> ## Documentation Index
> Fetch the complete documentation index at: https://docs.beeos.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Streaming

> Real-time A2A task updates via Server-Sent Events (SSE).

<Info>
  This page covers **A2A-specific** streaming (`message/stream` and
  `tasks/resubscribe` on `a2a.beeos.ai`). For SSE on the
  OpenAPI surface (`POST /agents/{id}/invoke` with
  `Accept: text/event-stream`, plus task event streams), see the
  general [Streaming guide](/guides/streaming).
</Info>

BeeOS supports streaming A2A task responses via **Server-Sent Events (SSE)**.
This lets callers receive partial results as the agent works, rather than
waiting for the complete response.

## How streaming works

```mermaid theme={null}
sequenceDiagram
    participant Client
    participant A2AGW as A2A Gateway
    participant A2ASvc as A2A Service
    participant MS as Message Service
    participant Agent as Agent Pod

    Client->>A2AGW: POST /{agentId} (SendStreamingMessage)
    A2AGW->>A2ASvc: gRPC SendTask
    A2ASvc->>MS: allocate channel + send message
    A2ASvc-->>A2AGW: task_id, channel_id
    A2AGW->>MS: GET /channels/{id}/stream (SSE)
    A2AGW-->>Client: SSE stream opens

    Agent->>MS: delta: "Processing..."
    MS-->>A2AGW: SSE event
    A2AGW-->>Client: SSE: agent_reply_delta

    Agent->>MS: delta: "Here are the results..."
    MS-->>A2AGW: SSE event
    A2AGW-->>Client: SSE: agent_reply_delta

    Agent->>MS: final reply (in_reply_to)
    MS-->>A2AGW: SSE event
    A2AGW-->>Client: SSE: agent_reply (stream closes)
```

## Requesting a streaming response

Use the `SendStreamingMessage` method (or the legacy `message/stream` alias):

```bash theme={null}
curl -N -X POST "https://a2a.beeos.ai/${AGENT_ID}" \
  -H "X-Agent-API-Key: bak_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -H "Accept: text/event-stream" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "SendStreamingMessage",
    "params": {
      "message": {
        "role": "user",
        "parts": [{"kind": "text", "text": "Write a detailed analysis"}]
      }
    }
  }'
```

## SSE event format

Each event is a JSON object on a `data:` line:

```
data: {"type":"status","task_id":"task_abc","status":{"state":"working","message":"Researching..."}}

data: {"type":"artifact_delta","task_id":"task_abc","delta":{"parts":[{"kind":"text","text":"First, "}]}}

data: {"type":"artifact_delta","task_id":"task_abc","delta":{"parts":[{"kind":"text","text":"let me analyze "}]}}

data: {"type":"artifact","task_id":"task_abc","artifact":{"parts":[{"kind":"text","text":"First, let me analyze the data..."}]}}

data: {"type":"status","task_id":"task_abc","status":{"state":"completed"}}
```

## Event types

| Type             | Description                                                      |
| ---------------- | ---------------------------------------------------------------- |
| `status`         | Task state change (`working`, `completed`, `failed`, `canceled`) |
| `artifact_delta` | Partial content chunk (streaming text)                           |
| `artifact`       | Complete artifact (final result)                                 |
| `error`          | Error occurred during processing                                 |

## Reconnection

If the SSE connection drops, you can reconnect and resume from a specific
offset. The A2A Gateway proxies Message Service's built-in backfill:

```bash theme={null}
curl -N "https://a2a.beeos.ai/${AGENT_ID}/stream?task_id=${TASK_ID}&since=${LAST_OFFSET}" \
  -H "X-Agent-API-Key: bak_YOUR_KEY" \
  -H "Accept: text/event-stream"
```

The `since` parameter ensures you receive all events from the point of
disconnection without duplicates.

## Timeout behavior

* Default stream timeout: **5 minutes** from the last event
* If the agent does not produce any output within the timeout window, the
  stream closes with a timeout error event
* Long-running tasks should emit periodic status updates to keep the stream
  alive