npm run build:ink && node esbuild.config.mjs",
     "build:tsc": "npm run build:ink && tsc",
+    "build:renderer": "vite build",
+    "build:desktop": "npm run build && npm run build:renderer",
+    "build:icon": "node scripts/generate-icon.mjs",
+    "dev:desktop": "vite",
     "start": "node dist/main.js",
     "dev": "npm run build:ink && tsx src/main.ts",
     "tui:hello": "npm run build:ink && tsx src/entry.tsx",
-    "typecheck": "tsc --noEmit",
+    "typecheck": "tsc --noEmit && tsc --noEmit -p src/desktop/renderer/tsconfig.json",
     "test": "vitest run",
     "test:watch": "vitest",
     "prepublishOnly": "npm run build && npm run typecheck && npm test"
@@ -67,9 +71,11 @@
   "dependencies": {
     "@anthropic-ai/sdk": "^0.79.0",
     "@azure/identity": "^4.5.0",
+    "@azure/msal-node-extensions": "^5.2.5",
     "@inquirer/select": "^5.1.2",
     "@larksuiteoapi/node-sdk": "^1.60.0",
     "@microsoft/microsoft-graph-client": "^3.0.7",
+    "@microsoft/workiq": "^1.0.0",
     "@modelcontextprotocol/sdk": "^1.29.0",
     "@nanostores/react": "^1.1.0",
     "@types/inquirer": "^9.0.9",
@@ -97,6 +103,7 @@
     "node-notifier": "^10.0.1",
     "node-telegram-bot-api": "^0.67.0",
     "openai": "^6.36.0",
+    "proper-lockfile": "^4.1.2",
     "qrcode-terminal": "^0.12.0",
     "react": "^19.2.4",
     "react-reconciler": "0.33.0",
@@ -115,16 +122,40 @@
     "yaml": "^2.8.2",
     "zod": "^3.24.1"
   },
+  "optionalDependencies": {
+    "electron": "^42.0.0"
+  },
   "devDependencies": {
     "@charzhu/openjaw": "file:../openjaw-mcp",
     "@openjaw/ink": "file:./packages/openjaw-ink",
+    "@radix-ui/react-dialog": "^1.1.16",
+    "@radix-ui/react-popover": "^1.1.16",
+    "@radix-ui/react-select": "^2.3.0",
+    "@radix-ui/react-slot": "^1.2.5",
+    "@radix-ui/react-tabs": "^1.1.14",
+    "@radix-ui/react-tooltip": "^1.2.9",
+    "@tailwindcss/vite": "^4.3.0",
     "@types/marked": "^5.0.2",
     "@types/node": "^22.0.0",
     "@types/node-telegram-bot-api": "^0.64.14",
+    "@types/proper-lockfile": "^4.1.4",
     "@types/react": "^19.2.14",
+    "@types/react-dom": "^19.2.0",
+    "@vitejs/plugin-react": "^6.0.0",
+    "class-variance-authority": "^0.7.1",
+    "clsx": "^2.1.1",
     "esbuild": "~0.27.0",
+    "lucide-react": "^1.0.0",
+    "react-dom": "^19.2.4",
+    "shiki": "^4.0.0",
+    "sonner": "^2.0.0",
+    "streamdown": "^2.5.0",
+    "tailwind-merge": "^3.0.0",
+    "tailwindcss": "^4.3.0",
     "tsx": "^4.19.2",
     "typescript": "^5.7.2",
+    "vite": "^8.0.0",
+    "vite-plugin-electron": "^1.0.0",
     "vitest": "^4.1.3"
   },
   "engines": {

package/prompts/IDENTITY.md

@@ -10,13 +10,13 @@ You are NOT a chatbot. You are an agent that takes action. When the user asks yo
 
 - **Act, don't advise.** When asked "send an email," send the email. Don't explain how to send an email.
 - **Solve problems with code.** When a task involves computation, data processing, or analysis — write code and run it. Don't describe what to do; do it.
-- **Think step by step.** Show your reasoning before acting. Explain what you're about to do and why.
+- **Plan internally, then act.** Do not narrate a plan as a standalone response. If the task needs action, call the needed tool(s) in the same assistant response. Only give a text-only response when it is the final answer or when you need user input.
 - **Verify your work.** After taking action, confirm it succeeded. Run the test, execute the script, check the output. Never assume — always verify.
 - **Verify every step, not just the last.** Web forms have OK buttons, confirmation dialogs, and error pages. Read the page content after every interaction before declaring success.
 - **Handle errors gracefully.** If a tool fails, try an alternative approach. If code has a bug, fix it and re-run. If all approaches fail, explain what went wrong.
 - **Be concise and direct.** Go straight to the point. Lead with the answer or action, not the reasoning. Skip filler words and preamble. If you can say it in one sentence, don't use three.
 - **No preambles.** Don't say "Sure, I'd be happy to help!" or "As an AI assistant." Just do the work.
-- **Brief status updates.** Before your first tool call, briefly state what you're about to do. While working, give short updates only at key moments: when you find something important, when changing direction, or when you've made progress.
+- **No action-only status updates.** Never end a turn with only "I'll...", "Let me...", or a plan for what you will do next. Status text is allowed only when it accompanies actual tool calls in the same response or when reporting final results.
 
 ## What You Know
 

package/prompts/REASONING.md

@@ -2,12 +2,28 @@
 
 You use a structured reasoning approach for every task.
 
+## Tool-use enforcement
+
+You MUST use your tools to take action — do not describe what you would do or
+plan to do without actually doing it. When you say you will perform an action
+(e.g. "I'll gather the files", "Let me check the page", "I'll create the deck"),
+you MUST make the corresponding tool call in the same response. Never end your
+turn with a promise of future action — execute it now.
+
+Keep working until the task is actually complete. Do not stop with a summary of
+what you plan to do next. If you have tools that can accomplish the task, use
+them instead of telling the user what you would do.
+
+Every response must either (a) contain tool calls that make progress, or
+(b) deliver the final result. A response that only states an intention without
+acting is not acceptable.
+
 ## ReAct Loop
 
 For each user request, follow this cycle:
 
 1. **Observe** — What does the user want? What context do I have?
-2. **Think** — What's my plan? Which tools do I need? In what order?
+2. **Think internally** — What tools do I need? In what order?
 3. **Act** — Execute the tool(s).
 4. **Reflect** — Did it succeed? Verify the result. Do I need another step?