experimental-agent 0.1.2 → 0.1.4

package/README.md

@@ -191,6 +191,13 @@ const stream = await session.stream();
 await session.tag.set("category", "support");
 ```
 
+### Handles vs Records
+
+- `myAgent.session(id)` and `myAgent.sandbox(id)` return orchestration handles.
+- Handles are use-case APIs (send, stream, run tools, resume sandbox lifecycle).
+- `myAgent.storage.*` is the raw record API (`Session`, `SandboxRecord`, `Message`, `Part`, `Command`, `SetupSnapshot`).
+- Use `storage` for record lookups and filters (for example existence checks and list queries).
+
 ### Durable UI
 
 Unlike `useChat` alone, messages survive refresh. They're persisted to storage.
@@ -293,6 +300,8 @@ The SDK stores: `Session`, `Message`, `Part`, `SandboxRecord`.
 
 Everything else—users, titles, access control—belongs in your database. Session ID is your join key.
 
+For session queries scoped to a sandbox, use `storage.session.listBySandbox({ sandboxId, ... })`.
+
 ### Vercel Agent Storage
 
 The default `{ type: "vercel" }` uses the hosted Vercel Agent Storage service:

package/dist/agent-workflow.d.mts

@@ -1,5 +1,5 @@
 import * as workflow from 'workflow';
-import { u as StorageConfig, g as RpcPayload, h as RpcResult } from './client-CKLwB-ES.mjs';
+import { u as StorageConfig, g as RpcPayload, h as RpcResult } from './client-Bkuq-Dfa.mjs';
 import 'errore';
 import 'ai';
 import 'zod';

package/dist/agent-workflow.d.ts

@@ -1,5 +1,5 @@
 import * as workflow from 'workflow';
-import { u as StorageConfig, g as RpcPayload, h as RpcResult } from './client-CKLwB-ES.js';
+import { u as StorageConfig, g as RpcPayload, h as RpcResult } from './client-Bkuq-Dfa.js';
 import 'errore';
 import 'ai';
 import 'zod';