{"title":"Session-Start-Pattern","slug":"session-start-pattern","category":"llm-cookbook","summary":"Die kanonische Session-Start-Sequenz, die jeder LLM-Agent befolgen sollte.","audience":["llm"],"tags":["cookbook","session","pattern","startup"],"difficulty":"beginner","updated":"2026-06-27","word_count":214,"read_minutes":1,"llm_context":"ALWAYS at session start: 1) GET /memory/recall, 2) GET /chat/poll, 3) GET /mind/tasks?status=in_progress\nBuild system prompt from recall output.\nProcess unread chat messages before doing new work.\nResume any in_progress tasks before starting new ones.\nStore new learnings as they happen — don't wait until session end.\n","lang":"de","translated":true,"requested_lang":"de","content_markdown":"\n# Session-Start-Pattern\n\nJede LLM-Agent-Session sollte dieser kanonischen Startup-Sequenz folgen. Das\nÜberspringen von Schritten führt zu verlorenem Kontext, verpassten Nachrichten\nund vergessenen Tasks.\n\n## Das Pattern\n\n```\n1. Recall all memories\n2. Poll for unread chat messages\n3. Check in-progress tasks\n4. Build context from results\n5. Process pending items before new work\n```\n\n## Implementierung\n\n### Schritt 1: Alle Memories abrufen\n\n> [!CRITICAL]\n> Das ist der wichtigste Aufruf. Ohne ihn hast du keine Erinnerung an vergangene\n> Sessions.\n\n```bash\ncurl -H \"Authorization: Bearer YOUR_MIND_KEY\" \\\n https://synapse.schaefer.zone/memory/recall\n```\n\nLiefert eine Klartext-Zusammenfassung aller Memories, sortiert nach Priorität.\n\n### Schritt 2: Auf ungelesene Chat-Nachrichten pollen\n\n```bash\ncurl -H \"Authorization: Bearer YOUR_MIND_KEY\" \\\n https://synapse.schaefer.zone/chat/poll\n```\n\nLiefert ungelesene Nachrichten vom Menschen. **Markiert sie automatisch als\ngelesen.**\n\n### Schritt 3: In-Progress-Tasks prüfen\n\n```bash\ncurl -H \"Authorization: Bearer YOUR_MIND_KEY\" \\\n \"https://synapse.schaefer.zone/mind/tasks?status=in_progress\"\n```\n\nLiefert Tasks, an denen du in der letzten Session gearbeitet hast.\n\n### Schritt 4: Kontext aufbauen\n\nKombiniere die drei Antworten in deinem System-Prompt:\n\n```python\ndef build_context(memories, messages, tasks):\n context = f\"\"\"# SESSION CONTEXT\n\n## Memories (from previous sessions)\n{memories}\n\n## Unread Messages from Human\n{format_messages(messages)}\n\n## Active Tasks\n{format_tasks(tasks)}\n\n## Instructions\n- Address unread messages first\n- Resume active tasks before starting new work\n- Store new learnings as they happen (POST /memory)\n- Poll for new messages every 30-60 seconds\n\"\"\"\n return context\n```\n\n### Schritt 5: Pending-Items verarbeiten\n\n```\nFor each unread message:\n - Acknowledge receipt (POST /chat/reply)\n - Address the message content\n - Store any new commitments as memories\n\nFor each in-progress task:\n - Recall why you were working on it\n - Continue from where you left off\n - Update task status as you progress\n```\n\n## Vollständiges Beispiel\n\n```python\nimport os\nimport requests\n\nURL = \"https://synapse.schaefer.zone\"\nKEY = os.environ[\"SYNAPSE_MIND_KEY\"]\n\ndef session_start():\n \"\"\"Canonical session start sequence.\"\"\"\n headers = {\"Authorization\": f\"Bearer {KEY}\"}\n \n # 1. Recall memories\n r = requests.get(f\"{URL}/memory/recall\", headers=headers)\n memories = r.text\n \n # 2. Poll chat\n r = requests.get(f\"{URL}/chat/poll\", headers=headers)\n messages = r.json().get(\"messages\", [])\n \n # 3. Check tasks\n r = requests.get(f\"{URL}/mind/tasks?status=in_progress\", headers=headers)\n tasks = r.json().get(\"tasks\", [])\n \n # 4. Build context\n context = f\"\"\"You are a Synapse-enabled AI assistant.\n\nMEMORIES FROM PREVIOUS SESSIONS:\n{memories}\n\nUNREAD MESSAGES FROM HUMAN:\n{chr(10).join(f'- {m[\"content\"]}' for m in messages) or 'None'}\n\nACTIVE TASKS:\n{chr(10).join(f'- [{t[\"id\"]}] {t[\"title\"]}: {t.get(\"description\", \"\")}' for t in tasks) or 'None'}\n\nINSTRUCTIONS:\n1. Acknowledge each unread message\n2. Resume active tasks\n3. Store new learnings via POST /memory\n4. Poll /chat/poll every 30-60 seconds\n\"\"\"\n return context\n\n# At session start\nsystem_prompt = session_start()\n# Pass to LLM...\n```\n\n## Häufige Fehler\n\n> [!WARNING]\n> - **Recall überspringen** — du startest ohne Kontext, wiederholst vergangene Fehler\n> - **Chat-Poll vergessen** — Nachrichten des Menschen bleiben unbeantwortet\n> - **Aktive Tasks ignorieren** — Arbeit wird mitten in der Ausführung vergessen\n> - **Nichts speichern** — Session produziert keinen persistenten Wert\n\n## Variationen\n\n### Minimales Pattern (Low-Context-LLMs)\n\nFür LLMs mit kleinem Kontext-Fenster überspringe den vollständigen Recall:\n\n```bash\n# Just get stats, not full content\ncurl -H \"Authorization: Bearer $KEY\" .../memory/stats\n```\n\nDann bei Bedarf nach spezifischen Themen suchen:\n\n```bash\ncurl -H \"Authorization: Bearer $KEY\" \".../memory/search?q=current+project\"\n```\n\n### Aggressives Pattern (langlaufende Agenten)\n\nFür Agenten, die stundenlang laufen, füge periodischen Re-Recall hinzu:\n\n```python\nwhile working:\n if time.time() - last_recall > 3600: # every hour\n memories = recall()\n last_recall = time.time()\n # ... do work ...\n```\n\n## Nächste Schritte\n\n- [Memory-Tagging-Strategie](/docs/llm-cookbook/memory-tagging-strategy)\n- [Task-getriebener Workflow](/docs/llm-cookbook/task-driven-workflow)\n- [Chat-Polling-Pattern](/docs/llm-cookbook/chat-polling-pattern)\n","content_html":"

Session-Start-Pattern

\n

Jede LLM-Agent-Session sollte dieser kanonischen Startup-Sequenz folgen. Das\nÜberspringen von Schritten führt zu verlorenem Kontext, verpassten Nachrichten\nund vergessenen Tasks.

\n

Das Pattern

\n
1. Recall all memories\n2. Poll for unread chat messages\n3. Check in-progress tasks\n4. Build context from results\n5. Process pending items before new work

Implementierung

\n

Schritt 1: Alle Memories abrufen

\n
Das ist der wichtigste Aufruf. Ohne ihn hast du keine Erinnerung an vergangene\nSessions.
curl -H "Authorization: Bearer YOUR_MIND_KEY" \\\n     https://synapse.schaefer.zone/memory/recall

Liefert eine Klartext-Zusammenfassung aller Memories, sortiert nach Priorität.

\n

Schritt 2: Auf ungelesene Chat-Nachrichten pollen

\n
curl -H "Authorization: Bearer YOUR_MIND_KEY" \\\n     https://synapse.schaefer.zone/chat/poll

Liefert ungelesene Nachrichten vom Menschen. Markiert sie automatisch als\ngelesen.

\n

Schritt 3: In-Progress-Tasks prüfen

\n
curl -H "Authorization: Bearer YOUR_MIND_KEY" \\\n     "https://synapse.schaefer.zone/mind/tasks?status=in_progress"

Liefert Tasks, an denen du in der letzten Session gearbeitet hast.

\n

Schritt 4: Kontext aufbauen

\n

Kombiniere die drei Antworten in deinem System-Prompt:

\n
def build_context(memories, messages, tasks):\n    context = f"""# SESSION CONTEXT\n\n## Memories (from previous sessions)\n{memories}\n\n## Unread Messages from Human\n{format_messages(messages)}\n\n## Active Tasks\n{format_tasks(tasks)}\n\n## Instructions\n- Address unread messages first\n- Resume active tasks before starting new work\n- Store new learnings as they happen (POST /memory)\n- Poll for new messages every 30-60 seconds\n"""\n    return context

Schritt 5: Pending-Items verarbeiten

\n
For each unread message:\n  - Acknowledge receipt (POST /chat/reply)\n  - Address the message content\n  - Store any new commitments as memories\n\nFor each in-progress task:\n  - Recall why you were working on it\n  - Continue from where you left off\n  - Update task status as you progress

Vollständiges Beispiel

\n
import os\nimport requests\n\nURL = "https://synapse.schaefer.zone"\nKEY = os.environ["SYNAPSE_MIND_KEY"]\n\ndef session_start():\n    """Canonical session start sequence."""\n    headers = {"Authorization": f"Bearer {KEY}"}\n    \n    # 1. Recall memories\n    r = requests.get(f"{URL}/memory/recall", headers=headers)\n    memories = r.text\n    \n    # 2. Poll chat\n    r = requests.get(f"{URL}/chat/poll", headers=headers)\n    messages = r.json().get("messages", [])\n    \n    # 3. Check tasks\n    r = requests.get(f"{URL}/mind/tasks?status=in_progress", headers=headers)\n    tasks = r.json().get("tasks", [])\n    \n    # 4. Build context\n    context = f"""You are a Synapse-enabled AI assistant.\n\nMEMORIES FROM PREVIOUS SESSIONS:\n{memories}\n\nUNREAD MESSAGES FROM HUMAN:\n{chr(10).join(f'- {m["content"]}' for m in messages) or 'None'}\n\nACTIVE TASKS:\n{chr(10).join(f'- [{t["id"]}] {t["title"]}: {t.get("description", "")}' for t in tasks) or 'None'}\n\nINSTRUCTIONS:\n1. Acknowledge each unread message\n2. Resume active tasks\n3. Store new learnings via POST /memory\n4. Poll /chat/poll every 30-60 seconds\n"""\n    return context\n\n# At session start\nsystem_prompt = session_start()\n# Pass to LLM...

Häufige Fehler

\n

Variationen

\n

Minimales Pattern (Low-Context-LLMs)

\n

Für LLMs mit kleinem Kontext-Fenster überspringe den vollständigen Recall:

\n
# Just get stats, not full content\ncurl -H "Authorization: Bearer $KEY" .../memory/stats

Dann bei Bedarf nach spezifischen Themen suchen:

\n
curl -H "Authorization: Bearer $KEY" ".../memory/search?q=current+project"

Aggressives Pattern (langlaufende Agenten)

\n

Für Agenten, die stundenlang laufen, füge periodischen Re-Recall hinzu:

\n
while working:\n    if time.time() - last_recall > 3600:  # every hour\n        memories = recall()\n        last_recall = time.time()\n    # ... do work ...

Nächste Schritte

\n\n","urls":{"html":"/docs/llm-cookbook/session-start-pattern","text":"/docs/llm-cookbook/session-start-pattern?format=text","json":"/docs/llm-cookbook/session-start-pattern?format=json","llm":"/docs/llm-cookbook/session-start-pattern?format=llm"},"translations_available":["en","zh","hi","es","fr","ar","pt","ru","ja","de","it","ko","nl","pl","tr","sv","vi","th","id","uk"]}