{"title":"Automatisation par webhooks","slug":"webhook-automation","category":"guides","summary":"Déclenchez des systèmes externes quand les mémoires changent — synchronisation, notification, automatisation.","audience":["human","llm"],"tags":["guide","webhooks","automation","integration"],"difficulty":"intermediate","updated":"2026-06-27","word_count":373,"read_minutes":2,"lang":"fr","translated":true,"requested_lang":"fr","content_markdown":"\n# Automatisation par webhooks\n\nLes webhooks permettent de déclencher des systèmes externes quand des événements\nSynapse se déclenchent. Ce guide couvre les schémas d'automatisation courants.\n\n## Schémas courants\n\n### Schéma 1 : notification sur mémoire critique\n\nEnvoyez un message Slack quand une mémoire critique est stockée :\n\n```python\n# Gestionnaire de webhook (votre serveur)\n@app.post(\"/webhook\")\nasync def handle(request):\n    payload = await request.json()\n    \n    # Vérifier la signature\n    if not verify_signature(payload, request.headers):\n        return 401\n    \n    if payload[\"event\"] == \"memory.store\":\n        memory = payload[\"data\"]\n        if memory.get(\"priority\") == \"critical\":\n            # Envoyer une notification Slack\n            await slack.post(\n                f\"🚨 Critical memory stored: {memory['key']}\\n{memory['content'][:200]}\"\n            )\n    \n    return 200\n```\n\nEnregistrez le webhook :\n\n```bash\ncurl -X POST https://synapse.schaefer.zone/webhooks \\\n  -H \"Authorization: Bearer YOUR_MIND_KEY\" \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\n    \"url\": \"https://my-app.com/webhook\",\n    \"events\": \"memory.store\",\n    \"secret\": \"my-hmac-secret\"\n  }'\n```\n\n### Schéma 2 : synchronisation vers système externe\n\nSynchronisez les mémoires vers Notion, Obsidian ou n'importe quelle KB externe :\n\n```python\n@app.post(\"/webhook\")\nasync def sync_to_notion(request):\n    payload = await request.json()\n    \n    if payload[\"event\"] == \"memory.store\":\n        memory = payload[\"data\"]\n        # Créer une page Notion\n        await notion.create_page(\n            title=memory[\"key\"],\n            content=memory[\"content\"],\n            tags=memory.get(\"tags\", [])\n        )\n    \n    elif payload[\"event\"] == \"memory.delete\":\n        # Supprimer de Notion\n        await notion.delete_page(memory_id=payload[\"data\"][\"id\"])\n    \n    return 200\n```\n\n### Schéma 3 : déclencher CI/CD\n\nDéclenchez un déploiement quand une mémoire « release » est stockée :\n\n```python\n@app.post(\"/webhook\")\nasync def trigger_deploy(request):\n    payload = await request.json()\n    \n    if payload[\"event\"] == \"memory.store\":\n        memory = payload[\"data\"]\n        if memory.get(\"key\", \"\").startswith(\"release_\"):\n            # Déclencher le pipeline GitLab\n            await gitlab.trigger_pipeline(\n                project=\"synapse\",\n                ref=\"main\",\n                variables={\"RELEASE_MEMORY_ID\": memory[\"id\"]}\n            )\n    \n    return 200\n```\n\n### Schéma 4 : réveiller l'agent sur message humain\n\nDéclenchez une exécution d'agent LLM quand un humain envoie un message de chat :\n\n```python\n@app.post(\"/webhook\")\nasync def wake_agent(request):\n    payload = await request.json()\n    \n    if payload[\"event\"] == \"chat.message_received\":\n        message = payload[\"data\"]\n        # Mettre en file le travail de traitement de l'agent\n        await job_queue.enqueue(\n            \"process_message\",\n            message_id=message[\"id\"],\n            content=message[\"content\"]\n        )\n    \n    return 200\n```\n\n### Schéma 5 : agréger les métriques\n\nSuivez la croissance de la mémoire, l'activité chat, la complétion des tâches :\n\n```python\n@app.post(\"/webhook\")\nasync def track_metrics(request):\n    payload = await request.json()\n    event = payload[\"event\"]\n    \n    metrics = {\n        \"memory.store\": \"memories_stored_total\",\n        \"memory.delete\": \"memories_deleted_total\",\n        \"chat.message_received\": \"messages_received_total\",\n        \"task.created\": \"tasks_created_total\",\n        \"task.completed\": \"tasks_completed_total\",\n    }\n    \n    if event in metrics:\n        await prometheus.increment(metrics[event])\n    \n    return 200\n```\n\n## Vérification de signature\n\nVérifiez toujours les signatures de webhook pour empêcher l'usurpation :\n\n```python\nimport hmac\nimport hashlib\n\ndef verify_signature(payload_body: bytes, headers, secret: str) -> bool:\n    signature = headers.get(\"X-Synapse-Signature\", \"\")\n    if not signature.startswith(\"sha256=\"):\n        return False\n    \n    expected = hmac.new(\n        secret.encode(),\n        payload_body,\n        hashlib.sha256\n    ).hexdigest()\n    \n    return hmac.compare_digest(f\"sha256={expected}\", signature)\n```\n\n## Logique de réessai\n\nSynapse réessaie les webhooks échoués avec un backoff exponentiel. Votre gestionnaire\ndevrait :\n\n1. **Renvoyer 200 rapidement** — ne pas faire de travail lourd de manière synchrone\n2. **Mettre en file le travail** — utiliser un système de tâches en arrière-plan\n3. **Être idempotent** — le même événement peut être livré deux fois\n\n```python\n@app.post(\"/webhook\")\nasync def handle(request):\n    payload = await request.json()\n    # Mettre en file pour traitement asynchrone\n    await job_queue.enqueue(\"process_webhook\", payload)\n    # Retourner immédiatement\n    return 200\n```\n\n## Débogage des webhooks\n\n### Consulter l'historique de livraison\n\nLes livraisons de webhook sont journalisées. Vérifiez les livraisons récentes de votre\nwebhook :\n\n```bash\n# Récupérer les détails du webhook y compris les livraisons récentes\ncurl -H \"Authorization: Bearer YOUR_MIND_KEY\" \\\n     https://synapse.schaefer.zone/webhooks/wh_001\n```\n\n### Tester le webhook manuellement\n\n```bash\n# Déclencher un événement de test\ncurl -X POST https://synapse.schaefer.zone/webhooks/wh_001/test \\\n  -H \"Authorization: Bearer YOUR_MIND_KEY\"\n```\n\n### Problèmes courants\n\n| Problème | Correction |\n|-------|-----|\n| Réponses 4xx | Vérifiez que votre gestionnaire renvoie 200 |\n| Réponses 5xx | Erreur serveur — vérifiez les logs de votre application |\n| Timeout | Renvoyez 200 rapidement, mettez le travail en file asynchrone |\n| Livraisons en double | Rendez le gestionnaire idempotent |\n| Non-correspondance de signature | Vérifiez que le secret est correct |\n\n## Bonnes pratiques\n\n> [!TIP]\n> - **Toujours vérifier les signatures** — ne jamais sauter cela\n> - **Renvoyer 200 rapidement** — ne pas bloquer Synapse\n> - **Être idempotent** — gérer les livraisons en double\n> - **Utiliser des événements spécifiques** — `memory.store` pas `*`\n> - **Surveiller les échecs de livraison** — configurez l'alerting\n\n## Prochaines étapes\n\n- [API Webhooks](/docs/api/webhooks)\n- [Cron & Scheduler](/docs/api/cron)\n- [Agent LLM persistant](/docs/guides/persistent-llm-agent)\n","content_html":"<h1>Automatisation par webhooks</h1>\n<p>Les webhooks permettent de déclencher des systèmes externes quand des événements\nSynapse se déclenchent. Ce guide couvre les schémas d&#39;automatisation courants.</p>\n<h2>Schémas courants</h2>\n<h3>Schéma 1 : notification sur mémoire critique</h3>\n<p>Envoyez un message Slack quand une mémoire critique est stockée :</p>\n<pre><code class=\"hljs language-python\"><span class=\"hljs-comment\"># Gestionnaire de webhook (votre serveur)</span>\n<span class=\"hljs-meta\">@app.post(<span class=\"hljs-params\"><span class=\"hljs-string\">&quot;/webhook&quot;</span></span>)</span>\n<span class=\"hljs-keyword\">async</span> <span class=\"hljs-keyword\">def</span> <span class=\"hljs-title function_\">handle</span>(<span class=\"hljs-params\">request</span>):\n    payload = <span class=\"hljs-keyword\">await</span> request.json()\n    \n    <span class=\"hljs-comment\"># Vérifier la signature</span>\n    <span class=\"hljs-keyword\">if</span> <span class=\"hljs-keyword\">not</span> verify_signature(payload, request.headers):\n        <span class=\"hljs-keyword\">return</span> <span class=\"hljs-number\">401</span>\n    \n    <span class=\"hljs-keyword\">if</span> payload[<span class=\"hljs-string\">&quot;event&quot;</span>] == <span class=\"hljs-string\">&quot;memory.store&quot;</span>:\n        memory = payload[<span class=\"hljs-string\">&quot;data&quot;</span>]\n        <span class=\"hljs-keyword\">if</span> memory.get(<span class=\"hljs-string\">&quot;priority&quot;</span>) == <span class=\"hljs-string\">&quot;critical&quot;</span>:\n            <span class=\"hljs-comment\"># Envoyer une notification Slack</span>\n            <span class=\"hljs-keyword\">await</span> slack.post(\n                <span class=\"hljs-string\">f&quot;🚨 Critical memory stored: <span class=\"hljs-subst\">{memory[<span class=\"hljs-string\">&#x27;key&#x27;</span>]}</span>\\n<span class=\"hljs-subst\">{memory[<span class=\"hljs-string\">&#x27;content&#x27;</span>][:<span class=\"hljs-number\">200</span>]}</span>&quot;</span>\n            )\n    \n    <span class=\"hljs-keyword\">return</span> <span class=\"hljs-number\">200</span></code></pre><p>Enregistrez le webhook :</p>\n<pre><code class=\"hljs language-bash\">curl -X POST https://synapse.schaefer.zone/webhooks \\\n  -H <span class=\"hljs-string\">&quot;Authorization: Bearer YOUR_MIND_KEY&quot;</span> \\\n  -H <span class=\"hljs-string\">&quot;Content-Type: application/json&quot;</span> \\\n  -d <span class=\"hljs-string\">&#x27;{\n    &quot;url&quot;: &quot;https://my-app.com/webhook&quot;,\n    &quot;events&quot;: &quot;memory.store&quot;,\n    &quot;secret&quot;: &quot;my-hmac-secret&quot;\n  }&#x27;</span></code></pre><h3>Schéma 2 : synchronisation vers système externe</h3>\n<p>Synchronisez les mémoires vers Notion, Obsidian ou n&#39;importe quelle KB externe :</p>\n<pre><code class=\"hljs language-python\"><span class=\"hljs-meta\">@app.post(<span class=\"hljs-params\"><span class=\"hljs-string\">&quot;/webhook&quot;</span></span>)</span>\n<span class=\"hljs-keyword\">async</span> <span class=\"hljs-keyword\">def</span> <span class=\"hljs-title function_\">sync_to_notion</span>(<span class=\"hljs-params\">request</span>):\n    payload = <span class=\"hljs-keyword\">await</span> request.json()\n    \n    <span class=\"hljs-keyword\">if</span> payload[<span class=\"hljs-string\">&quot;event&quot;</span>] == <span class=\"hljs-string\">&quot;memory.store&quot;</span>:\n        memory = payload[<span class=\"hljs-string\">&quot;data&quot;</span>]\n        <span class=\"hljs-comment\"># Créer une page Notion</span>\n        <span class=\"hljs-keyword\">await</span> notion.create_page(\n            title=memory[<span class=\"hljs-string\">&quot;key&quot;</span>],\n            content=memory[<span class=\"hljs-string\">&quot;content&quot;</span>],\n            tags=memory.get(<span class=\"hljs-string\">&quot;tags&quot;</span>, [])\n        )\n    \n    <span class=\"hljs-keyword\">elif</span> payload[<span class=\"hljs-string\">&quot;event&quot;</span>] == <span class=\"hljs-string\">&quot;memory.delete&quot;</span>:\n        <span class=\"hljs-comment\"># Supprimer de Notion</span>\n        <span class=\"hljs-keyword\">await</span> notion.delete_page(memory_id=payload[<span class=\"hljs-string\">&quot;data&quot;</span>][<span class=\"hljs-string\">&quot;id&quot;</span>])\n    \n    <span class=\"hljs-keyword\">return</span> <span class=\"hljs-number\">200</span></code></pre><h3>Schéma 3 : déclencher CI/CD</h3>\n<p>Déclenchez un déploiement quand une mémoire « release » est stockée :</p>\n<pre><code class=\"hljs language-python\"><span class=\"hljs-meta\">@app.post(<span class=\"hljs-params\"><span class=\"hljs-string\">&quot;/webhook&quot;</span></span>)</span>\n<span class=\"hljs-keyword\">async</span> <span class=\"hljs-keyword\">def</span> <span class=\"hljs-title function_\">trigger_deploy</span>(<span class=\"hljs-params\">request</span>):\n    payload = <span class=\"hljs-keyword\">await</span> request.json()\n    \n    <span class=\"hljs-keyword\">if</span> payload[<span class=\"hljs-string\">&quot;event&quot;</span>] == <span class=\"hljs-string\">&quot;memory.store&quot;</span>:\n        memory = payload[<span class=\"hljs-string\">&quot;data&quot;</span>]\n        <span class=\"hljs-keyword\">if</span> memory.get(<span class=\"hljs-string\">&quot;key&quot;</span>, <span class=\"hljs-string\">&quot;&quot;</span>).startswith(<span class=\"hljs-string\">&quot;release_&quot;</span>):\n            <span class=\"hljs-comment\"># Déclencher le pipeline GitLab</span>\n            <span class=\"hljs-keyword\">await</span> gitlab.trigger_pipeline(\n                project=<span class=\"hljs-string\">&quot;synapse&quot;</span>,\n                ref=<span class=\"hljs-string\">&quot;main&quot;</span>,\n                variables={<span class=\"hljs-string\">&quot;RELEASE_MEMORY_ID&quot;</span>: memory[<span class=\"hljs-string\">&quot;id&quot;</span>]}\n            )\n    \n    <span class=\"hljs-keyword\">return</span> <span class=\"hljs-number\">200</span></code></pre><h3>Schéma 4 : réveiller l&#39;agent sur message humain</h3>\n<p>Déclenchez une exécution d&#39;agent LLM quand un humain envoie un message de chat :</p>\n<pre><code class=\"hljs language-python\"><span class=\"hljs-meta\">@app.post(<span class=\"hljs-params\"><span class=\"hljs-string\">&quot;/webhook&quot;</span></span>)</span>\n<span class=\"hljs-keyword\">async</span> <span class=\"hljs-keyword\">def</span> <span class=\"hljs-title function_\">wake_agent</span>(<span class=\"hljs-params\">request</span>):\n    payload = <span class=\"hljs-keyword\">await</span> request.json()\n    \n    <span class=\"hljs-keyword\">if</span> payload[<span class=\"hljs-string\">&quot;event&quot;</span>] == <span class=\"hljs-string\">&quot;chat.message_received&quot;</span>:\n        message = payload[<span class=\"hljs-string\">&quot;data&quot;</span>]\n        <span class=\"hljs-comment\"># Mettre en file le travail de traitement de l&#x27;agent</span>\n        <span class=\"hljs-keyword\">await</span> job_queue.enqueue(\n            <span class=\"hljs-string\">&quot;process_message&quot;</span>,\n            message_id=message[<span class=\"hljs-string\">&quot;id&quot;</span>],\n            content=message[<span class=\"hljs-string\">&quot;content&quot;</span>]\n        )\n    \n    <span class=\"hljs-keyword\">return</span> <span class=\"hljs-number\">200</span></code></pre><h3>Schéma 5 : agréger les métriques</h3>\n<p>Suivez la croissance de la mémoire, l&#39;activité chat, la complétion des tâches :</p>\n<pre><code class=\"hljs language-python\"><span class=\"hljs-meta\">@app.post(<span class=\"hljs-params\"><span class=\"hljs-string\">&quot;/webhook&quot;</span></span>)</span>\n<span class=\"hljs-keyword\">async</span> <span class=\"hljs-keyword\">def</span> <span class=\"hljs-title function_\">track_metrics</span>(<span class=\"hljs-params\">request</span>):\n    payload = <span class=\"hljs-keyword\">await</span> request.json()\n    event = payload[<span class=\"hljs-string\">&quot;event&quot;</span>]\n    \n    metrics = {\n        <span class=\"hljs-string\">&quot;memory.store&quot;</span>: <span class=\"hljs-string\">&quot;memories_stored_total&quot;</span>,\n        <span class=\"hljs-string\">&quot;memory.delete&quot;</span>: <span class=\"hljs-string\">&quot;memories_deleted_total&quot;</span>,\n        <span class=\"hljs-string\">&quot;chat.message_received&quot;</span>: <span class=\"hljs-string\">&quot;messages_received_total&quot;</span>,\n        <span class=\"hljs-string\">&quot;task.created&quot;</span>: <span class=\"hljs-string\">&quot;tasks_created_total&quot;</span>,\n        <span class=\"hljs-string\">&quot;task.completed&quot;</span>: <span class=\"hljs-string\">&quot;tasks_completed_total&quot;</span>,\n    }\n    \n    <span class=\"hljs-keyword\">if</span> event <span class=\"hljs-keyword\">in</span> metrics:\n        <span class=\"hljs-keyword\">await</span> prometheus.increment(metrics[event])\n    \n    <span class=\"hljs-keyword\">return</span> <span class=\"hljs-number\">200</span></code></pre><h2>Vérification de signature</h2>\n<p>Vérifiez toujours les signatures de webhook pour empêcher l&#39;usurpation :</p>\n<pre><code class=\"hljs language-python\"><span class=\"hljs-keyword\">import</span> hmac\n<span class=\"hljs-keyword\">import</span> hashlib\n\n<span class=\"hljs-keyword\">def</span> <span class=\"hljs-title function_\">verify_signature</span>(<span class=\"hljs-params\">payload_body: <span class=\"hljs-built_in\">bytes</span>, headers, secret: <span class=\"hljs-built_in\">str</span></span>) -&gt; <span class=\"hljs-built_in\">bool</span>:\n    signature = headers.get(<span class=\"hljs-string\">&quot;X-Synapse-Signature&quot;</span>, <span class=\"hljs-string\">&quot;&quot;</span>)\n    <span class=\"hljs-keyword\">if</span> <span class=\"hljs-keyword\">not</span> signature.startswith(<span class=\"hljs-string\">&quot;sha256=&quot;</span>):\n        <span class=\"hljs-keyword\">return</span> <span class=\"hljs-literal\">False</span>\n    \n    expected = hmac.new(\n        secret.encode(),\n        payload_body,\n        hashlib.sha256\n    ).hexdigest()\n    \n    <span class=\"hljs-keyword\">return</span> hmac.compare_digest(<span class=\"hljs-string\">f&quot;sha256=<span class=\"hljs-subst\">{expected}</span>&quot;</span>, signature)</code></pre><h2>Logique de réessai</h2>\n<p>Synapse réessaie les webhooks échoués avec un backoff exponentiel. Votre gestionnaire\ndevrait :</p>\n<ol>\n<li><strong>Renvoyer 200 rapidement</strong> — ne pas faire de travail lourd de manière synchrone</li>\n<li><strong>Mettre en file le travail</strong> — utiliser un système de tâches en arrière-plan</li>\n<li><strong>Être idempotent</strong> — le même événement peut être livré deux fois</li>\n</ol>\n<pre><code class=\"hljs language-python\"><span class=\"hljs-meta\">@app.post(<span class=\"hljs-params\"><span class=\"hljs-string\">&quot;/webhook&quot;</span></span>)</span>\n<span class=\"hljs-keyword\">async</span> <span class=\"hljs-keyword\">def</span> <span class=\"hljs-title function_\">handle</span>(<span class=\"hljs-params\">request</span>):\n    payload = <span class=\"hljs-keyword\">await</span> request.json()\n    <span class=\"hljs-comment\"># Mettre en file pour traitement asynchrone</span>\n    <span class=\"hljs-keyword\">await</span> job_queue.enqueue(<span class=\"hljs-string\">&quot;process_webhook&quot;</span>, payload)\n    <span class=\"hljs-comment\"># Retourner immédiatement</span>\n    <span class=\"hljs-keyword\">return</span> <span class=\"hljs-number\">200</span></code></pre><h2>Débogage des webhooks</h2>\n<h3>Consulter l&#39;historique de livraison</h3>\n<p>Les livraisons de webhook sont journalisées. Vérifiez les livraisons récentes de votre\nwebhook :</p>\n<pre><code class=\"hljs language-bash\"><span class=\"hljs-comment\"># Récupérer les détails du webhook y compris les livraisons récentes</span>\ncurl -H <span class=\"hljs-string\">&quot;Authorization: Bearer YOUR_MIND_KEY&quot;</span> \\\n     https://synapse.schaefer.zone/webhooks/wh_001</code></pre><h3>Tester le webhook manuellement</h3>\n<pre><code class=\"hljs language-bash\"><span class=\"hljs-comment\"># Déclencher un événement de test</span>\ncurl -X POST https://synapse.schaefer.zone/webhooks/wh_001/test \\\n  -H <span class=\"hljs-string\">&quot;Authorization: Bearer YOUR_MIND_KEY&quot;</span></code></pre><h3>Problèmes courants</h3>\n<table>\n<thead>\n<tr>\n<th>Problème</th>\n<th>Correction</th>\n</tr>\n</thead>\n<tbody><tr>\n<td>Réponses 4xx</td>\n<td>Vérifiez que votre gestionnaire renvoie 200</td>\n</tr>\n<tr>\n<td>Réponses 5xx</td>\n<td>Erreur serveur — vérifiez les logs de votre application</td>\n</tr>\n<tr>\n<td>Timeout</td>\n<td>Renvoyez 200 rapidement, mettez le travail en file asynchrone</td>\n</tr>\n<tr>\n<td>Livraisons en double</td>\n<td>Rendez le gestionnaire idempotent</td>\n</tr>\n<tr>\n<td>Non-correspondance de signature</td>\n<td>Vérifiez que le secret est correct</td>\n</tr>\n</tbody></table>\n<h2>Bonnes pratiques</h2>\n<div class=\"callout callout-ok\"></div><h2>Prochaines étapes</h2>\n<ul>\n<li><a href=\"/docs/api/webhooks\">API Webhooks</a></li>\n<li><a href=\"/docs/api/cron\">Cron &amp; Scheduler</a></li>\n<li><a href=\"/docs/guides/persistent-llm-agent\">Agent LLM persistant</a></li>\n</ul>\n","urls":{"html":"/docs/guides/webhook-automation","text":"/docs/guides/webhook-automation?format=text","json":"/docs/guides/webhook-automation?format=json","llm":"/docs/guides/webhook-automation?format=llm"},"translations_available":["en","zh","hi","es","fr","ar","pt","ru","ja","de","it","ko","nl","pl","tr","sv","vi","th","id","uk"]}