Streaming von Nachrichten

Bei der Erstellung einer Nachricht können Sie "stream": true festlegen, um die Antwort inkrementell mithilfe von Server-Sent Events (SSE) zu streamen.

Streaming mit SDKs

Unsere Python und TypeScript SDKs bieten mehrere Möglichkeiten zum Streaming. Das Python SDK ermöglicht sowohl synchrone als auch asynchrone Streams. Weitere Details finden Sie in der Dokumentation der jeweiligen SDKs.

import anthropic

client = anthropic.Anthropic()

with client.messages.stream(
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello"}],
    model="claude-opus-4-20250514",
) as stream:
  for text in stream.text_stream:
      print(text, end="", flush=True)

Event-Typen

Jedes Server-Sent Event enthält einen benannten Event-Typ und zugehörige JSON-Daten. Jedes Event verwendet einen SSE-Event-Namen (z.B. event: message_stop) und enthält den passenden Event-type in seinen Daten. Jeder Stream verwendet den folgenden Event-Ablauf:

message_start: enthält ein Message-Objekt mit leerem content.
Eine Reihe von Inhaltsblöcken, von denen jeder ein content_block_start, ein oder mehrere content_block_delta-Events und ein content_block_stop-Event hat. Jeder Inhaltsblock hat einen index, der seinem Index im endgültigen Message content-Array entspricht.
Ein oder mehrere message_delta-Events, die Änderungen auf oberster Ebene am endgültigen Message-Objekt anzeigen.
Ein abschließendes message_stop-Event.

Die in den usage-Feldern des message_delta-Events angezeigten Token-Zählungen sind kumulativ.

Ping-Events

Event-Streams können auch eine beliebige Anzahl von ping-Events enthalten.

Fehler-Events

Gelegentlich können wir Fehler im Event-Stream senden. Zum Beispiel können Sie während Zeiten hoher Auslastung einen overloaded_error erhalten, der normalerweise einem HTTP 529 in einem Nicht-Streaming-Kontext entsprechen würde:

Beispiel-Fehler

event: error
data: {"type": "error", "error": {"type": "overloaded_error", "message": "Overloaded"}}

Andere Events

Gemäß unserer Versionierungsrichtlinie können wir neue Event-Typen hinzufügen, und Ihr Code sollte unbekannte Event-Typen problemlos verarbeiten können.

Content-Block-Delta-Typen

Jedes content_block_delta-Event enthält ein delta eines Typs, der den content-Block an einem bestimmten index aktualisiert.

Text-Delta

Ein text-Content-Block-Delta sieht wie folgt aus:

Text-Delta

event: content_block_delta
data: {"type": "content_block_delta","index": 0,"delta": {"type": "text_delta", "text": "ello frien"}}

Input-JSON-Delta

Die Deltas für tool_use-Content-Blöcke entsprechen Aktualisierungen für das input-Feld des Blocks. Um maximale Granularität zu unterstützen, sind die Deltas partielle JSON-Strings, während das endgültige tool_use.input immer ein Objekt ist. Sie können die String-Deltas akkumulieren und das JSON parsen, sobald Sie ein content_block_stop-Event erhalten, indem Sie eine Bibliothek wie Pydantic für partielles JSON-Parsing verwenden oder unsere SDKs, die Hilfsfunktionen für den Zugriff auf geparste inkrementelle Werte bieten. Ein tool_use-Content-Block-Delta sieht wie folgt aus:

Input-JSON-Delta

event: content_block_delta
data: {"type": "content_block_delta","index": 1,"delta": {"type": "input_json_delta","partial_json": "{\"location\": \"San Fra"}}}

Hinweis: Unsere aktuellen Modelle unterstützen nur die Ausgabe einer vollständigen Schlüssel- und Werteigenschaft von input auf einmal. Daher kann es bei der Verwendung von Tools zu Verzögerungen zwischen Streaming-Events kommen, während das Modell arbeitet. Sobald ein input-Schlüssel und -Wert akkumuliert sind, geben wir sie als mehrere content_block_delta-Events mit gestückeltem Partial-JSON aus, sodass das Format automatisch eine feinere Granularität in zukünftigen Modellen unterstützen kann.

Thinking-Delta

Bei Verwendung von erweitertem Denken mit aktiviertem Streaming erhalten Sie Denkinhalte über thinking_delta-Events. Diese Deltas entsprechen dem thinking-Feld der thinking-Content-Blöcke. Für Denkinhalte wird ein spezielles signature_delta-Event kurz vor dem content_block_stop-Event gesendet. Diese Signatur wird verwendet, um die Integrität des Denkblocks zu überprüfen. Ein typisches Thinking-Delta sieht wie folgt aus:

Thinking-Delta

event: content_block_delta
data: {"type": "content_block_delta", "index": 0, "delta": {"type": "thinking_delta", "thinking": "Let me solve this step by step:\n\n1. First break down 27 * 453"}}

Das Signatur-Delta sieht wie folgt aus:

Signatur-Delta

event: content_block_delta
data: {"type": "content_block_delta", "index": 0, "delta": {"type": "signature_delta", "signature": "EqQBCgIYAhIM1gbcDa9GJwZA2b3hGgxBdjrkzLoky3dl1pkiMOYds..."}}

Vollständige HTTP-Stream-Antwort

Wir empfehlen dringend, unsere Client-SDKs zu verwenden, wenn Sie den Streaming-Modus nutzen. Wenn Sie jedoch eine direkte API-Integration erstellen, müssen Sie diese Events selbst verarbeiten. Eine Stream-Antwort besteht aus:

Einem message_start-Event
Möglicherweise mehreren Inhaltsblöcken, von denen jeder enthält:
- Ein content_block_start-Event
- Möglicherweise mehrere content_block_delta-Events
- Ein content_block_stop-Event
Einem message_delta-Event
Einem message_stop-Event

Es können auch ping-Events über die gesamte Antwort verteilt sein. Weitere Details zum Format finden Sie unter Event-Typen.

Einfache Streaming-Anfrage

Shell

curl https://api.anthropic.com/v1/messages \
     --header "anthropic-version: 2023-06-01" \
     --header "content-type: application/json" \
     --header "x-api-key: $ANTHROPIC_API_KEY" \
     --data \
'{
  "model": "claude-opus-4-20250514",
  "messages": [{"role": "user", "content": "Hello"}],
  "max_tokens": 256,
  "stream": true
}'

Antwort

event: message_start
data: {"type": "message_start", "message": {"id": "msg_1nZdL29xx5MUA1yADyHTEsnR8uuvGzszyY", "type": "message", "role": "assistant", "content": [], "model": "claude-opus-4-20250514", "stop_reason": null, "stop_sequence": null, "usage": {"input_tokens": 25, "output_tokens": 1}}}

event: content_block_start
data: {"type": "content_block_start", "index": 0, "content_block": {"type": "text", "text": ""}}

event: ping
data: {"type": "ping"}

event: content_block_delta
data: {"type": "content_block_delta", "index": 0, "delta": {"type": "text_delta", "text": "Hello"}}

event: content_block_delta
data: {"type": "content_block_delta", "index": 0, "delta": {"type": "text_delta", "text": "!"}}

event: content_block_stop
data: {"type": "content_block_stop", "index": 0}

event: message_delta
data: {"type": "message_delta", "delta": {"stop_reason": "end_turn", "stop_sequence":null}, "usage": {"output_tokens": 15}}

event: message_stop
data: {"type": "message_stop"}

Streaming-Anfrage mit Tool-Verwendung

In dieser Anfrage bitten wir Claude, ein Tool zu verwenden, um uns das Wetter mitzuteilen.

Shell

  curl https://api.anthropic.com/v1/messages \
    -H "content-type: application/json" \
    -H "x-api-key: $ANTHROPIC_API_KEY" \
    -H "anthropic-version: 2023-06-01" \
    -d '{
      "model": "claude-opus-4-20250514",
      "max_tokens": 1024,
      "tools": [
        {
          "name": "get_weather",
          "description": "Get the current weather in a given location",
          "input_schema": {
            "type": "object",
            "properties": {
              "location": {
                "type": "string",
                "description": "The city and state, e.g. San Francisco, CA"
              }
            },
            "required": ["location"]
          }
        }
      ],
      "tool_choice": {"type": "any"},
      "messages": [
        {
          "role": "user",
          "content": "What is the weather like in San Francisco?"
        }
      ],
      "stream": true
    }'

Antwort

event: message_start
data: {"type":"message_start","message":{"id":"msg_014p7gG3wDgGV9EUtLvnow3U","type":"message","role":"assistant","model":"claude-3-haiku-20240307","stop_sequence":null,"usage":{"input_tokens":472,"output_tokens":2},"content":[],"stop_reason":null}}

event: content_block_start
data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""}}

event: ping
data: {"type": "ping"}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Okay"}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":","}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" let"}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"'s"}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" check"}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" the"}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" weather"}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" for"}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" San"}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Francisco"}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":","}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" CA"}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":":"}}

event: content_block_stop
data: {"type":"content_block_stop","index":0}

event: content_block_start
data: {"type":"content_block_start","index":1,"content_block":{"type":"tool_use","id":"toolu_01T1x1fJ34qAmk2tNTrN7Up6","name":"get_weather","input":{}}}

event: content_block_delta
data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":""}}

event: content_block_delta
data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"{\"location\":"}}

event: content_block_delta
data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":" \"San"}}

event: content_block_delta
data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":" Francisc"}}

event: content_block_delta
data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"o,"}}

event: content_block_delta
data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":" CA\""}}

event: content_block_delta
data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":", "}}

event: content_block_delta
data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"\"unit\": \"fah"}}

event: content_block_delta
data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"renheit\"}"}}

event: content_block_stop
data: {"type":"content_block_stop","index":1}

event: message_delta
data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"output_tokens":89}}

event: message_stop
data: {"type":"message_stop"}

Streaming-Anfrage mit erweitertem Denken

In dieser Anfrage aktivieren wir erweitertes Denken mit Streaming, um Claudes schrittweise Überlegungen zu sehen.

Shell

curl https://api.anthropic.com/v1/messages \
     --header "x-api-key: $ANTHROPIC_API_KEY" \
     --header "anthropic-version: 2023-06-01" \
     --header "content-type: application/json" \
     --data \
'{
    "model": "claude-opus-4-20250514",
    "max_tokens": 20000,
    "stream": true,
    "thinking": {
        "type": "enabled",
        "budget_tokens": 16000
    },
    "messages": [
        {
            "role": "user",
            "content": "What is 27 * 453?"
        }
    ]
}'

Antwort

event: message_start
data: {"type": "message_start", "message": {"id": "msg_01...", "type": "message", "role": "assistant", "content": [], "model": "claude-opus-4-20250514", "stop_reason": null, "stop_sequence": null}}

event: content_block_start
data: {"type": "content_block_start", "index": 0, "content_block": {"type": "thinking", "thinking": ""}}

event: content_block_delta
data: {"type": "content_block_delta", "index": 0, "delta": {"type": "thinking_delta", "thinking": "Let me solve this step by step:\n\n1. First break down 27 * 453"}}

event: content_block_delta
data: {"type": "content_block_delta", "index": 0, "delta": {"type": "thinking_delta", "thinking": "\n2. 453 = 400 + 50 + 3"}}

event: content_block_delta
data: {"type": "content_block_delta", "index": 0, "delta": {"type": "thinking_delta", "thinking": "\n3. 27 * 400 = 10,800"}}

event: content_block_delta
data: {"type": "content_block_delta", "index": 0, "delta": {"type": "thinking_delta", "thinking": "\n4. 27 * 50 = 1,350"}}

event: content_block_delta
data: {"type": "content_block_delta", "index": 0, "delta": {"type": "thinking_delta", "thinking": "\n5. 27 * 3 = 81"}}

event: content_block_delta
data: {"type": "content_block_delta", "index": 0, "delta": {"type": "thinking_delta", "thinking": "\n6. 10,800 + 1,350 + 81 = 12,231"}}

event: content_block_delta
data: {"type": "content_block_delta", "index": 0, "delta": {"type": "signature_delta", "signature": "EqQBCgIYAhIM1gbcDa9GJwZA2b3hGgxBdjrkzLoky3dl1pkiMOYds..."}}

event: content_block_stop
data: {"type": "content_block_stop", "index": 0}

event: content_block_start
data: {"type": "content_block_start", "index": 1, "content_block": {"type": "text", "text": ""}}

event: content_block_delta
data: {"type": "content_block_delta", "index": 1, "delta": {"type": "text_delta", "text": "27 * 453 = 12,231"}}

event: content_block_stop
data: {"type": "content_block_stop", "index": 1}

event: message_delta
data: {"type": "message_delta", "delta": {"stop_reason": "end_turn", "stop_sequence": null}}

event: message_stop
data: {"type": "message_stop"}

Streaming-Anfrage mit Websuche-Tool-Verwendung

Antwort

event: message_start
data: {"type":"message_start","message":{"id":"msg_01G...","type":"message","role":"assistant","model":"claude-3-7-sonnet-20250219","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2679,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":3}}}

event: content_block_start
data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"I'll check"}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" the current weather in New York City for you"}}

event: ping
data: {"type": "ping"}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"."}}

event: content_block_stop
data: {"type":"content_block_stop","index":0}

event: content_block_start
data: {"type":"content_block_start","index":1,"content_block":{"type":"server_tool_use","id":"srvtoolu_014hJH82Qum7Td6UV8gDXThB","name":"web_search","input":{}}}

event: content_block_delta
data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":""}}

event: content_block_delta
data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"{\"query"}}

event: content_block_delta
data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"\":"}}

event: content_block_delta
data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":" \"weather"}}

event: content_block_delta
data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":" NY"}}

event: content_block_delta
data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"C to"}}

event: content_block_delta
data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"day\"}"}}

event: content_block_stop
data: {"type":"content_block_stop","index":1 }

event: content_block_start
data: {"type":"content_block_start","index":2,"content_block":{"type":"web_search_tool_result","tool_use_id":"srvtoolu_014hJH82Qum7Td6UV8gDXThB","content":[{"type":"web_search_result","title":"Weather in New York City in May 2025 (New York) - detailed Weather Forecast for a month","url":"https://world-weather.info/forecast/usa/new_york/may-2025/","encrypted_content":"Ev0DCioIAxgCIiQ3NmU4ZmI4OC1k...","page_age":null},...]}}

event: content_block_stop
data: {"type":"content_block_stop","index":2}

event: content_block_start
data: {"type":"content_block_start","index":3,"content_block":{"type":"text","text":""}}

event: content_block_delta
data: {"type":"content_block_delta","index":3,"delta":{"type":"text_delta","text":"Here's the current weather information for New York"}}

event: content_block_delta
data: {"type":"content_block_delta","index":3,"delta":{"type":"text_delta","text":" City:\n\n# Weather"}}

event: content_block_delta
data: {"type":"content_block_delta","index":3,"delta":{"type":"text_delta","text":" in New York City"}}

event: content_block_delta
data: {"type":"content_block_delta","index":3,"delta":{"type":"text_delta","text":"\n\n"}}

...

event: content_block_stop
data: {"type":"content_block_stop","index":17}

event: message_delta
data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":10682,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":510,"server_tool_use":{"web_search_requests":1}}}

event: message_stop
data: {"type":"message_stop"}

Erste Schritte

Modelle & Preise

Mehr über Claude erfahren

Funktionen

Werkzeuge

Model Context Protocol (MCP)

Anwendungsfälle

Prompt Engineering

Testen & bewerten

Schutzmaßnahmen verstärken

Rechtszentrum

Streaming von Nachrichten

Streaming mit SDKs

Event-Typen

Ping-Events

Fehler-Events

Andere Events

Content-Block-Delta-Typen

Text-Delta

Input-JSON-Delta

Thinking-Delta

Vollständige HTTP-Stream-Antwort

Einfache Streaming-Anfrage

Streaming-Anfrage mit Tool-Verwendung

Streaming-Anfrage mit erweitertem Denken

Streaming-Anfrage mit Websuche-Tool-Verwendung

Erste Schritte

Modelle & Preise

Mehr über Claude erfahren

Funktionen

Werkzeuge

Model Context Protocol (MCP)

Anwendungsfälle

Prompt Engineering

Testen & bewerten

Schutzmaßnahmen verstärken

Rechtszentrum

​Streaming mit SDKs

​Event-Typen

​Ping-Events

​Fehler-Events

​Andere Events

​Content-Block-Delta-Typen

​Text-Delta

​Input-JSON-Delta

​Thinking-Delta

​Vollständige HTTP-Stream-Antwort

​Einfache Streaming-Anfrage

​Streaming-Anfrage mit Tool-Verwendung

​Streaming-Anfrage mit erweitertem Denken

​Streaming-Anfrage mit Websuche-Tool-Verwendung

Streaming mit SDKs

Event-Typen

Ping-Events

Fehler-Events

Andere Events

Content-Block-Delta-Typen

Text-Delta

Input-JSON-Delta

Thinking-Delta

Vollständige HTTP-Stream-Antwort

Einfache Streaming-Anfrage

Streaming-Anfrage mit Tool-Verwendung

Streaming-Anfrage mit erweitertem Denken

Streaming-Anfrage mit Websuche-Tool-Verwendung