> For the complete documentation index, see [llms.txt](https://docs.jetadmin.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.jetadmin.io/api-reference/javascript-sdk/agents.md).

# AI agents and conversations

Use `jet.agent(agentUid)` to prompt a configured agent and manage its conversations. These methods use the agent API; they are distinct from editing a database collection named `assistant_conversation`.

### Create a conversation and send a prompt

```typescript
import { Jet } from '@jet-admin/jet-sdk';

async function startConversation(jet: Jet, prompt: string) {
  const agent = jet.agent('YOUR_AGENT_UID');
  const conversation = await agent.createConversation();
  const result = await agent.prompt({ conversation: conversation.id, prompt });
  return {
    conversationId: conversation.id,
    completed: result.result,
    messages: result.messagesCreated,
  };
}
```

Retain `conversation.id` and send it with later prompts to continue the same conversation. `prompt()` resolves to `AgentPromptResult` with `result: boolean`, `messagesCreated: AgentMessage[]`, and `conversationName: string`. It does not expose a conversation ID, so explicitly creating a conversation is useful when you need to retain it.

#### Prompt options

Both `prompt(options?)` and `promptSse(options?)` accept:

| Option               | Type                      | Purpose                                                          |
| -------------------- | ------------------------- | ---------------------------------------------------------------- |
| `conversation`       | `string`                  | Conversation identifier                                          |
| `prompt`             | `string`                  | User's prompt                                                    |
| `params`             | `Record<string, unknown>` | Parameters supported by the configured agent                     |
| `promptImageBase64`  | `string[]`                | Base64 image values; accepted encoding details depend on the API |
| `promptImageURLs`    | `string[]`                | Image URLs                                                       |
| `promptDocumentURLs` | `string[]`                | Document URLs                                                    |
| `confirmMessage`     | `string`                  | Confirmation message value expected by the agent flow            |
| `confirmContext`     | `Record<string, unknown>` | Context expected by that confirmation flow                       |

All options are optional in the SDK types; the server determines valid combinations. The SDK forwards confirmation values without defining their payload contract. Obtain these values from your agent's confirmation flow rather than inventing them.

### Stream agent events

```typescript
import {
  Jet, AgentMessage, AgentPromptRequest, AgentPromptResult,
} from '@jet-admin/jet-sdk';

async function streamReply(jet: Jet, conversationId: string, prompt: string) {
  const agent = jet.agent('YOUR_AGENT_UID');
  for await (const event of agent.promptSse({ conversation: conversationId, prompt })) {
    if (event instanceof AgentMessage) {
      console.log('New message', event.id, event.content);
    } else if (event instanceof AgentPromptRequest) {
      console.log('Request', event.request);
    } else if (event instanceof AgentPromptResult) {
      console.log('Result', event.result, event.messagesCreated);
    }
  }
}
```

`promptSse()` reads Server-Sent Events through `fetch` and yields class instances:

| Server event      | Yielded value                               |
| ----------------- | ------------------------------------------- |
| `message_created` | `AgentMessage`                              |
| `request`         | `AgentPromptRequest` with `request: string` |
| `result`          | `AgentPromptResult`                         |
| `error`           | Throws `JetRequestError`                    |

Other event types are ignored. The SDK has no dedicated token-delta event contract, automatic reconnection, or `AbortSignal` option. Exiting the loop cancels the response reader; it does not guarantee cancellation of the agent's server-side work. See [streaming limitations](/api-reference/javascript-sdk/troubleshooting.md).

### Read conversation history

```typescript
import { Jet } from '@jet-admin/jet-sdk';

async function loadHistory(jet: Jet, conversationId: string) {
  const agent = jet.agent('YOUR_AGENT_UID');
  const page = await agent.getMessages({ conversationId, page: 1 });
  return { messages: page.results, hasMore: page.hasMore };
}
```

| Method                               | Arguments                                    | Resolved value                                                                          |
| ------------------------------------ | -------------------------------------------- | --------------------------------------------------------------------------------------- |
| `getConversations(options?)`         | `{ search?: string, page?: number }`         | `results: AgentConversation[]`, `hasMore`                                               |
| `getConversation(id)`                | Conversation ID                              | `AgentConversation`                                                                     |
| `createConversation()`               | None                                         | `AgentConversation`                                                                     |
| `renameConversation(id, name)`       | ID and new name                              | `AgentConversation`                                                                     |
| `archiveConversation(id)`            | ID                                           | `{ result, otherConversation }`; replacement conversation is populated only if returned |
| `getMessages(options?)`              | `{ page?: number, conversationId?: string }` | `results: AgentMessage[]`, `hasMore`, and conversation when returned                    |
| `getConversationFiles(id, options?)` | ID and `{ page?: number, perPage?: number }` | `results: StorageObject[]`, `hasMore`, `perPage`                                        |

For listing methods, request the next page while the server returns `hasMore: true`. Pagination is manual. `getConversationFiles()` sends the page-size query parameter as `perPage` in this version; confirm compatibility with your endpoint if it ignores the page size.

Conversations include `id`, `name`, `objectType`, `objectId`, `params`, `dateAdd`, `filesCount`, and optional `lastMessage`. Messages include `id`, `type`, `content`, `params`, `dateAdd`, and `parentId`, with user/local ID data when returned.

### Display message content

`message.content` is an array of `AgentContentBlock`, not a single text string. Text is available through `block.markdown?.content`, `block.plainText?.content`, or `block.html?.content`. Other parsed payloads include images, files, calls, confirmations, follow-ups, and questions.

Render only supported block types in your UI, handle unknown types gracefully, and sanitize HTML or rendered Markdown before inserting it into the DOM. The SDK parses data; it does not provide a chat UI or sanitize content. Some enum values have no dedicated payload parser in this version.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.jetadmin.io/api-reference/javascript-sdk/agents.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.