@mobileai/react-native 0.5.9 → 0.5.11

package/README.md

@@ -32,6 +32,8 @@
 - 🧭 **Auto-navigation** — Navigates between screens to complete multi-step tasks.
 - 🧩 **Custom actions** — Expose any business logic (checkout, API calls) as AI-callable tools with `useAction`.
 - 🌐 **MCP bridge** — Let external AI agents (OpenClaw, Claude Desktop) control your app remotely.
+- 🧠 **Knowledge base** — Give the AI domain knowledge (policies, FAQs, product info) it can query on demand. Static array or bring your own retriever.
+- 💡 **Knowledge-only mode** — Set `enableUIControl={false}` for a lightweight AI assistant with no UI interaction — single LLM call, ~70% fewer tokens.
 
 
 ### 🎤 Voice Mode (Live Agent)
@@ -182,6 +184,77 @@ export default function RootLayout() {
 
 A floating chat bar appears automatically. Ask the AI to navigate, tap buttons, fill forms — it reads your live UI and acts.
 
+## 🧠 Knowledge Base
+
+Give the AI domain-specific knowledge it can query on demand — policies, FAQs, product details, etc. The AI uses a `query_knowledge` tool to fetch relevant entries only when needed (no token waste).
+
+### Static Array (Simple)
+
+Pass an array of entries — the SDK handles keyword-based retrieval internally:
+
+```tsx
+import type { KnowledgeEntry } from '@mobileai/react-native';
+
+const KNOWLEDGE: KnowledgeEntry[] = [
+  {
+    id: 'shipping',
+    title: 'Shipping Policy',
+    content: 'Free shipping on orders over $75. Standard: 5-7 days ($4.99). Express: 2-3 days ($12.99).',
+    tags: ['shipping', 'delivery', 'free shipping'],
+  },
+  {
+    id: 'returns',
+    title: 'Return Policy',
+    content: '30-day returns on all items. Refunds processed in 5-7 business days.',
+    tags: ['return', 'refund', 'exchange'],
+    screens: ['product/[id]', 'order-history'], // optional: only surface on these screens
+  },
+];
+
+<AIAgent knowledgeBase={KNOWLEDGE} />
+```
+
+### Custom Retriever (Advanced)
+
+Bring your own retrieval logic — call an API, vector database, or any async source:
+
+```tsx
+<AIAgent
+  knowledgeBase={{
+    retrieve: async (query: string, screenName?: string) => {
+      const results = await fetch(`/api/knowledge?q=${query}&screen=${screenName}`);
+      return results.json();
+    },
+  }}
+/>
+```
+
+The retriever receives the user's question and current screen name, and returns a formatted string with the relevant knowledge.
+
+### Knowledge-Only Mode
+
+Set `enableUIControl={false}` to disable all UI interactions. The AI becomes a lightweight FAQ / support assistant — no screen analysis, no multi-step agent loop, just question → answer:
+
+```tsx
+<AIAgent
+  enableUIControl={false}
+  knowledgeBase={KNOWLEDGE}
+/>
+```
+
+**How it differs from full mode:**
+
+| | Full mode (default) | Knowledge-only mode |
+|---|---|---|
+| UI tree analysis | ✅ Full fiber walk | ❌ Skipped |
+| Screen content sent to LLM | ✅ ~500-2000 tokens | ❌ Only screen name |
+| Screenshots | ✅ Optional | ❌ Skipped |
+| Agent loop | Up to 10 steps | Single LLM call |
+| Available tools | 7 (tap, type, navigate, ...) | 2 (done, query_knowledge) |
+| System prompt | ~1,500 tokens | ~400 tokens |
+
+The AI still knows the current **screen name** (from navigation state, zero cost), so `screens`-filtered knowledge entries work correctly. It just can't see what's *on* the screen — ideal for domain Q&A where answers come from knowledge, not UI.
+
 ## 🔌 API Reference
 
 ### `<AIAgent>` Component
@@ -415,77 +488,6 @@ This starts two servers:
 | `ask_user(question)` | Ask the user for clarification. |
 | `query_knowledge(question)` | Search the knowledge base. Only available when `knowledgeBase` is configured. |
 
-## 🧠 Knowledge Base
-
-Give the AI domain-specific knowledge it can query on demand — policies, FAQs, product details, etc. The AI uses a `query_knowledge` tool to fetch relevant entries only when needed (no token waste).
-
-### Static Array (Simple)
-
-Pass an array of entries — the SDK handles keyword-based retrieval internally:
-
-```tsx
-import type { KnowledgeEntry } from '@mobileai/react-native';
-
-const KNOWLEDGE: KnowledgeEntry[] = [
-  {
-    id: 'shipping',
-    title: 'Shipping Policy',
-    content: 'Free shipping on orders over $75. Standard: 5-7 days ($4.99). Express: 2-3 days ($12.99).',
-    tags: ['shipping', 'delivery', 'free shipping'],
-  },
-  {
-    id: 'returns',
-    title: 'Return Policy',
-    content: '30-day returns on all items. Refunds processed in 5-7 business days.',
-    tags: ['return', 'refund', 'exchange'],
-    screens: ['product/[id]', 'order-history'], // optional: only surface on these screens
-  },
-];
-
-<AIAgent knowledgeBase={KNOWLEDGE} />
-```
-
-### Custom Retriever (Advanced)
-
-Bring your own retrieval logic — call an API, vector database, or any async source:
-
-```tsx
-<AIAgent
-  knowledgeBase={{
-    retrieve: async (query: string, screenName?: string) => {
-      const results = await fetch(`/api/knowledge?q=${query}&screen=${screenName}`);
-      return results.json();
-    },
-  }}
-/>
-```
-
-The retriever receives the user's question and current screen name, and returns a formatted string with the relevant knowledge.
-
-### Knowledge-Only Mode
-
-Set `enableUIControl={false}` to disable all UI interactions. The AI becomes a lightweight FAQ / support assistant — no screen analysis, no multi-step agent loop, just question → answer:
-
-```tsx
-<AIAgent
-  enableUIControl={false}
-  knowledgeBase={KNOWLEDGE}
-/>
-```
-
-**How it differs from full mode:**
-
-| | Full mode (default) | Knowledge-only mode |
-|---|---|---|
-| UI tree analysis | ✅ Full fiber walk | ❌ Skipped |
-| Screen content sent to LLM | ✅ ~500-2000 tokens | ❌ Only screen name |
-| Screenshots | ✅ Optional | ❌ Skipped |
-| Agent loop | Up to 10 steps | Single LLM call |
-| Available tools | 7 (tap, type, navigate, ...) | 2 (done, query_knowledge) |
-| System prompt | ~1,500 tokens | ~400 tokens |
-
-The AI still knows the current **screen name** (from navigation state, zero cost), so `screens`-filtered knowledge entries work correctly. It just can't see what's *on* the screen — ideal for domain Q&A where answers come from knowledge, not UI.
-
 ## 📋 Requirements
 
 - React Native 0.72+

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mobileai/react-native",
-  "version": "0.5.9",
+  "version": "0.5.11",
   "description": "Build autonomous AI agents for React Native and Expo apps. Provides AI-native UI traversal, tool calling, and structured reasoning.",
   "main": "./lib/module/index.js",
   "source": "./src/index.ts",