[{"data":1,"prerenderedAt":6392},["ShallowReactive",2],{"navigation_fr":3,"posts_fr:/fr/blog/overvibing":51,"posts_fr:/fr/blog/overvibing:surround":1319},[4],{"title":5,"path":6,"stem":7,"children":8,"page":50},"Fr","/fr","fr",[9],{"title":10,"path":11,"stem":12,"children":13,"page":50},"Blog","/fr/blog","fr/blog",[14,18,22,26,30,34,38,42,46],{"title":15,"path":16,"stem":17},"Serveurs MCP comme Serveurs de Ressources OAuth : Une Approche Simplifiée","/fr/blog/mcp-aouth","fr/blog/1. mcp-aouth",{"title":19,"path":20,"stem":21},"Pratiques de Sécurité pour MCP Utilisant JSON-RPC","/fr/blog/mcp-security","fr/blog/1. mcp-security",{"title":23,"path":24,"stem":25},"Serveurs MCP : Connecter l'IA aux Données en Temps Réel","/fr/blog/mcp-create","fr/blog/2. mcp-create",{"title":27,"path":28,"stem":29},"Agents MCP Tiny On-Premises : S'affranchir des Dépendances Cloud","/fr/blog/mcp-tiny-agents-on-premises","fr/blog/5. mcp-tiny-agents-on-premises",{"title":31,"path":32,"stem":33},"De l'Idée à l'API en 2 Jours : Construction de Bankly avec les Flux de Travail Agentiques","/fr/blog/overvibing","fr/blog/6. overvibing",{"title":35,"path":36,"stem":37},"Chatbot IA, Serveur MCP construit avec le Protocole de Flux de Travail Agentique pour démo sur Checkatrade.com","/fr/blog/mcp-in-chat-demo","fr/blog/8. mcp-in-chat-demo",{"title":39,"path":40,"stem":41},"DDD et les Frameworks Full-Stack","/fr/blog/architecture","fr/blog/architecture",{"title":43,"path":44,"stem":45},"Qualité : Comment protéger les Frontends alimentés par l'IA ?","/fr/blog/qa-front","fr/blog/qa-front",{"title":47,"path":48,"stem":49},"Consultation et Stratégie IT","/fr/blog/strategy","fr/blog/strategy",false,{"id":52,"title":31,"authors":53,"badge":60,"body":64,"date":1311,"description":1312,"extension":1313,"image":1314,"meta":1316,"navigation":291,"path":32,"seo":1317,"stem":33,"__hash__":1318},"posts_fr/fr/blog/6. overvibing.md",[54],{"name":55,"description":56,"to":57,"avatar":58},"Michael Wybraniec","Freelance, MCP Servers, Développement Full-Stack, Architecture","https://www.linkedin.com/in/one-front/",{"src":59},"https://media.licdn.com/dms/image/v2/D4D03AQEGvIVcrTTS2g/profile-displayphoto-shrink_800_800/B4DZbtSf7LIEAc-/0/1747737772225?e=1753315200&v=beta&t=EOviZQkM396PoGctVjDCdlG8U3vN5UKWiIeNQ8IFUgQ",{"label":61,"color":62,"variant":63},"Développement Agentique, Overvibing","success","soft",{"type":65,"value":66,"toc":1296},"minimark",[67,88,95,101,329,350,353,381,396,406,421,575,590,626,631,654,669,675,747,753,795,808,813,822,828,1067,1084,1099,1131,1146,1197,1212,1248,1263,1279,1292],[68,69,72],"div",{"className":70},[71],"text-justify",[73,74,75,79,80,83,84,87],"p",{},[76,77,78],"strong",{},"Bankly",", un système bancaire backend construit en seulement deux jours en utilisant ",[76,81,82],{},"Apollo GraphQL"," et ",[76,85,86],{},"Prisma",". Conçu comme un test technique personnel et une preuve de concept qui démontre comment les flux de travail agentiques et le développement assisté par l'IA peuvent accélérer la livraison de logiciels modernes tout en maintenant la qualité et la structure.",[89,90,92],"h2",{"id":91},"chronologie-du-développement",[76,93,94],{},"Chronologie du Développement",[68,96,98],{"className":97},[71],[73,99,100],{},"Voici la progression complète commit par commit montrant comment l'AWP a permis un développement systématique sans perdre d'élan :",[102,103,108],"pre",{"className":104,"code":105,"language":106,"meta":107,"style":107},"language-mermaid shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","---\ntitle: Chronologie du Développement de Bankly - 24-25 Juin 2025 @michaelwybraniec\n---\ngitGraph\n commit id: \"Diagrammes ER\" tag: \"3.1.2\" type: HIGHLIGHT\n commit id: \"Modèles Prisma\" tag: \"3.1.1\"\n commit id: \"Politique AWP → Protocole\" tag: \"9.1\" type: REVERSE\n commit id: \"BD Sécurisée Transaction\" tag: \"3.3\"\n commit id: \"Dépendances\" tag: \"1.3\"\n commit id: \"API GraphQL\" tag: \"4.4\" type: HIGHLIGHT\n commit id: \"ESLint + Outillage\" tag: \"1.4\"\n commit id: \"User Stories\" tag: \"5.1\"\n commit id: \"Producteur Kafka\" tag: \"5.1\" type: HIGHLIGHT\n commit id: \"Émission d'Événements\" tag: \"5.2\"\n commit id: \"Étapes Kafka Terminées\" tag: \"5.0\"\n commit id: \"Procédures AWP\" tag: \"9.2\"\n branch pre-prod\n checkout pre-prod\n commit id: \"Tests Unitaires\" tag: \"8.1\"\n commit id: \"Docker Compose\" tag: \"10.1\"\n commit id: \"Release Sémantique\" tag: \"7.4\" type: HIGHLIGHT\n commit id: \"Winston + Prometheus\" tag: \"10.3\" type: HIGHLIGHT\n commit id: \"Contrôles Santé\" tag: \"10.3\"\n commit id: \"Événements d'Audit\" tag: \"6.1\"\n commit id: \"Actions GitHub\" tag: \"7.4\"\n commit id: \"Docs Observabilité\" tag: \"10.5\"\n commit id: \"Structure Projet\" tag: \"11\"\n checkout main\n merge pre-prod tag: \"v0.9.0\" type: HIGHLIGHT\n\n%%{init: {\n \"gitGraph\": {\n \"titleTopMargin\": 80,\n \"diagramPadding\": 20\n }\n }}%%\n","mermaid","",[109,110,111,119,125,130,136,142,148,154,160,166,172,178,184,190,196,202,208,214,220,226,232,238,244,250,256,262,268,274,280,286,293,299,305,311,317,323],"code",{"__ignoreMap":107},[112,113,116],"span",{"class":114,"line":115},"line",1,[112,117,118],{},"---\n",[112,120,122],{"class":114,"line":121},2,[112,123,124],{},"title: Chronologie du Développement de Bankly - 24-25 Juin 2025 @michaelwybraniec\n",[112,126,128],{"class":114,"line":127},3,[112,129,118],{},[112,131,133],{"class":114,"line":132},4,[112,134,135],{},"gitGraph\n",[112,137,139],{"class":114,"line":138},5,[112,140,141],{}," commit id: \"Diagrammes ER\" tag: \"3.1.2\" type: HIGHLIGHT\n",[112,143,145],{"class":114,"line":144},6,[112,146,147],{}," commit id: \"Modèles Prisma\" tag: \"3.1.1\"\n",[112,149,151],{"class":114,"line":150},7,[112,152,153],{}," commit id: \"Politique AWP → Protocole\" tag: \"9.1\" type: REVERSE\n",[112,155,157],{"class":114,"line":156},8,[112,158,159],{}," commit id: \"BD Sécurisée Transaction\" tag: \"3.3\"\n",[112,161,163],{"class":114,"line":162},9,[112,164,165],{}," commit id: \"Dépendances\" tag: \"1.3\"\n",[112,167,169],{"class":114,"line":168},10,[112,170,171],{}," commit id: \"API GraphQL\" tag: \"4.4\" type: HIGHLIGHT\n",[112,173,175],{"class":114,"line":174},11,[112,176,177],{}," commit id: \"ESLint + Outillage\" tag: \"1.4\"\n",[112,179,181],{"class":114,"line":180},12,[112,182,183],{}," commit id: \"User Stories\" tag: \"5.1\"\n",[112,185,187],{"class":114,"line":186},13,[112,188,189],{}," commit id: \"Producteur Kafka\" tag: \"5.1\" type: HIGHLIGHT\n",[112,191,193],{"class":114,"line":192},14,[112,194,195],{}," commit id: \"Émission d'Événements\" tag: \"5.2\"\n",[112,197,199],{"class":114,"line":198},15,[112,200,201],{}," commit id: \"Étapes Kafka Terminées\" tag: \"5.0\"\n",[112,203,205],{"class":114,"line":204},16,[112,206,207],{}," commit id: \"Procédures AWP\" tag: \"9.2\"\n",[112,209,211],{"class":114,"line":210},17,[112,212,213],{}," branch pre-prod\n",[112,215,217],{"class":114,"line":216},18,[112,218,219],{}," checkout pre-prod\n",[112,221,223],{"class":114,"line":222},19,[112,224,225],{}," commit id: \"Tests Unitaires\" tag: \"8.1\"\n",[112,227,229],{"class":114,"line":228},20,[112,230,231],{}," commit id: \"Docker Compose\" tag: \"10.1\"\n",[112,233,235],{"class":114,"line":234},21,[112,236,237],{}," commit id: \"Release Sémantique\" tag: \"7.4\" type: HIGHLIGHT\n",[112,239,241],{"class":114,"line":240},22,[112,242,243],{}," commit id: \"Winston + Prometheus\" tag: \"10.3\" type: HIGHLIGHT\n",[112,245,247],{"class":114,"line":246},23,[112,248,249],{}," commit id: \"Contrôles Santé\" tag: \"10.3\"\n",[112,251,253],{"class":114,"line":252},24,[112,254,255],{}," commit id: \"Événements d'Audit\" tag: \"6.1\"\n",[112,257,259],{"class":114,"line":258},25,[112,260,261],{}," commit id: \"Actions GitHub\" tag: \"7.4\"\n",[112,263,265],{"class":114,"line":264},26,[112,266,267],{}," commit id: \"Docs Observabilité\" tag: \"10.5\"\n",[112,269,271],{"class":114,"line":270},27,[112,272,273],{}," commit id: \"Structure Projet\" tag: \"11\"\n",[112,275,277],{"class":114,"line":276},28,[112,278,279],{}," checkout main\n",[112,281,283],{"class":114,"line":282},29,[112,284,285],{}," merge pre-prod tag: \"v0.9.0\" type: HIGHLIGHT\n",[112,287,289],{"class":114,"line":288},30,[112,290,292],{"emptyLinePlaceholder":291},true,"\n",[112,294,296],{"class":114,"line":295},31,[112,297,298],{},"%%{init: {\n",[112,300,302],{"class":114,"line":301},32,[112,303,304],{}," \"gitGraph\": {\n",[112,306,308],{"class":114,"line":307},33,[112,309,310],{}," \"titleTopMargin\": 80,\n",[112,312,314],{"class":114,"line":313},34,[112,315,316],{}," \"diagramPadding\": 20\n",[112,318,320],{"class":114,"line":319},35,[112,321,322],{}," }\n",[112,324,326],{"class":114,"line":325},36,[112,327,328],{}," }}%%\n",[68,330,335,341],{"className":331},[332,333,334],"flex","justify-between","items-center",[89,336,338],{"id":337},"vibe-coding",[76,339,340],{},"Vibe-Coding",[73,342,343],{},[344,345,347],"a",{"href":346},"#top",[76,348,349],{},"⤴",[73,351,352],{},"Comment a-t-il été construit ? En utilisant le Vibe-Coding avec la Stratégie de Flux de Travail Agentique.",[68,354,356,362,378],{"className":355},[71],[73,357,358,361],{},[76,359,360],{},"Le Vibecoding avec Flux de Travail Agentique"," signifie :",[363,364,365,369,372,375],"ul",{},[366,367,368],"li",{},"Travailler dans un état de flux de haute créativité et de rapidité",[366,370,371],{},"Laisser les co-pilotes IA gérer les tâches dans un cadre structuré",[366,373,374],{},"Éviter l'épuisement cognitif ou le \"over-vibing\" grâce à des limites systématiques",[366,376,377],{},"Maintenir la clarté et le contexte tout au long du processus de développement",[73,379,380],{},"Le marché du travail évolue, et les ingénieurs doivent s'adapter. L'IA n'est plus optionnelle—c'est un multiplicateur. Mais l'ingénierie à grande échelle avec l'IA nécessite maturité, structure et conscience des risques.",[68,382,384,390],{"className":383},[332,333,334],[89,385,387],{"id":386},"over-vibing",[76,388,389],{},"Over-Vibing",[73,391,392],{},[344,393,394],{"href":346},[76,395,349],{},[68,397,399],{"className":398},[71],[73,400,401,402,405],{},"J'appelle ce risque ",[76,403,404],{},"\"over-vibing\"","—un état où l'on pousse trop loin, trop vite, et perd le contexte. Les outils d'IA (en particulier les LLM) souffrent de la dérive du contexte. Les humains aussi. Lorsque vous êtes dans le flux mais perdez de vue l'image globale, la productivité peut rapidement se transformer en dette technique. Pour gérer ce risque, j'ai créé un \"AWP\" qui maintient la structure tout en préservant l'élan créatif.",[68,407,409,415],{"className":408},[332,333,334],[89,410,412],{"id":411},"awp-protocole-de-flux-de-travail-agentique",[76,413,414],{},"AWP : Protocole de Flux de Travail Agentique",[73,416,417],{},[344,418,419],{"href":346},[76,420,349],{},[68,422,424,443],{"className":423},[71],[73,425,426,427,430,431,434,435,438,439,442],{},"Le secret derrière le développement rapide de Bankly réside dans une méthodologie que j'appelle ",[76,428,429],{},"\"Vibecoding avec Flux de Travail Agentique\"",". Cette approche combine des états de flux de haute créativité avec une collaboration IA structurée, évitant les pièges de l'automatisation incontrôlée ou du \"over-vibing\". ",[76,432,433],{},"La preuve est dans les commits"," : Bankly a été construit les ",[76,436,437],{},"24-25 juin 2025"," en deux jours avec ",[76,440,441],{},"plus de 35 commits structurés"," suivant le Protocole de Flux de Travail Agentique (AWP). Chaque commit représente une étape délibérée dans le processus de développement, de la configuration initiale aux fonctionnalités de pré-production.",[102,444,446],{"className":104,"code":445,"language":106,"meta":107,"style":107},"---\ntitle: AWP - Protocole de Flux de Travail Agentique\n---\ngraph TD\n subgraph Planification\n A[Définir les Tâches] --> B[Itérer les Tâches]\n end\n \n subgraph Développement\n C[Revoir la Portée] --> D[Développement]\n D --> E[Vérifier le Contexte]\n E --> F[Exécuter les Tests]\n end\n \n subgraph Documentation\n G[Documentation] --> H[Vérification Finale]\n end\n \n subgraph Contrôle de Version\n I[Revue Git] --> J[Tâche Suivante]\n end\n \n B --> C\n F --> G\n H --> I\n J --> B\n E --> C\n",[109,447,448,452,457,461,466,471,476,481,486,491,496,501,506,510,514,519,524,528,532,537,542,546,550,555,560,565,570],{"__ignoreMap":107},[112,449,450],{"class":114,"line":115},[112,451,118],{},[112,453,454],{"class":114,"line":121},[112,455,456],{},"title: AWP - Protocole de Flux de Travail Agentique\n",[112,458,459],{"class":114,"line":127},[112,460,118],{},[112,462,463],{"class":114,"line":132},[112,464,465],{},"graph TD\n",[112,467,468],{"class":114,"line":138},[112,469,470],{}," subgraph Planification\n",[112,472,473],{"class":114,"line":144},[112,474,475],{}," A[Définir les Tâches] --> B[Itérer les Tâches]\n",[112,477,478],{"class":114,"line":150},[112,479,480],{}," end\n",[112,482,483],{"class":114,"line":156},[112,484,485],{}," \n",[112,487,488],{"class":114,"line":162},[112,489,490],{}," subgraph Développement\n",[112,492,493],{"class":114,"line":168},[112,494,495],{}," C[Revoir la Portée] --> D[Développement]\n",[112,497,498],{"class":114,"line":174},[112,499,500],{}," D --> E[Vérifier le Contexte]\n",[112,502,503],{"class":114,"line":180},[112,504,505],{}," E --> F[Exécuter les Tests]\n",[112,507,508],{"class":114,"line":186},[112,509,480],{},[112,511,512],{"class":114,"line":192},[112,513,485],{},[112,515,516],{"class":114,"line":198},[112,517,518],{}," subgraph Documentation\n",[112,520,521],{"class":114,"line":204},[112,522,523],{}," G[Documentation] --> H[Vérification Finale]\n",[112,525,526],{"class":114,"line":210},[112,527,480],{},[112,529,530],{"class":114,"line":216},[112,531,485],{},[112,533,534],{"class":114,"line":222},[112,535,536],{}," subgraph Contrôle de Version\n",[112,538,539],{"class":114,"line":228},[112,540,541],{}," I[Revue Git] --> J[Tâche Suivante]\n",[112,543,544],{"class":114,"line":234},[112,545,480],{},[112,547,548],{"class":114,"line":240},[112,549,485],{},[112,551,552],{"class":114,"line":246},[112,553,554],{}," B --> C\n",[112,556,557],{"class":114,"line":252},[112,558,559],{}," F --> G\n",[112,561,562],{"class":114,"line":258},[112,563,564],{}," H --> I\n",[112,566,567],{"class":114,"line":264},[112,568,569],{}," J --> B\n",[112,571,572],{"class":114,"line":270},[112,573,574],{}," E --> C\n",[68,576,578,584],{"className":577},[332,333,334],[89,579,581],{"id":580},"mcp-outils-sldc-agentiques",[76,582,583],{},"MCP : Outils SLDC Agentiques",[73,585,586],{},[344,587,588],{"href":346},[76,589,349],{},[68,591,593,607,612],{"className":592},[71],[73,594,595,598,599,602,603,606],{},[76,596,597],{},"MCP SLDC Agentique"," (Cycle de Développement Logiciel) est un ",[76,600,601],{},"protocole open-source"," conçu pour aider les équipes et les agents IA à maintenir la ",[76,604,605],{},"clarté, la transparence et le flux"," tout au long du développement.",[73,608,609],{},[76,610,611],{},"Avantages Clés :",[363,613,614,617,620,623],{},[366,615,616],{},"Toujours savoir quelle est la prochaine étape",[366,618,619],{},"Restaurer le contexte après les pauses ou les réinitialisations",[366,621,622],{},"Maintenir la documentation et le code synchronisés",[366,624,625],{},"Équilibrer efficacement la collaboration humain-IA",[73,627,628],{},[76,629,630],{},"Considérations Clés en 2025 :",[68,632,634],{"className":633},[71],[363,635,636,642,648],{},[366,637,638,641],{},[76,639,640],{},"VIBE-CODING"," : Flux créatif productif en développement",[366,643,644,647],{},[76,645,646],{},"OVER-VIBING"," : Perte de contrôle du projet due à une automatisation excessive ou à une dérive",[366,649,650,653],{},[76,651,652],{},"Collaboration Agentique"," : Attribution de rôles et de responsabilités structurés aux humains et aux IA",[68,655,657,663],{"className":656},[332,333,334],[89,658,660],{"id":659},"exemples-de-flux-de-travail",[76,661,662],{},"Exemples de Flux de Travail",[73,664,665],{},[344,666,667],{"href":346},[76,668,349],{},[68,670,672],{"className":671},[71],[73,673,674],{},"Voici un diagramme simplifié de la boucle de développement :",[102,676,678],{"className":104,"code":677,"language":106,"meta":107,"style":107},"---\ntitle: SLDC Agentique - Boucle de Développement\n---\nflowchart TD\n Start([\"Début Étape (ex., 4.2)\"])\n Code[\"Implémenter Fonctionnalité/Tâche\"]\n Update[\"mcp:update\u003Cbr/>Mettre à jour README.md & mcp-agentic-sldc.yml\u003Cbr/>Marquer l'étape comme terminée\"]\n Commit[\"mcp:commit\u003Cbr/>Commit avec message Conventional Commit\"]\n Next[\"mcp:next\u003Cbr/>Passer à l'étape suivante\"]\n Review[\"Revoir Progression\u003Cbr/>(Si blocages, les signaler)\"]\n Done([\"Étape Terminée\"])\n\n Start --> Code --> Update --> Commit --> Next --> Review --> Done\n Review -- \"Si pas prêt\" --> Start\n",[109,679,680,684,689,693,698,703,708,713,718,723,728,733,737,742],{"__ignoreMap":107},[112,681,682],{"class":114,"line":115},[112,683,118],{},[112,685,686],{"class":114,"line":121},[112,687,688],{},"title: SLDC Agentique - Boucle de Développement\n",[112,690,691],{"class":114,"line":127},[112,692,118],{},[112,694,695],{"class":114,"line":132},[112,696,697],{},"flowchart TD\n",[112,699,700],{"class":114,"line":138},[112,701,702],{}," Start([\"Début Étape (ex., 4.2)\"])\n",[112,704,705],{"class":114,"line":144},[112,706,707],{}," Code[\"Implémenter Fonctionnalité/Tâche\"]\n",[112,709,710],{"class":114,"line":150},[112,711,712],{}," Update[\"mcp:update\u003Cbr/>Mettre à jour README.md & mcp-agentic-sldc.yml\u003Cbr/>Marquer l'étape comme terminée\"]\n",[112,714,715],{"class":114,"line":156},[112,716,717],{}," Commit[\"mcp:commit\u003Cbr/>Commit avec message Conventional Commit\"]\n",[112,719,720],{"class":114,"line":162},[112,721,722],{}," Next[\"mcp:next\u003Cbr/>Passer à l'étape suivante\"]\n",[112,724,725],{"class":114,"line":168},[112,726,727],{}," Review[\"Revoir Progression\u003Cbr/>(Si blocages, les signaler)\"]\n",[112,729,730],{"class":114,"line":174},[112,731,732],{}," Done([\"Étape Terminée\"])\n",[112,734,735],{"class":114,"line":180},[112,736,292],{"emptyLinePlaceholder":291},[112,738,739],{"class":114,"line":186},[112,740,741],{}," Start --> Code --> Update --> Commit --> Next --> Review --> Done\n",[112,743,744],{"class":114,"line":192},[112,745,746],{}," Review -- \"Si pas prêt\" --> Start\n",[68,748,750],{"className":749},[71],[73,751,752],{},"Et comment retrouver le contexte après une pause :",[102,754,756],{"className":104,"code":755,"language":106,"meta":107,"style":107},"---\ntitle: SLDC Agentique - Retrouver le contexte\n---\nflowchart TD\n Done --> MemoryLoss[\"Perte de Mémoire ?\"]\n MemoryLoss -- \"Oui\" --> Recall --> Start\n MemoryLoss -- \"Non\" --> Start\n Recall[\"Consulter mcp-agentic-sldc.yml\u003Cbr/>Restaurer contexte, voir progression, actions suivantes\"]\n",[109,757,758,762,767,771,775,780,785,790],{"__ignoreMap":107},[112,759,760],{"class":114,"line":115},[112,761,118],{},[112,763,764],{"class":114,"line":121},[112,765,766],{},"title: SLDC Agentique - Retrouver le contexte\n",[112,768,769],{"class":114,"line":127},[112,770,118],{},[112,772,773],{"class":114,"line":132},[112,774,697],{},[112,776,777],{"class":114,"line":138},[112,778,779],{}," Done --> MemoryLoss[\"Perte de Mémoire ?\"]\n",[112,781,782],{"class":114,"line":144},[112,783,784],{}," MemoryLoss -- \"Oui\" --> Recall --> Start\n",[112,786,787],{"class":114,"line":150},[112,788,789],{}," MemoryLoss -- \"Non\" --> Start\n",[112,791,792],{"class":114,"line":156},[112,793,794],{}," Recall[\"Consulter mcp-agentic-sldc.yml\u003Cbr/>Restaurer contexte, voir progression, actions suivantes\"]\n",[68,796,798,802],{"className":797},[332,333,334],[89,799,801],{"id":800},"prochaines-étapes","Prochaines étapes",[73,803,804],{},[344,805,806],{"href":346},[76,807,349],{},[809,810,812],"h3",{"id":811},"mcp","MCP",[73,814,815],{},[76,816,817,818,821],{},"Le Fichier ",[109,819,820],{},"mcp-agentic-sldc.yml"," : Votre Feuille de Route Vivante",[68,823,825],{"className":824},[71],[73,826,827],{},"Comme nous sommes en phase initiale, chaque projet utilisant le SLDC Agentique avec le Protocole de Flux de Travail Agentique devrait inclure un fichier structuré comme cet exemple :",[102,829,833],{"className":830,"code":831,"language":832,"meta":107,"style":107},"language-yaml shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","author: Michael Wybraniec\ngoal: Construire une API backend moderne pour la banque\noverview:\n - Configuration\n - Schéma & Modèles\n - Auth & Permissions\n - Tests & Documentation\noutcome: Une API bancaire prête à être étendue, construite avec GraphQL et Prisma\nsteps:\n - number: 1\n name: Configurer le projet\n owner: human\n done: true\n - number: 2\n name: Définir le schéma GraphQL\n owner: ai\n done: true\n - number: 3\n name: Connecter les modèles Prisma\n owner: ai\n done: true\n - number: 4\n name: Auth & middleware\n owner: human\n done: false\n","yaml",[109,834,835,849,859,867,875,882,889,896,906,913,926,936,946,957,968,977,986,994,1005,1014,1022,1030,1041,1050,1058],{"__ignoreMap":107},[112,836,837,841,845],{"class":114,"line":115},[112,838,840],{"class":839},"swJcz","author",[112,842,844],{"class":843},"sMK4o",":",[112,846,848],{"class":847},"sfazB"," Michael Wybraniec\n",[112,850,851,854,856],{"class":114,"line":121},[112,852,853],{"class":839},"goal",[112,855,844],{"class":843},[112,857,858],{"class":847}," Construire une API backend moderne pour la banque\n",[112,860,861,864],{"class":114,"line":127},[112,862,863],{"class":839},"overview",[112,865,866],{"class":843},":\n",[112,868,869,872],{"class":114,"line":132},[112,870,871],{"class":843}," -",[112,873,874],{"class":847}," Configuration\n",[112,876,877,879],{"class":114,"line":138},[112,878,871],{"class":843},[112,880,881],{"class":847}," Schéma & Modèles\n",[112,883,884,886],{"class":114,"line":144},[112,885,871],{"class":843},[112,887,888],{"class":847}," Auth & Permissions\n",[112,890,891,893],{"class":114,"line":150},[112,892,871],{"class":843},[112,894,895],{"class":847}," Tests & Documentation\n",[112,897,898,901,903],{"class":114,"line":156},[112,899,900],{"class":839},"outcome",[112,902,844],{"class":843},[112,904,905],{"class":847}," Une API bancaire prête à être étendue, construite avec GraphQL et Prisma\n",[112,907,908,911],{"class":114,"line":162},[112,909,910],{"class":839},"steps",[112,912,866],{"class":843},[112,914,915,917,920,922],{"class":114,"line":168},[112,916,871],{"class":843},[112,918,919],{"class":839}," number",[112,921,844],{"class":843},[112,923,925],{"class":924},"sbssI"," 1\n",[112,927,928,931,933],{"class":114,"line":174},[112,929,930],{"class":839}," name",[112,932,844],{"class":843},[112,934,935],{"class":847}," Configurer le projet\n",[112,937,938,941,943],{"class":114,"line":180},[112,939,940],{"class":839}," owner",[112,942,844],{"class":843},[112,944,945],{"class":847}," human\n",[112,947,948,951,953],{"class":114,"line":186},[112,949,950],{"class":839}," done",[112,952,844],{"class":843},[112,954,956],{"class":955},"sfNiH"," true\n",[112,958,959,961,963,965],{"class":114,"line":192},[112,960,871],{"class":843},[112,962,919],{"class":839},[112,964,844],{"class":843},[112,966,967],{"class":924}," 2\n",[112,969,970,972,974],{"class":114,"line":198},[112,971,930],{"class":839},[112,973,844],{"class":843},[112,975,976],{"class":847}," Définir le schéma GraphQL\n",[112,978,979,981,983],{"class":114,"line":204},[112,980,940],{"class":839},[112,982,844],{"class":843},[112,984,985],{"class":847}," ai\n",[112,987,988,990,992],{"class":114,"line":210},[112,989,950],{"class":839},[112,991,844],{"class":843},[112,993,956],{"class":955},[112,995,996,998,1000,1002],{"class":114,"line":216},[112,997,871],{"class":843},[112,999,919],{"class":839},[112,1001,844],{"class":843},[112,1003,1004],{"class":924}," 3\n",[112,1006,1007,1009,1011],{"class":114,"line":222},[112,1008,930],{"class":839},[112,1010,844],{"class":843},[112,1012,1013],{"class":847}," Connecter les modèles Prisma\n",[112,1015,1016,1018,1020],{"class":114,"line":228},[112,1017,940],{"class":839},[112,1019,844],{"class":843},[112,1021,985],{"class":847},[112,1023,1024,1026,1028],{"class":114,"line":234},[112,1025,950],{"class":839},[112,1027,844],{"class":843},[112,1029,956],{"class":955},[112,1031,1032,1034,1036,1038],{"class":114,"line":240},[112,1033,871],{"class":843},[112,1035,919],{"class":839},[112,1037,844],{"class":843},[112,1039,1040],{"class":924}," 4\n",[112,1042,1043,1045,1047],{"class":114,"line":246},[112,1044,930],{"class":839},[112,1046,844],{"class":843},[112,1048,1049],{"class":847}," Auth & middleware\n",[112,1051,1052,1054,1056],{"class":114,"line":252},[112,1053,940],{"class":839},[112,1055,844],{"class":843},[112,1057,945],{"class":847},[112,1059,1060,1062,1064],{"class":114,"line":258},[112,1061,950],{"class":839},[112,1063,844],{"class":843},[112,1065,1066],{"class":955}," false\n",[68,1068,1070],{"className":1069},[71],[73,1071,1072,1073,1076,1077,1076,1080,1083],{},"Le YAML définit également des procédures comme ",[109,1074,1075],{},"update",", ",[109,1078,1079],{},"commit",[109,1081,1082],{},"next",", et des règles pour les transferts humain-IA.",[68,1085,1087,1093],{"className":1086},[332,333,334],[809,1088,1090],{"id":1089},"meilleures-pratiques",[76,1091,1092],{},"Meilleures Pratiques",[73,1094,1095],{},[344,1096,1097],{"href":346},[76,1098,349],{},[68,1100,1102,1105],{"className":1101},[71],[73,1103,1104],{},"Comme le Vibe-Coding est nouveau, ainsi que tous les concepts, vous gardez le contrôle total, que vous le vouliez ou non. Par conséquent, gardez cette liste à l'esprit :",[363,1106,1107,1110,1113,1116,1119,1125,1128],{},[366,1108,1109],{},"Utiliser des commits conventionnels liés à chaque étape",[366,1111,1112],{},"Maintenir README.md et le YAML SLDC toujours à jour",[366,1114,1115],{},"Définir la propriété : Humain vs IA",[366,1117,1118],{},"Permettre un transfert structuré entre agents",[366,1120,1121,1122],{},"Récupérer d'une perte de mémoire avec une seule commande ",[109,1123,1124],{},"check",[366,1126,1127],{},"Maintenir une architecture de code modulaire et testable",[366,1129,1130],{},"Documenter les décisions et compromis au fur et à mesure",[68,1132,1134,1140],{"className":1133},[332,333,334],[809,1135,1137],{"id":1136},"pourquoi-ça-marche",[76,1138,1139],{},"Pourquoi Ça Marche ?",[73,1141,1142],{},[344,1143,1144],{"href":346},[76,1145,349],{},[68,1147,1149,1156,1190],{"className":1148},[71],[73,1150,1151,1152,1155],{},"Bankly a été complété en ",[76,1153,1154],{},"2 jours"," parce que :",[363,1157,1158,1161,1164,1167,1170,1173,1179],{},[366,1159,1160],{},"La feuille de route était claire et structurée via le protocole AWP",[366,1162,1163],{},"Le travail était modulaire et agentique avec une propriété définie",[366,1165,1166],{},"L'IA gérait les tâches structurées comme le schéma, la génération de modèles et la configuration",[366,1168,1169],{},"L'attention humaine restait sur l'architecture, la logique et la revue",[366,1171,1172],{},"Le contexte n'a jamais été perdu grâce au protocole SLDC",[366,1174,1175,1178],{},[76,1176,1177],{},"35+ commits"," montrent une progression systématique sans dérive fonctionnelle",[366,1180,1181,1182,1185,1186,1189],{},"Le ",[76,1183,1184],{},"Jour 1"," s'est concentré sur les fonctionnalités de base, le ",[76,1187,1188],{},"Jour 2"," sur la préparation à la production",[73,1191,1192,1193,1196],{},"Ce n'est pas juste de la productivité—c'est une ",[76,1194,1195],{},"vélocité durable",". Le système prévient l'épuisement tout en maintenant la qualité et en permettant une itération rapide.",[68,1198,1200,1206],{"className":1199},[332,333,334],[89,1201,1203],{"id":1202},"réflexions-finales",[76,1204,1205],{},"Réflexions Finales",[73,1207,1208],{},[344,1209,1210],{"href":346},[76,1211,349],{},[68,1213,1215,1218,1245],{"className":1214},[71],[73,1216,1217],{},"MCP SLDC Agentique ou mon AWP - Protocole de Flux de Travail Agentique - n'est pas juste un outil—c'est une philosophie pour le développement moderne :",[363,1219,1220,1230,1236,1242],{},[366,1221,1222,1223,1226,1227],{},"Construire ",[76,1224,1225],{},"plus rapidement",", mais rester ",[76,1228,1229],{},"ancré",[366,1231,1232,1233],{},"Collaborer avec l'IA, mais maintenir la ",[76,1234,1235],{},"stratégie et la propriété",[366,1237,1238,1239],{},"Utiliser la clarté et le processus pour ",[76,1240,1241],{},"éviter l'épuisement et la confusion",[366,1243,1244],{},"La structure est la nouvelle vitesse dans un monde assisté par l'IA",[73,1246,1247],{},"Que vous fassiez du Vibe-Coding en solo ou que vous gériez une équipe d'agents, la collaboration structurée est la clé d'une livraison de logiciels durable et de haute qualité.",[68,1249,1251,1257],{"className":1250},[332,333,334],[89,1252,1254],{"id":1253},"participez",[76,1255,1256],{},"Participez !",[73,1258,1259],{},[344,1260,1261],{"href":346},[76,1262,349],{},[68,1264,1266],{"className":1265},[71],[73,1267,1268,1269,1271,1272,1278],{},"Contribuez au ",[76,1270,597],{}," et rejoignez le mouvement vers un développement centré sur l'humain et assisté par l'IA -> ",[344,1273,1277],{"href":1274,"rel":1275},"https://one-front.com/contact",[1276],"nofollow","ici",".",[68,1280,1286],{"className":1281},[1282,1283,1284,1285],"text-md","text-center","mt-10","font-bold",[73,1287,1288],{},[344,1289,1290],{"href":346},[76,1291,349],{},[1293,1294,1295],"style",{},"html .light .shiki span {color: var(--shiki-light);background: var(--shiki-light-bg);font-style: var(--shiki-light-font-style);font-weight: var(--shiki-light-font-weight);text-decoration: var(--shiki-light-text-decoration);}html.light .shiki span {color: var(--shiki-light);background: var(--shiki-light-bg);font-style: var(--shiki-light-font-style);font-weight: var(--shiki-light-font-weight);text-decoration: var(--shiki-light-text-decoration);}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html pre.shiki code .swJcz, html code.shiki .swJcz{--shiki-light:#E53935;--shiki-default:#F07178;--shiki-dark:#F07178}html pre.shiki code .sMK4o, html code.shiki .sMK4o{--shiki-light:#39ADB5;--shiki-default:#89DDFF;--shiki-dark:#89DDFF}html pre.shiki code .sfazB, html code.shiki .sfazB{--shiki-light:#91B859;--shiki-default:#C3E88D;--shiki-dark:#C3E88D}html pre.shiki code .sbssI, html code.shiki .sbssI{--shiki-light:#F76D47;--shiki-default:#F78C6C;--shiki-dark:#F78C6C}html pre.shiki code .sfNiH, html code.shiki .sfNiH{--shiki-light:#FF5370;--shiki-default:#FF9CAC;--shiki-dark:#FF9CAC}",{"title":107,"searchDepth":121,"depth":121,"links":1297},[1298,1299,1300,1301,1302,1303,1304,1309,1310],{"id":91,"depth":121,"text":94},{"id":337,"depth":121,"text":340},{"id":386,"depth":121,"text":389},{"id":411,"depth":121,"text":414},{"id":580,"depth":121,"text":583},{"id":659,"depth":121,"text":662},{"id":800,"depth":121,"text":801,"children":1305},[1306,1307,1308],{"id":811,"depth":127,"text":812},{"id":1089,"depth":127,"text":1092},{"id":1136,"depth":127,"text":1139},{"id":1202,"depth":121,"text":1205},{"id":1253,"depth":121,"text":1256},"2025-06-29","Découvrez comment Bankly, un système bancaire backend, a été construit en seulement deux jours en utilisant Apollo GraphQL, Prisma et le Flux de Travail Agentique - un protocole pour le développement durable assisté par l'IA.","md",{"src":1315},"/blog/mcp/mcp-cover.png",{},{"title":31,"description":1312},"YK4sakb9zslMRc8p8ii8F05oAMn1YxZECH7H7RpY94E",[1320,4485],{"id":1321,"title":35,"authors":1322,"badge":1326,"body":1329,"date":4479,"description":4480,"extension":1313,"image":4481,"meta":4482,"navigation":291,"path":36,"seo":4483,"stem":37,"__hash__":4484},"posts_fr/fr/blog/8. mcp-in-chat-demo.md",[1323],{"name":55,"description":1324,"to":57,"avatar":1325},"Freelance, Serveurs MCP, Développement Full-Stack, Architecture",{"src":59},{"label":1327,"color":1328,"variant":63},"MCP, E-commerce, Demo","error",{"type":65,"value":1330,"toc":4429},[1331,1342,1357,1373,1377,1392,1396,1428,1432,1517,1521,1534,1538,1576,1580,2037,2052,2058,2062,2220,2226,2230,2441,2446,2450,2630,2635,2639,2831,2836,2851,2857,2861,2867,2872,2880,2884,2889,2893,2899,2903,2908,2912,2918,2933,2939,2943,3193,3197,3200,3250,3254,3615,3621,3636,3642,3646,3678,3682,3688,3692,3794,3809,3815,3819,3833,3837,3937,3941,3944,3964,3968,3971,3997,4012,4018,4022,4053,4057,4128,4132,4158,4173,4179,4183,4261,4265,4268,4300,4315,4321,4325,4357,4360,4363,4394,4397,4400,4405,4426],[68,1332,1334],{"className":1333},[71],[73,1335,1181,1336,1341],{},[344,1337,1340],{"href":1338,"rel":1339},"https://github.com/michaelwybraniec/mcp-in-chat",[1276],"projet MCP-in-Chat"," représente une démonstration complète et prête pour la production de comment les serveurs du Protocole de Contexte de Modèle (MCP) peuvent transformer les interactions de chatbots e-commerce. Cet exemple complet montre un système de maintenance et de vente de chaudières qui connecte des chatbots IA externes (Claude, ChatGPT, Cursor) à la logique métier réelle à travers une architecture sécurisée et évolutive. Construit en utilisant la méthodologie du Protocole de Flux de Travail Agentique (AWP), ce projet démontre comment créer des serveurs MCP sophistiqués qui connectent l'IA conversationnelle aux systèmes d'entreprise.",[68,1343,1345,1351],{"className":1344},[332,333,334],[89,1346,1348],{"id":1347},"_1-awp-protocole-de-flux-de-travail-agentique",[76,1349,1350],{},"1. AWP - Protocole de Flux de Travail Agentique",[73,1352,1353],{},[344,1354,1355],{"href":346},[76,1356,349],{},[68,1358,1360],{"className":1359},[71],[73,1361,1181,1362,1365,1366,1368,1369,1372],{},[76,1363,1364],{},"Protocole de Flux de Travail Agentique (AWP)"," est une méthodologie de développement agentique créée par ",[76,1367,55],{}," et maintenue par la ",[76,1370,1371],{},"communauté overvibing.com",". Cette approche systématique assure des résultats de développement cohérents et de haute qualité en fournissant des flux de travail structurés qui préviennent la dérive de contexte appelée overvibing et permettent une livraison rapide et fiable de projets tout en évoluant le développement avec l'IA.",[809,1374,1376],{"id":1375},"vidéo-demo","Vidéo Demo",[68,1378,1380],{"className":1379},[1283],[68,1381,1383],{"style":1382},"padding:95.12% 0 0 0;position:relative;",[1384,1385],"iframe",{"src":1386,"frameBorder":1387,"allow":1388,"referrerPolicy":1389,"style":1390,"title":1391},"https://player.vimeo.com/video/1106052609?badge=0&autopause=0&player_id=0&app_id=58479","0","autoplay; fullscreen; picture-in-picture; clipboard-write; encrypted-media; web-share","strict-origin-when-cross-origin","position:absolute;top:0;left:0;width:100%;height:100%;","Agentic Workflow preventing Overvibing.com - #GreenTokens #MCP #VibeCoding",[809,1393,1395],{"id":1394},"résultats-dimplémentation","Résultats d'Implémentation",[363,1397,1398,1404,1410,1416,1422],{},[366,1399,1400,1403],{},[76,1401,1402],{},"Calendrier",": 4 heures (incluant les tâches non planifiées)",[366,1405,1406,1409],{},[76,1407,1408],{},"Portée",": 100% des fonctionnalités planifiées implémentées",[366,1411,1412,1415],{},[76,1413,1414],{},"Qualité",": 100% de succès aux tests",[366,1417,1418,1421],{},[76,1419,1420],{},"Documentation",": Complète et compréhensive",[366,1423,1424,1427],{},[76,1425,1426],{},"Prêt pour Demo",": Entièrement fonctionnel et testé",[809,1429,1431],{"id":1430},"phases-de-développement","Phases de Développement",[1433,1434,1435,1455,1478,1497],"ol",{},[366,1436,1437,1440,1441],{},[76,1438,1439],{},"Configuration du Projet et Architecture de Sécurité"," (45 minutes)",[363,1442,1443,1446,1449,1452],{},[366,1444,1445],{},"Initialiser le projet Node.js avec TypeScript",[366,1447,1448],{},"Configurer l'architecture en couches (MCP → API → Services → Base de Données)",[366,1450,1451],{},"Créer des fichiers JSON de données simulées avec contrôles d'accès",[366,1453,1454],{},"Définir les interfaces TypeScript et types de sécurité",[366,1456,1457,1460,1461],{},[76,1458,1459],{},"Développement des Outils MCP Principaux"," (90 minutes)",[363,1462,1463,1466,1469,1472,1475],{},[366,1464,1465],{},"Implémenter l'outil d'information de chaudières avec validation d'API",[366,1467,1468],{},"Implémenter l'outil de planification de maintenance avec contrôles d'accès",[366,1470,1471],{},"Implémenter l'outil d'inventaire avec limitation de débit",[366,1473,1474],{},"Implémenter l'outil d'achat avec validation de transactions",[366,1476,1477],{},"Implémenter l'outil email avec filtrage de contenu",[366,1479,1480,1440,1483],{},[76,1481,1482],{},"Tests et Préparation de Demo",[363,1484,1485,1488,1491,1494],{},[366,1486,1487],{},"Tester le Serveur MCP avec des outils locaux",[366,1489,1490],{},"Créer des instructions de demo pour l'intégration de chatbot",[366,1492,1493],{},"Préparer des flux de conversation d'exemple",[366,1495,1496],{},"Créer README avec instructions de configuration et d'utilisation",[366,1498,1499,1502,1503],{},[76,1500,1501],{},"Intégration de Chatbot Externe"," (60 minutes)",[363,1504,1505,1508,1511,1514],{},[366,1506,1507],{},"Tester le Serveur MCP avec Claude (si disponible)",[366,1509,1510],{},"Tester le Serveur MCP avec ChatGPT (si disponible)",[366,1512,1513],{},"Documenter le processus d'intégration et les résultats",[366,1515,1516],{},"Créer des matériaux de présentation de demo",[809,1518,1520],{"id":1519},"aperçu-du-projet","Aperçu du Projet",[68,1522,1524,1527],{"className":1523},[71],[73,1525,1526],{},"Serveur MCP de Maintenance de Chaudières",[73,1528,1529,1530,1533],{},"Le projet MCP-in-Chat est une ",[76,1531,1532],{},"démonstration complétée avec succès"," qui montre comment les serveurs MCP peuvent s'intégrer avec des chatbots externes pour fournir une valeur commerciale réelle. Le système gère les services de maintenance de chaudières, les informations client, la gestion d'inventaire et le traitement des achats à travers une interface conversationnelle.",[809,1535,1537],{"id":1536},"fonctionnalités-clés","Fonctionnalités Clés",[363,1539,1540,1546,1552,1558,1564,1570],{},[366,1541,1542,1545],{},[76,1543,1544],{},"4 Outils MCP",": Information de chaudières, planification de maintenance, traitement d'achat et confirmation par email",[366,1547,1548,1551],{},[76,1549,1550],{},"4 Endpoints API",": Backend RESTful avec authentification et validation",[366,1553,1554,1557],{},[76,1555,1556],{},"8 Services Simulés",": Simulation complète de logique métier avec données réalistes",[366,1559,1560,1563],{},[76,1561,1562],{},"Intégration Externe",": Prêt pour les tests avec Claude, ChatGPT et Cursor IDE",[366,1565,1566,1569],{},[76,1567,1568],{},"100% Couverture de Tests",": Suite de tests complète avec 100% de succès",[366,1571,1572,1575],{},[76,1573,1574],{},"Prêt pour Production",": Middleware de sécurité, limitation de débit et gestion d'erreurs",[809,1577,1579],{"id":1578},"aperçu-de-larchitecture","Aperçu de l'Architecture",[102,1581,1583],{"className":104,"code":1582,"language":106,"meta":107,"style":107},"---\nheader: MCP-in-Chat System Architecture\nlegend:\n - color: \"#3B82F6\"\n text: \"External Systems\"\n - color: \"#8B5CF6\"\n text: \"MCP Protocol Layer\"\n - color: \"#10B981\"\n text: \"API & Security Layer\"\n - color: \"#F59E0B\"\n text: \"Business Services\"\n - color: \"#EF4444\"\n text: \"Data Layer\"\n---\ngraph TD\n subgraph \"External Chatbots\"\n A[Claude AI]\n B[ChatGPT]\n C[Cursor IDE]\n end\n \n subgraph \"MCP Server Layer\"\n D[MCP Server]\n E[Boiler Info Tool]\n F[Maintenance Tool]\n G[Purchase Tool]\n H[Email Tool]\n end\n \n subgraph \"Backend API Layer\"\n I[Express API Server]\n J[Authentication]\n K[Validation]\n L[Rate Limiting]\n end\n \n subgraph \"Business Services\"\n M[Customer Service]\n N[Boiler Service]\n O[Payment Service]\n P[Email Service]\n end\n \n subgraph \"Mock Data Layer\"\n Q[Customers JSON]\n R[Boilers JSON]\n S[Inventory JSON]\n T[Orders JSON]\n end\n \n A --> D\n B --> D\n C --> D\n D --> E\n D --> F\n D --> G\n D --> H\n E --> I\n F --> I\n G --> I\n H --> I\n I --> J\n J --> K\n K --> L\n L --> M\n L --> N\n L --> O\n L --> P\n M --> Q\n N --> R\n O --> S\n P --> T\n \n classDef external fill:#3B82F6,stroke:#1E40AF,stroke-width:2px,color:#fff\n classDef mcp fill:#8B5CF6,stroke:#7C3AED,stroke-width:2px,color:#fff\n classDef api fill:#10B981,stroke:#059669,stroke-width:2px,color:#fff\n classDef service fill:#F59E0B,stroke:#D97706,stroke-width:2px,color:#fff\n classDef data fill:#EF4444,stroke:#DC2626,stroke-width:2px,color:#fff\n \n class A,B,C external\n class D,E,F,G,H mcp\n class I,J,K,L api\n class M,N,O,P service\n class Q,R,S,T data\n",[109,1584,1585,1589,1594,1599,1604,1609,1614,1619,1624,1629,1634,1639,1644,1649,1653,1657,1662,1667,1672,1677,1681,1685,1690,1695,1700,1705,1710,1715,1719,1723,1728,1733,1738,1743,1748,1752,1756,1762,1768,1774,1780,1786,1791,1796,1802,1808,1814,1820,1826,1831,1836,1842,1848,1854,1860,1866,1872,1878,1884,1890,1896,1901,1907,1913,1919,1925,1931,1937,1943,1949,1955,1961,1967,1972,1978,1984,1990,1996,2002,2007,2013,2019,2025,2031],{"__ignoreMap":107},[112,1586,1587],{"class":114,"line":115},[112,1588,118],{},[112,1590,1591],{"class":114,"line":121},[112,1592,1593],{},"header: MCP-in-Chat System Architecture\n",[112,1595,1596],{"class":114,"line":127},[112,1597,1598],{},"legend:\n",[112,1600,1601],{"class":114,"line":132},[112,1602,1603],{}," - color: \"#3B82F6\"\n",[112,1605,1606],{"class":114,"line":138},[112,1607,1608],{}," text: \"External Systems\"\n",[112,1610,1611],{"class":114,"line":144},[112,1612,1613],{}," - color: \"#8B5CF6\"\n",[112,1615,1616],{"class":114,"line":150},[112,1617,1618],{}," text: \"MCP Protocol Layer\"\n",[112,1620,1621],{"class":114,"line":156},[112,1622,1623],{}," - color: \"#10B981\"\n",[112,1625,1626],{"class":114,"line":162},[112,1627,1628],{}," text: \"API & Security Layer\"\n",[112,1630,1631],{"class":114,"line":168},[112,1632,1633],{}," - color: \"#F59E0B\"\n",[112,1635,1636],{"class":114,"line":174},[112,1637,1638],{}," text: \"Business Services\"\n",[112,1640,1641],{"class":114,"line":180},[112,1642,1643],{}," - color: \"#EF4444\"\n",[112,1645,1646],{"class":114,"line":186},[112,1647,1648],{}," text: \"Data Layer\"\n",[112,1650,1651],{"class":114,"line":192},[112,1652,118],{},[112,1654,1655],{"class":114,"line":198},[112,1656,465],{},[112,1658,1659],{"class":114,"line":204},[112,1660,1661],{}," subgraph \"External Chatbots\"\n",[112,1663,1664],{"class":114,"line":210},[112,1665,1666],{}," A[Claude AI]\n",[112,1668,1669],{"class":114,"line":216},[112,1670,1671],{}," B[ChatGPT]\n",[112,1673,1674],{"class":114,"line":222},[112,1675,1676],{}," C[Cursor IDE]\n",[112,1678,1679],{"class":114,"line":228},[112,1680,480],{},[112,1682,1683],{"class":114,"line":234},[112,1684,485],{},[112,1686,1687],{"class":114,"line":240},[112,1688,1689],{}," subgraph \"MCP Server Layer\"\n",[112,1691,1692],{"class":114,"line":246},[112,1693,1694],{}," D[MCP Server]\n",[112,1696,1697],{"class":114,"line":252},[112,1698,1699],{}," E[Boiler Info Tool]\n",[112,1701,1702],{"class":114,"line":258},[112,1703,1704],{}," F[Maintenance Tool]\n",[112,1706,1707],{"class":114,"line":264},[112,1708,1709],{}," G[Purchase Tool]\n",[112,1711,1712],{"class":114,"line":270},[112,1713,1714],{}," H[Email Tool]\n",[112,1716,1717],{"class":114,"line":276},[112,1718,480],{},[112,1720,1721],{"class":114,"line":282},[112,1722,485],{},[112,1724,1725],{"class":114,"line":288},[112,1726,1727],{}," subgraph \"Backend API Layer\"\n",[112,1729,1730],{"class":114,"line":295},[112,1731,1732],{}," I[Express API Server]\n",[112,1734,1735],{"class":114,"line":301},[112,1736,1737],{}," J[Authentication]\n",[112,1739,1740],{"class":114,"line":307},[112,1741,1742],{}," K[Validation]\n",[112,1744,1745],{"class":114,"line":313},[112,1746,1747],{}," L[Rate Limiting]\n",[112,1749,1750],{"class":114,"line":319},[112,1751,480],{},[112,1753,1754],{"class":114,"line":325},[112,1755,485],{},[112,1757,1759],{"class":114,"line":1758},37,[112,1760,1761],{}," subgraph \"Business Services\"\n",[112,1763,1765],{"class":114,"line":1764},38,[112,1766,1767],{}," M[Customer Service]\n",[112,1769,1771],{"class":114,"line":1770},39,[112,1772,1773],{}," N[Boiler Service]\n",[112,1775,1777],{"class":114,"line":1776},40,[112,1778,1779],{}," O[Payment Service]\n",[112,1781,1783],{"class":114,"line":1782},41,[112,1784,1785],{}," P[Email Service]\n",[112,1787,1789],{"class":114,"line":1788},42,[112,1790,480],{},[112,1792,1794],{"class":114,"line":1793},43,[112,1795,485],{},[112,1797,1799],{"class":114,"line":1798},44,[112,1800,1801],{}," subgraph \"Mock Data Layer\"\n",[112,1803,1805],{"class":114,"line":1804},45,[112,1806,1807],{}," Q[Customers JSON]\n",[112,1809,1811],{"class":114,"line":1810},46,[112,1812,1813],{}," R[Boilers JSON]\n",[112,1815,1817],{"class":114,"line":1816},47,[112,1818,1819],{}," S[Inventory JSON]\n",[112,1821,1823],{"class":114,"line":1822},48,[112,1824,1825],{}," T[Orders JSON]\n",[112,1827,1829],{"class":114,"line":1828},49,[112,1830,480],{},[112,1832,1834],{"class":114,"line":1833},50,[112,1835,485],{},[112,1837,1839],{"class":114,"line":1838},51,[112,1840,1841],{}," A --> D\n",[112,1843,1845],{"class":114,"line":1844},52,[112,1846,1847],{}," B --> D\n",[112,1849,1851],{"class":114,"line":1850},53,[112,1852,1853],{}," C --> D\n",[112,1855,1857],{"class":114,"line":1856},54,[112,1858,1859],{}," D --> E\n",[112,1861,1863],{"class":114,"line":1862},55,[112,1864,1865],{}," D --> F\n",[112,1867,1869],{"class":114,"line":1868},56,[112,1870,1871],{}," D --> G\n",[112,1873,1875],{"class":114,"line":1874},57,[112,1876,1877],{}," D --> H\n",[112,1879,1881],{"class":114,"line":1880},58,[112,1882,1883],{}," E --> I\n",[112,1885,1887],{"class":114,"line":1886},59,[112,1888,1889],{}," F --> I\n",[112,1891,1893],{"class":114,"line":1892},60,[112,1894,1895],{}," G --> I\n",[112,1897,1899],{"class":114,"line":1898},61,[112,1900,564],{},[112,1902,1904],{"class":114,"line":1903},62,[112,1905,1906],{}," I --> J\n",[112,1908,1910],{"class":114,"line":1909},63,[112,1911,1912],{}," J --> K\n",[112,1914,1916],{"class":114,"line":1915},64,[112,1917,1918],{}," K --> L\n",[112,1920,1922],{"class":114,"line":1921},65,[112,1923,1924],{}," L --> M\n",[112,1926,1928],{"class":114,"line":1927},66,[112,1929,1930],{}," L --> N\n",[112,1932,1934],{"class":114,"line":1933},67,[112,1935,1936],{}," L --> O\n",[112,1938,1940],{"class":114,"line":1939},68,[112,1941,1942],{}," L --> P\n",[112,1944,1946],{"class":114,"line":1945},69,[112,1947,1948],{}," M --> Q\n",[112,1950,1952],{"class":114,"line":1951},70,[112,1953,1954],{}," N --> R\n",[112,1956,1958],{"class":114,"line":1957},71,[112,1959,1960],{}," O --> S\n",[112,1962,1964],{"class":114,"line":1963},72,[112,1965,1966],{}," P --> T\n",[112,1968,1970],{"class":114,"line":1969},73,[112,1971,485],{},[112,1973,1975],{"class":114,"line":1974},74,[112,1976,1977],{}," classDef external fill:#3B82F6,stroke:#1E40AF,stroke-width:2px,color:#fff\n",[112,1979,1981],{"class":114,"line":1980},75,[112,1982,1983],{}," classDef mcp fill:#8B5CF6,stroke:#7C3AED,stroke-width:2px,color:#fff\n",[112,1985,1987],{"class":114,"line":1986},76,[112,1988,1989],{}," classDef api fill:#10B981,stroke:#059669,stroke-width:2px,color:#fff\n",[112,1991,1993],{"class":114,"line":1992},77,[112,1994,1995],{}," classDef service fill:#F59E0B,stroke:#D97706,stroke-width:2px,color:#fff\n",[112,1997,1999],{"class":114,"line":1998},78,[112,2000,2001],{}," classDef data fill:#EF4444,stroke:#DC2626,stroke-width:2px,color:#fff\n",[112,2003,2005],{"class":114,"line":2004},79,[112,2006,485],{},[112,2008,2010],{"class":114,"line":2009},80,[112,2011,2012],{}," class A,B,C external\n",[112,2014,2016],{"class":114,"line":2015},81,[112,2017,2018],{}," class D,E,F,G,H mcp\n",[112,2020,2022],{"class":114,"line":2021},82,[112,2023,2024],{}," class I,J,K,L api\n",[112,2026,2028],{"class":114,"line":2027},83,[112,2029,2030],{}," class M,N,O,P service\n",[112,2032,2034],{"class":114,"line":2033},84,[112,2035,2036],{}," class Q,R,S,T data\n",[68,2038,2040,2046],{"className":2039},[332,333,334],[89,2041,2043],{"id":2042},"_2-mcp-protocole-de-contexte-de-modèle",[76,2044,2045],{},"2. MCP - Protocole de Contexte de Modèle",[73,2047,2048],{},[344,2049,2050],{"href":346},[76,2051,349],{},[68,2053,2055],{"className":2054},[71],[73,2056,2057],{},"Le serveur MCP expose quatre outils spécialisés qui permettent aux chatbots d'interagir avec le système de maintenance de chaudières. Chaque outil est conçu avec une logique métier spécifique et des contrôles de sécurité.",[809,2059,2061],{"id":2060},"outil-dinformation-de-chaudières","Outil d'Information de Chaudières",[102,2063,2067],{"className":2064,"code":2065,"language":2066,"meta":107,"style":107},"language-typescript shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","// Outil MCP: Obtenir les détails de chaudières du client et informations de garantie\nconst boilerInfoTool = {\n name: \"boiler-info\",\n description: \"Obtenir les informations de chaudières du client et détails de garantie\",\n inputSchema: {\n type: \"object\",\n properties: {\n customer_id: { type: \"string\" }\n },\n required: [\"customer_id\"]\n }\n}\n","typescript",[109,2068,2069,2075,2091,2110,2126,2135,2151,2160,2185,2190,2210,2215],{"__ignoreMap":107},[112,2070,2071],{"class":114,"line":115},[112,2072,2074],{"class":2073},"sHwdD","// Outil MCP: Obtenir les détails de chaudières du client et informations de garantie\n",[112,2076,2077,2081,2085,2088],{"class":114,"line":121},[112,2078,2080],{"class":2079},"spNyl","const",[112,2082,2084],{"class":2083},"sTEyZ"," boilerInfoTool ",[112,2086,2087],{"class":843},"=",[112,2089,2090],{"class":843}," {\n",[112,2092,2093,2096,2098,2101,2104,2107],{"class":114,"line":127},[112,2094,2095],{"class":839}," name",[112,2097,844],{"class":843},[112,2099,2100],{"class":843}," \"",[112,2102,2103],{"class":847},"boiler-info",[112,2105,2106],{"class":843},"\"",[112,2108,2109],{"class":843},",\n",[112,2111,2112,2115,2117,2119,2122,2124],{"class":114,"line":132},[112,2113,2114],{"class":839}," description",[112,2116,844],{"class":843},[112,2118,2100],{"class":843},[112,2120,2121],{"class":847},"Obtenir les informations de chaudières du client et détails de garantie",[112,2123,2106],{"class":843},[112,2125,2109],{"class":843},[112,2127,2128,2131,2133],{"class":114,"line":138},[112,2129,2130],{"class":839}," inputSchema",[112,2132,844],{"class":843},[112,2134,2090],{"class":843},[112,2136,2137,2140,2142,2144,2147,2149],{"class":114,"line":144},[112,2138,2139],{"class":839}," type",[112,2141,844],{"class":843},[112,2143,2100],{"class":843},[112,2145,2146],{"class":847},"object",[112,2148,2106],{"class":843},[112,2150,2109],{"class":843},[112,2152,2153,2156,2158],{"class":114,"line":150},[112,2154,2155],{"class":839}," properties",[112,2157,844],{"class":843},[112,2159,2090],{"class":843},[112,2161,2162,2165,2167,2170,2173,2175,2177,2180,2182],{"class":114,"line":156},[112,2163,2164],{"class":839}," customer_id",[112,2166,844],{"class":843},[112,2168,2169],{"class":843}," {",[112,2171,2172],{"class":839}," type",[112,2174,844],{"class":843},[112,2176,2100],{"class":843},[112,2178,2179],{"class":847},"string",[112,2181,2106],{"class":843},[112,2183,2184],{"class":843}," }\n",[112,2186,2187],{"class":114,"line":162},[112,2188,2189],{"class":843}," },\n",[112,2191,2192,2195,2197,2200,2202,2205,2207],{"class":114,"line":168},[112,2193,2194],{"class":839}," required",[112,2196,844],{"class":843},[112,2198,2199],{"class":2083}," [",[112,2201,2106],{"class":843},[112,2203,2204],{"class":847},"customer_id",[112,2206,2106],{"class":843},[112,2208,2209],{"class":2083},"]\n",[112,2211,2212],{"class":114,"line":174},[112,2213,2214],{"class":843}," }\n",[112,2216,2217],{"class":114,"line":180},[112,2218,2219],{"class":843},"}\n",[73,2221,2222,2225],{},[76,2223,2224],{},"Cas d'Usage",": Quand un client demande \"Quel est le statut de garantie de ma chaudière?\", le chatbot peut récupérer des informations détaillées sur son modèle de chaudière spécifique, date d'installation, couverture de garantie et historique de maintenance.",[809,2227,2229],{"id":2228},"outil-de-planification-de-maintenance","Outil de Planification de Maintenance",[102,2231,2233],{"className":2064,"code":2232,"language":2066,"meta":107,"style":107},"// Outil MCP: Planifier les services de maintenance avec considération météo\nconst maintenanceTool = {\n name: \"maintenance\",\n description: \"Planifier les services de maintenance\",\n inputSchema: {\n type: \"object\",\n properties: {\n customer_id: { type: \"string\" },\n service_type: { type: \"string\", enum: [\"routine\", \"emergency\"] },\n preferred_date: { type: \"string\" }\n },\n required: [\"customer_id\", \"service_type\"]\n }\n}\n",[109,2234,2235,2240,2251,2266,2281,2289,2303,2311,2332,2383,2404,2408,2433,2437],{"__ignoreMap":107},[112,2236,2237],{"class":114,"line":115},[112,2238,2239],{"class":2073},"// Outil MCP: Planifier les services de maintenance avec considération météo\n",[112,2241,2242,2244,2247,2249],{"class":114,"line":121},[112,2243,2080],{"class":2079},[112,2245,2246],{"class":2083}," maintenanceTool ",[112,2248,2087],{"class":843},[112,2250,2090],{"class":843},[112,2252,2253,2255,2257,2259,2262,2264],{"class":114,"line":127},[112,2254,2095],{"class":839},[112,2256,844],{"class":843},[112,2258,2100],{"class":843},[112,2260,2261],{"class":847},"maintenance",[112,2263,2106],{"class":843},[112,2265,2109],{"class":843},[112,2267,2268,2270,2272,2274,2277,2279],{"class":114,"line":132},[112,2269,2114],{"class":839},[112,2271,844],{"class":843},[112,2273,2100],{"class":843},[112,2275,2276],{"class":847},"Planifier les services de maintenance",[112,2278,2106],{"class":843},[112,2280,2109],{"class":843},[112,2282,2283,2285,2287],{"class":114,"line":138},[112,2284,2130],{"class":839},[112,2286,844],{"class":843},[112,2288,2090],{"class":843},[112,2290,2291,2293,2295,2297,2299,2301],{"class":114,"line":144},[112,2292,2139],{"class":839},[112,2294,844],{"class":843},[112,2296,2100],{"class":843},[112,2298,2146],{"class":847},[112,2300,2106],{"class":843},[112,2302,2109],{"class":843},[112,2304,2305,2307,2309],{"class":114,"line":150},[112,2306,2155],{"class":839},[112,2308,844],{"class":843},[112,2310,2090],{"class":843},[112,2312,2313,2315,2317,2319,2321,2323,2325,2327,2329],{"class":114,"line":156},[112,2314,2164],{"class":839},[112,2316,844],{"class":843},[112,2318,2169],{"class":843},[112,2320,2172],{"class":839},[112,2322,844],{"class":843},[112,2324,2100],{"class":843},[112,2326,2179],{"class":847},[112,2328,2106],{"class":843},[112,2330,2331],{"class":843}," },\n",[112,2333,2334,2337,2339,2341,2343,2345,2347,2349,2351,2354,2357,2359,2361,2363,2366,2368,2370,2372,2375,2377,2380],{"class":114,"line":162},[112,2335,2336],{"class":839}," service_type",[112,2338,844],{"class":843},[112,2340,2169],{"class":843},[112,2342,2172],{"class":839},[112,2344,844],{"class":843},[112,2346,2100],{"class":843},[112,2348,2179],{"class":847},[112,2350,2106],{"class":843},[112,2352,2353],{"class":843},",",[112,2355,2356],{"class":839}," enum",[112,2358,844],{"class":843},[112,2360,2199],{"class":2083},[112,2362,2106],{"class":843},[112,2364,2365],{"class":847},"routine",[112,2367,2106],{"class":843},[112,2369,2353],{"class":843},[112,2371,2100],{"class":843},[112,2373,2374],{"class":847},"emergency",[112,2376,2106],{"class":843},[112,2378,2379],{"class":2083},"] ",[112,2381,2382],{"class":843},"},\n",[112,2384,2385,2388,2390,2392,2394,2396,2398,2400,2402],{"class":114,"line":168},[112,2386,2387],{"class":839}," preferred_date",[112,2389,844],{"class":843},[112,2391,2169],{"class":843},[112,2393,2172],{"class":839},[112,2395,844],{"class":843},[112,2397,2100],{"class":843},[112,2399,2179],{"class":847},[112,2401,2106],{"class":843},[112,2403,2184],{"class":843},[112,2405,2406],{"class":114,"line":174},[112,2407,2189],{"class":843},[112,2409,2410,2412,2414,2416,2418,2420,2422,2424,2426,2429,2431],{"class":114,"line":180},[112,2411,2194],{"class":839},[112,2413,844],{"class":843},[112,2415,2199],{"class":2083},[112,2417,2106],{"class":843},[112,2419,2204],{"class":847},[112,2421,2106],{"class":843},[112,2423,2353],{"class":843},[112,2425,2100],{"class":843},[112,2427,2428],{"class":847},"service_type",[112,2430,2106],{"class":843},[112,2432,2209],{"class":2083},[112,2434,2435],{"class":114,"line":186},[112,2436,2214],{"class":843},[112,2438,2439],{"class":114,"line":192},[112,2440,2219],{"class":843},[73,2442,2443,2445],{},[76,2444,2224],{},": Les clients peuvent demander la planification de maintenance, et le système considère automatiquement les conditions météorologiques, la disponibilité des techniciens et les préférences client pour réserver des rendez-vous optimaux.",[809,2447,2449],{"id":2448},"outil-de-traitement-dachat","Outil de Traitement d'Achat",[102,2451,2453],{"className":2064,"code":2452,"language":2066,"meta":107,"style":107},"// Outil MCP: Traiter les commandes d'achat de chaudières\nconst purchaseTool = {\n name: \"purchase\",\n description: \"Traiter les commandes d'achat de chaudières\",\n inputSchema: {\n type: \"object\",\n properties: {\n customer_id: { type: \"string\" },\n boiler_model: { type: \"string\" },\n installation_date: { type: \"string\" }\n },\n required: [\"customer_id\", \"boiler_model\"]\n }\n}\n",[109,2454,2455,2460,2471,2486,2501,2509,2523,2531,2551,2572,2593,2597,2622,2626],{"__ignoreMap":107},[112,2456,2457],{"class":114,"line":115},[112,2458,2459],{"class":2073},"// Outil MCP: Traiter les commandes d'achat de chaudières\n",[112,2461,2462,2464,2467,2469],{"class":114,"line":121},[112,2463,2080],{"class":2079},[112,2465,2466],{"class":2083}," purchaseTool ",[112,2468,2087],{"class":843},[112,2470,2090],{"class":843},[112,2472,2473,2475,2477,2479,2482,2484],{"class":114,"line":127},[112,2474,2095],{"class":839},[112,2476,844],{"class":843},[112,2478,2100],{"class":843},[112,2480,2481],{"class":847},"purchase",[112,2483,2106],{"class":843},[112,2485,2109],{"class":843},[112,2487,2488,2490,2492,2494,2497,2499],{"class":114,"line":132},[112,2489,2114],{"class":839},[112,2491,844],{"class":843},[112,2493,2100],{"class":843},[112,2495,2496],{"class":847},"Traiter les commandes d'achat de chaudières",[112,2498,2106],{"class":843},[112,2500,2109],{"class":843},[112,2502,2503,2505,2507],{"class":114,"line":138},[112,2504,2130],{"class":839},[112,2506,844],{"class":843},[112,2508,2090],{"class":843},[112,2510,2511,2513,2515,2517,2519,2521],{"class":114,"line":144},[112,2512,2139],{"class":839},[112,2514,844],{"class":843},[112,2516,2100],{"class":843},[112,2518,2146],{"class":847},[112,2520,2106],{"class":843},[112,2522,2109],{"class":843},[112,2524,2525,2527,2529],{"class":114,"line":150},[112,2526,2155],{"class":839},[112,2528,844],{"class":843},[112,2530,2090],{"class":843},[112,2532,2533,2535,2537,2539,2541,2543,2545,2547,2549],{"class":114,"line":156},[112,2534,2164],{"class":839},[112,2536,844],{"class":843},[112,2538,2169],{"class":843},[112,2540,2172],{"class":839},[112,2542,844],{"class":843},[112,2544,2100],{"class":843},[112,2546,2179],{"class":847},[112,2548,2106],{"class":843},[112,2550,2331],{"class":843},[112,2552,2553,2556,2558,2560,2562,2564,2566,2568,2570],{"class":114,"line":162},[112,2554,2555],{"class":839}," boiler_model",[112,2557,844],{"class":843},[112,2559,2169],{"class":843},[112,2561,2172],{"class":839},[112,2563,844],{"class":843},[112,2565,2100],{"class":843},[112,2567,2179],{"class":847},[112,2569,2106],{"class":843},[112,2571,2331],{"class":843},[112,2573,2574,2577,2579,2581,2583,2585,2587,2589,2591],{"class":114,"line":168},[112,2575,2576],{"class":839}," installation_date",[112,2578,844],{"class":843},[112,2580,2169],{"class":843},[112,2582,2172],{"class":839},[112,2584,844],{"class":843},[112,2586,2100],{"class":843},[112,2588,2179],{"class":847},[112,2590,2106],{"class":843},[112,2592,2184],{"class":843},[112,2594,2595],{"class":114,"line":174},[112,2596,2189],{"class":843},[112,2598,2599,2601,2603,2605,2607,2609,2611,2613,2615,2618,2620],{"class":114,"line":180},[112,2600,2194],{"class":839},[112,2602,844],{"class":843},[112,2604,2199],{"class":2083},[112,2606,2106],{"class":843},[112,2608,2204],{"class":847},[112,2610,2106],{"class":843},[112,2612,2353],{"class":843},[112,2614,2100],{"class":843},[112,2616,2617],{"class":847},"boiler_model",[112,2619,2106],{"class":843},[112,2621,2209],{"class":2083},[112,2623,2624],{"class":114,"line":186},[112,2625,2214],{"class":843},[112,2627,2628],{"class":114,"line":192},[112,2629,2219],{"class":843},[73,2631,2632,2634],{},[76,2633,2224],{},": Quand les clients veulent mettre à niveau leur chaudière, le chatbot peut montrer les modèles disponibles, comparer les spécifications, traiter les paiements et planifier l'installation.",[809,2636,2638],{"id":2637},"outil-de-confirmation-par-email","Outil de Confirmation par Email",[102,2640,2642],{"className":2064,"code":2641,"language":2066,"meta":107,"style":107},"// Outil MCP: Envoyer des emails de confirmation\nconst emailTool = {\n name: \"email\",\n description: \"Envoyer des emails de confirmation\",\n inputSchema: {\n type: \"object\",\n properties: {\n customer_id: { type: \"string\" },\n email_type: { type: \"string\", enum: [\"maintenance\", \"purchase\", \"reminder\"] }\n },\n required: [\"customer_id\", \"email_type\"]\n }\n}\n",[109,2643,2644,2649,2660,2675,2690,2698,2712,2720,2740,2794,2798,2823,2827],{"__ignoreMap":107},[112,2645,2646],{"class":114,"line":115},[112,2647,2648],{"class":2073},"// Outil MCP: Envoyer des emails de confirmation\n",[112,2650,2651,2653,2656,2658],{"class":114,"line":121},[112,2652,2080],{"class":2079},[112,2654,2655],{"class":2083}," emailTool ",[112,2657,2087],{"class":843},[112,2659,2090],{"class":843},[112,2661,2662,2664,2666,2668,2671,2673],{"class":114,"line":127},[112,2663,2095],{"class":839},[112,2665,844],{"class":843},[112,2667,2100],{"class":843},[112,2669,2670],{"class":847},"email",[112,2672,2106],{"class":843},[112,2674,2109],{"class":843},[112,2676,2677,2679,2681,2683,2686,2688],{"class":114,"line":132},[112,2678,2114],{"class":839},[112,2680,844],{"class":843},[112,2682,2100],{"class":843},[112,2684,2685],{"class":847},"Envoyer des emails de confirmation",[112,2687,2106],{"class":843},[112,2689,2109],{"class":843},[112,2691,2692,2694,2696],{"class":114,"line":138},[112,2693,2130],{"class":839},[112,2695,844],{"class":843},[112,2697,2090],{"class":843},[112,2699,2700,2702,2704,2706,2708,2710],{"class":114,"line":144},[112,2701,2139],{"class":839},[112,2703,844],{"class":843},[112,2705,2100],{"class":843},[112,2707,2146],{"class":847},[112,2709,2106],{"class":843},[112,2711,2109],{"class":843},[112,2713,2714,2716,2718],{"class":114,"line":150},[112,2715,2155],{"class":839},[112,2717,844],{"class":843},[112,2719,2090],{"class":843},[112,2721,2722,2724,2726,2728,2730,2732,2734,2736,2738],{"class":114,"line":156},[112,2723,2164],{"class":839},[112,2725,844],{"class":843},[112,2727,2169],{"class":843},[112,2729,2172],{"class":839},[112,2731,844],{"class":843},[112,2733,2100],{"class":843},[112,2735,2179],{"class":847},[112,2737,2106],{"class":843},[112,2739,2331],{"class":843},[112,2741,2742,2745,2747,2749,2751,2753,2755,2757,2759,2761,2763,2765,2767,2769,2771,2773,2775,2777,2779,2781,2783,2785,2788,2790,2792],{"class":114,"line":162},[112,2743,2744],{"class":839}," email_type",[112,2746,844],{"class":843},[112,2748,2169],{"class":843},[112,2750,2172],{"class":839},[112,2752,844],{"class":843},[112,2754,2100],{"class":843},[112,2756,2179],{"class":847},[112,2758,2106],{"class":843},[112,2760,2353],{"class":843},[112,2762,2356],{"class":839},[112,2764,844],{"class":843},[112,2766,2199],{"class":2083},[112,2768,2106],{"class":843},[112,2770,2261],{"class":847},[112,2772,2106],{"class":843},[112,2774,2353],{"class":843},[112,2776,2100],{"class":843},[112,2778,2481],{"class":847},[112,2780,2106],{"class":843},[112,2782,2353],{"class":843},[112,2784,2100],{"class":843},[112,2786,2787],{"class":847},"reminder",[112,2789,2106],{"class":843},[112,2791,2379],{"class":2083},[112,2793,2219],{"class":843},[112,2795,2796],{"class":114,"line":168},[112,2797,2189],{"class":843},[112,2799,2800,2802,2804,2806,2808,2810,2812,2814,2816,2819,2821],{"class":114,"line":174},[112,2801,2194],{"class":839},[112,2803,844],{"class":843},[112,2805,2199],{"class":2083},[112,2807,2106],{"class":843},[112,2809,2204],{"class":847},[112,2811,2106],{"class":843},[112,2813,2353],{"class":843},[112,2815,2100],{"class":843},[112,2817,2818],{"class":847},"email_type",[112,2820,2106],{"class":843},[112,2822,2209],{"class":2083},[112,2824,2825],{"class":114,"line":180},[112,2826,2214],{"class":843},[112,2828,2829],{"class":114,"line":186},[112,2830,2219],{"class":843},[73,2832,2833,2835],{},[76,2834,2224],{},": Après toute réservation de service ou achat, le système envoie automatiquement des emails de confirmation appropriés avec des détails pertinents et les prochaines étapes.",[68,2837,2839,2845],{"className":2838},[332,333,334],[89,2840,2842],{"id":2841},"_3-scénarios-de-demo-et-flux",[76,2843,2844],{},"3. Scénarios de Demo et Flux",[73,2846,2847],{},[344,2848,2849],{"href":346},[76,2850,349],{},[68,2852,2854],{"className":2853},[71],[73,2855,2856],{},"Le projet MCP-in-Chat inclut des scénarios de demo complets qui montrent les modèles d'utilisation du monde réel et démontrent les capacités du système.",[809,2858,2860],{"id":2859},"scénario-1-requête-dinformation-de-chaudière","Scénario 1: Requête d'Information de Chaudière",[73,2862,2863,2866],{},[76,2864,2865],{},"Client",": \"Quel est le statut de garantie de ma chaudière?\"",[73,2868,2869,844],{},[76,2870,2871],{},"Réponse du Chatbot",[102,2873,2878],{"className":2874,"code":2876,"language":2877},[2875],"language-text","Je vais vérifier les informations de votre chaudière immédiatement.\n\nSelon vos dossiers, vous avez une Worcester Bosch 8000 Style installée le 15 mars 2023. \nVotre garantie est valide jusqu'au 15 mars 2028 (garantie fabricant de 5 ans).\n\nDétails clés:\n- Modèle: Worcester Bosch 8000 Style 30kW\n- Date d'Installation: 15 mars 2023\n- Statut de Garantie: Actif (3 ans restants)\n- Dernier Service: 12 janvier 2024\n- Prochain Service: 12 janvier 2025\n\nVoulez-vous que je planifie votre service annuel ou vérifie s'il y a des réclamations de garantie?\n","text",[109,2879,2876],{"__ignoreMap":107},[809,2881,2883],{"id":2882},"scénario-2-planification-de-maintenance","Scénario 2: Planification de Maintenance",[73,2885,2886,2888],{},[76,2887,2865],{},": \"J'ai besoin de planifier la maintenance de ma chaudière\"",[73,2890,2891,844],{},[76,2892,2871],{},[102,2894,2897],{"className":2895,"code":2896,"language":2877},[2875],"Je vais vous aider à planifier la maintenance de votre chaudière. Laissez-moi vérifier votre planning actuel et les créneaux disponibles.\n\nSelon votre localisation et préférences, j'ai trouvé ces rendez-vous disponibles:\n\nSemaine Prochaine:\n- Mardi 21 janvier: 9h00 - 11h00\n- Jeudi 23 janvier: 14h00 - 16h00\n\nLes prévisions météo montrent des conditions claires pour les deux jours. Votre technicien préféré, John Smith (enregistré Gas Safe), est disponible mardi.\n\nJe recommande mardi à 9h00. Le service inclura:\n- Inspection complète du système\n- Vérifications de sécurité\n- Tests d'efficacité\n- Nettoyage des filtres\n\nCoût: £120 (inclut pièces et main d'œuvre)\n\nDois-je réserver ce rendez-vous pour vous?\n",[109,2898,2896],{"__ignoreMap":107},[809,2900,2902],{"id":2901},"scénario-3-achat-de-mise-à-niveau-de-chaudière","Scénario 3: Achat de Mise à Niveau de Chaudière",[73,2904,2905,2907],{},[76,2906,2865],{},": \"Je veux mettre à niveau ma chaudière vers un modèle plus efficace\"",[73,2909,2910,844],{},[76,2911,2871],{},[102,2913,2916],{"className":2914,"code":2915,"language":2877},[2875],"Excellent choix! Laissez-moi vous montrer les modèles les plus efficaces disponibles et vous aider avec la mise à niveau.\n\nChaudières Haute Efficacité Disponibles:\n\n1. **Worcester Bosch 8000 Style** (Classe A)\n - Efficacité: 94%\n - Prix: £2,800\n - Garantie 10 ans\n - Thermostat intelligent inclus\n\n2. **Vaillant ecoTEC plus** (Classe A)\n - Efficacité: 92%\n - Prix: £2,600\n - Garantie 7 ans\n - Design compact\n\n3. **Ideal Logic+** (Classe A)\n - Efficacité: 90%\n - Prix: £2,200\n - Garantie 10 ans\n - Certifié Quiet Mark\n\nBasé sur votre système actuel, je recommande la Worcester Bosch 8000 Style pour une efficacité et fiabilité maximales.\n\nVoulez-vous que je:\n1. Montre les spécifications détaillées\n2. Vérifie la disponibilité d'installation\n3. Traite l'achat\n4. Compare avec votre chaudière actuelle\n",[109,2917,2915],{"__ignoreMap":107},[68,2919,2921,2927],{"className":2920},[332,333,334],[89,2922,2924],{"id":2923},"_4-spécifications-techniques",[76,2925,2926],{},"4. Spécifications Techniques",[73,2928,2929],{},[344,2930,2931],{"href":346},[76,2932,349],{},[68,2934,2936],{"className":2935},[71],[73,2937,2938],{},"Le projet MCP-in-Chat implémente une architecture sophistiquée en couches qui assure la sécurité, l'évolutivité et la maintenabilité.",[809,2940,2942],{"id":2941},"architecture-de-sécurité","Architecture de Sécurité",[102,2944,2946],{"className":104,"code":2945,"language":106,"meta":107,"style":107},"---\nheader: MCP Security Flow and Authentication\n---\nsequenceDiagram\n participant Chatbot as External Chatbot\n participant MCP as MCP Server\n participant API as Backend API\n participant Auth as Authentication\n participant Service as Business Service\n participant DB as Mock Database\n \n rect rgb(59, 130, 246)\n note right of Chatbot: External Systems\n Chatbot->>+MCP: Request with customer_id\n end\n \n rect rgb(139, 92, 246)\n note right of MCP: MCP Protocol Layer\n MCP->>+API: API call with validation\n end\n \n rect rgb(16, 185, 129)\n note right of API: Security & API Layer\n API->>+Auth: Verify authentication\n Auth-->>-API: Authentication result\n end\n \n rect rgb(245, 158, 11)\n note right of Service: Business Logic Layer\n API->>+Service: Process business logic\n Service->>+DB: Query with access controls\n end\n \n rect rgb(239, 68, 68)\n note right of DB: Data Storage Layer\n DB-->>-Service: Return filtered data\n end\n \n rect rgb(245, 158, 11)\n note right of Service: Business Logic Layer\n Service-->>-API: Processed response\n end\n \n rect rgb(16, 185, 129)\n note right of API: Security & API Layer\n API-->>-MCP: API response\n end\n \n rect rgb(139, 92, 246)\n note right of MCP: MCP Protocol Layer\n MCP-->>-Chatbot: Formatted result\n end\n \n Note over Chatbot,DB: Secure communication flow with\u003Cbr/>authentication at each layer\n",[109,2947,2948,2952,2957,2961,2966,2971,2976,2981,2986,2991,2996,3000,3005,3010,3015,3019,3023,3028,3033,3038,3042,3046,3051,3056,3061,3066,3070,3074,3079,3084,3089,3094,3098,3102,3107,3112,3117,3121,3125,3129,3133,3138,3142,3146,3150,3154,3159,3163,3167,3171,3175,3180,3184,3188],{"__ignoreMap":107},[112,2949,2950],{"class":114,"line":115},[112,2951,118],{},[112,2953,2954],{"class":114,"line":121},[112,2955,2956],{},"header: MCP Security Flow and Authentication\n",[112,2958,2959],{"class":114,"line":127},[112,2960,118],{},[112,2962,2963],{"class":114,"line":132},[112,2964,2965],{},"sequenceDiagram\n",[112,2967,2968],{"class":114,"line":138},[112,2969,2970],{}," participant Chatbot as External Chatbot\n",[112,2972,2973],{"class":114,"line":144},[112,2974,2975],{}," participant MCP as MCP Server\n",[112,2977,2978],{"class":114,"line":150},[112,2979,2980],{}," participant API as Backend API\n",[112,2982,2983],{"class":114,"line":156},[112,2984,2985],{}," participant Auth as Authentication\n",[112,2987,2988],{"class":114,"line":162},[112,2989,2990],{}," participant Service as Business Service\n",[112,2992,2993],{"class":114,"line":168},[112,2994,2995],{}," participant DB as Mock Database\n",[112,2997,2998],{"class":114,"line":174},[112,2999,485],{},[112,3001,3002],{"class":114,"line":180},[112,3003,3004],{}," rect rgb(59, 130, 246)\n",[112,3006,3007],{"class":114,"line":186},[112,3008,3009],{}," note right of Chatbot: External Systems\n",[112,3011,3012],{"class":114,"line":192},[112,3013,3014],{}," Chatbot->>+MCP: Request with customer_id\n",[112,3016,3017],{"class":114,"line":198},[112,3018,480],{},[112,3020,3021],{"class":114,"line":204},[112,3022,485],{},[112,3024,3025],{"class":114,"line":210},[112,3026,3027],{}," rect rgb(139, 92, 246)\n",[112,3029,3030],{"class":114,"line":216},[112,3031,3032],{}," note right of MCP: MCP Protocol Layer\n",[112,3034,3035],{"class":114,"line":222},[112,3036,3037],{}," MCP->>+API: API call with validation\n",[112,3039,3040],{"class":114,"line":228},[112,3041,480],{},[112,3043,3044],{"class":114,"line":234},[112,3045,485],{},[112,3047,3048],{"class":114,"line":240},[112,3049,3050],{}," rect rgb(16, 185, 129)\n",[112,3052,3053],{"class":114,"line":246},[112,3054,3055],{}," note right of API: Security & API Layer\n",[112,3057,3058],{"class":114,"line":252},[112,3059,3060],{}," API->>+Auth: Verify authentication\n",[112,3062,3063],{"class":114,"line":258},[112,3064,3065],{}," Auth-->>-API: Authentication result\n",[112,3067,3068],{"class":114,"line":264},[112,3069,480],{},[112,3071,3072],{"class":114,"line":270},[112,3073,485],{},[112,3075,3076],{"class":114,"line":276},[112,3077,3078],{}," rect rgb(245, 158, 11)\n",[112,3080,3081],{"class":114,"line":282},[112,3082,3083],{}," note right of Service: Business Logic Layer\n",[112,3085,3086],{"class":114,"line":288},[112,3087,3088],{}," API->>+Service: Process business logic\n",[112,3090,3091],{"class":114,"line":295},[112,3092,3093],{}," Service->>+DB: Query with access controls\n",[112,3095,3096],{"class":114,"line":301},[112,3097,480],{},[112,3099,3100],{"class":114,"line":307},[112,3101,485],{},[112,3103,3104],{"class":114,"line":313},[112,3105,3106],{}," rect rgb(239, 68, 68)\n",[112,3108,3109],{"class":114,"line":319},[112,3110,3111],{}," note right of DB: Data Storage Layer\n",[112,3113,3114],{"class":114,"line":325},[112,3115,3116],{}," DB-->>-Service: Return filtered data\n",[112,3118,3119],{"class":114,"line":1758},[112,3120,480],{},[112,3122,3123],{"class":114,"line":1764},[112,3124,485],{},[112,3126,3127],{"class":114,"line":1770},[112,3128,3078],{},[112,3130,3131],{"class":114,"line":1776},[112,3132,3083],{},[112,3134,3135],{"class":114,"line":1782},[112,3136,3137],{}," Service-->>-API: Processed response\n",[112,3139,3140],{"class":114,"line":1788},[112,3141,480],{},[112,3143,3144],{"class":114,"line":1793},[112,3145,485],{},[112,3147,3148],{"class":114,"line":1798},[112,3149,3050],{},[112,3151,3152],{"class":114,"line":1804},[112,3153,3055],{},[112,3155,3156],{"class":114,"line":1810},[112,3157,3158],{}," API-->>-MCP: API response\n",[112,3160,3161],{"class":114,"line":1816},[112,3162,480],{},[112,3164,3165],{"class":114,"line":1822},[112,3166,485],{},[112,3168,3169],{"class":114,"line":1828},[112,3170,3027],{},[112,3172,3173],{"class":114,"line":1833},[112,3174,3032],{},[112,3176,3177],{"class":114,"line":1838},[112,3178,3179],{}," MCP-->>-Chatbot: Formatted result\n",[112,3181,3182],{"class":114,"line":1844},[112,3183,480],{},[112,3185,3186],{"class":114,"line":1850},[112,3187,485],{},[112,3189,3190],{"class":114,"line":1856},[112,3191,3192],{}," Note over Chatbot,DB: Secure communication flow with\u003Cbr/>authentication at each layer\n",[809,3194,3196],{"id":3195},"implémentation-des-services-simulés","Implémentation des Services Simulés",[73,3198,3199],{},"Le projet inclut 8 services simulés complets qui simulent des opérations métier réelles:",[1433,3201,3202,3208,3214,3220,3226,3232,3238,3244],{},[366,3203,3204,3207],{},[76,3205,3206],{},"Service Client",": Gère les profils clients et préférences",[366,3209,3210,3213],{},[76,3211,3212],{},"Service Chaudières",": Gère les informations de chaudières et données de garantie",[366,3215,3216,3219],{},[76,3217,3218],{},"Service Paiement",": Traite les transactions et validation de paiements",[366,3221,3222,3225],{},[76,3223,3224],{},"Service Email",": Gère la communication et notifications",[366,3227,3228,3231],{},[76,3229,3230],{},"Service Météo",": Fournit les données météo pour la planification",[366,3233,3234,3237],{},[76,3235,3236],{},"Service Garantie",": Gère les informations de garantie et réclamations",[366,3239,3240,3243],{},[76,3241,3242],{},"Service Prédiction IA",": Simule les prédictions de maintenance",[366,3245,3246,3249],{},[76,3247,3248],{},"Service Technicien",": Gère la disponibilité et planification des techniciens",[809,3251,3253],{"id":3252},"structure-de-couche-de-données","Structure de Couche de Données",[102,3255,3259],{"className":3256,"code":3257,"language":3258,"meta":107,"style":107},"language-json shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","{\n \"customers\": {\n \"CUST001\": {\n \"name\": \"John Smith\",\n \"email\": \"john.smith@email.com\",\n \"address\": \"123 Main St, London\",\n \"phone\": \"+44 20 1234 5678\",\n \"preferences\": {\n \"contact_method\": \"email\",\n \"appointment_time\": \"morning\",\n \"technician_preference\": \"John Smith\"\n }\n }\n },\n \"boilers\": {\n \"BOIL001\": {\n \"customer_id\": \"CUST001\",\n \"model\": \"Worcester Bosch 8000 Style\",\n \"installation_date\": \"2023-03-15\",\n \"warranty_expiry\": \"2028-03-15\",\n \"last_service\": \"2024-01-12\",\n \"next_service_due\": \"2025-01-12\"\n }\n }\n}\n","json",[109,3260,3261,3266,3280,3295,3316,3335,3355,3375,3388,3408,3428,3446,3451,3456,3461,3474,3487,3505,3525,3545,3565,3585,3603,3607,3611],{"__ignoreMap":107},[112,3262,3263],{"class":114,"line":115},[112,3264,3265],{"class":843},"{\n",[112,3267,3268,3271,3274,3276,3278],{"class":114,"line":121},[112,3269,3270],{"class":843}," \"",[112,3272,3273],{"class":2079},"customers",[112,3275,2106],{"class":843},[112,3277,844],{"class":843},[112,3279,2090],{"class":843},[112,3281,3282,3285,3289,3291,3293],{"class":114,"line":127},[112,3283,3284],{"class":843}," \"",[112,3286,3288],{"class":3287},"sBMFI","CUST001",[112,3290,2106],{"class":843},[112,3292,844],{"class":843},[112,3294,2090],{"class":843},[112,3296,3297,3300,3303,3305,3307,3309,3312,3314],{"class":114,"line":132},[112,3298,3299],{"class":843}," \"",[112,3301,3302],{"class":924},"name",[112,3304,2106],{"class":843},[112,3306,844],{"class":843},[112,3308,2100],{"class":843},[112,3310,3311],{"class":847},"John Smith",[112,3313,2106],{"class":843},[112,3315,2109],{"class":843},[112,3317,3318,3320,3322,3324,3326,3328,3331,3333],{"class":114,"line":138},[112,3319,3299],{"class":843},[112,3321,2670],{"class":924},[112,3323,2106],{"class":843},[112,3325,844],{"class":843},[112,3327,2100],{"class":843},[112,3329,3330],{"class":847},"john.smith@email.com",[112,3332,2106],{"class":843},[112,3334,2109],{"class":843},[112,3336,3337,3339,3342,3344,3346,3348,3351,3353],{"class":114,"line":144},[112,3338,3299],{"class":843},[112,3340,3341],{"class":924},"address",[112,3343,2106],{"class":843},[112,3345,844],{"class":843},[112,3347,2100],{"class":843},[112,3349,3350],{"class":847},"123 Main St, London",[112,3352,2106],{"class":843},[112,3354,2109],{"class":843},[112,3356,3357,3359,3362,3364,3366,3368,3371,3373],{"class":114,"line":150},[112,3358,3299],{"class":843},[112,3360,3361],{"class":924},"phone",[112,3363,2106],{"class":843},[112,3365,844],{"class":843},[112,3367,2100],{"class":843},[112,3369,3370],{"class":847},"+44 20 1234 5678",[112,3372,2106],{"class":843},[112,3374,2109],{"class":843},[112,3376,3377,3379,3382,3384,3386],{"class":114,"line":156},[112,3378,3299],{"class":843},[112,3380,3381],{"class":924},"preferences",[112,3383,2106],{"class":843},[112,3385,844],{"class":843},[112,3387,2090],{"class":843},[112,3389,3390,3393,3396,3398,3400,3402,3404,3406],{"class":114,"line":162},[112,3391,3392],{"class":843}," \"",[112,3394,3395],{"class":839},"contact_method",[112,3397,2106],{"class":843},[112,3399,844],{"class":843},[112,3401,2100],{"class":843},[112,3403,2670],{"class":847},[112,3405,2106],{"class":843},[112,3407,2109],{"class":843},[112,3409,3410,3412,3415,3417,3419,3421,3424,3426],{"class":114,"line":168},[112,3411,3392],{"class":843},[112,3413,3414],{"class":839},"appointment_time",[112,3416,2106],{"class":843},[112,3418,844],{"class":843},[112,3420,2100],{"class":843},[112,3422,3423],{"class":847},"morning",[112,3425,2106],{"class":843},[112,3427,2109],{"class":843},[112,3429,3430,3432,3435,3437,3439,3441,3443],{"class":114,"line":174},[112,3431,3392],{"class":843},[112,3433,3434],{"class":839},"technician_preference",[112,3436,2106],{"class":843},[112,3438,844],{"class":843},[112,3440,2100],{"class":843},[112,3442,3311],{"class":847},[112,3444,3445],{"class":843},"\"\n",[112,3447,3448],{"class":114,"line":180},[112,3449,3450],{"class":843}," }\n",[112,3452,3453],{"class":114,"line":186},[112,3454,3455],{"class":843}," }\n",[112,3457,3458],{"class":114,"line":192},[112,3459,3460],{"class":843}," },\n",[112,3462,3463,3465,3468,3470,3472],{"class":114,"line":198},[112,3464,3270],{"class":843},[112,3466,3467],{"class":2079},"boilers",[112,3469,2106],{"class":843},[112,3471,844],{"class":843},[112,3473,2090],{"class":843},[112,3475,3476,3478,3481,3483,3485],{"class":114,"line":204},[112,3477,3284],{"class":843},[112,3479,3480],{"class":3287},"BOIL001",[112,3482,2106],{"class":843},[112,3484,844],{"class":843},[112,3486,2090],{"class":843},[112,3488,3489,3491,3493,3495,3497,3499,3501,3503],{"class":114,"line":210},[112,3490,3299],{"class":843},[112,3492,2204],{"class":924},[112,3494,2106],{"class":843},[112,3496,844],{"class":843},[112,3498,2100],{"class":843},[112,3500,3288],{"class":847},[112,3502,2106],{"class":843},[112,3504,2109],{"class":843},[112,3506,3507,3509,3512,3514,3516,3518,3521,3523],{"class":114,"line":216},[112,3508,3299],{"class":843},[112,3510,3511],{"class":924},"model",[112,3513,2106],{"class":843},[112,3515,844],{"class":843},[112,3517,2100],{"class":843},[112,3519,3520],{"class":847},"Worcester Bosch 8000 Style",[112,3522,2106],{"class":843},[112,3524,2109],{"class":843},[112,3526,3527,3529,3532,3534,3536,3538,3541,3543],{"class":114,"line":222},[112,3528,3299],{"class":843},[112,3530,3531],{"class":924},"installation_date",[112,3533,2106],{"class":843},[112,3535,844],{"class":843},[112,3537,2100],{"class":843},[112,3539,3540],{"class":847},"2023-03-15",[112,3542,2106],{"class":843},[112,3544,2109],{"class":843},[112,3546,3547,3549,3552,3554,3556,3558,3561,3563],{"class":114,"line":228},[112,3548,3299],{"class":843},[112,3550,3551],{"class":924},"warranty_expiry",[112,3553,2106],{"class":843},[112,3555,844],{"class":843},[112,3557,2100],{"class":843},[112,3559,3560],{"class":847},"2028-03-15",[112,3562,2106],{"class":843},[112,3564,2109],{"class":843},[112,3566,3567,3569,3572,3574,3576,3578,3581,3583],{"class":114,"line":234},[112,3568,3299],{"class":843},[112,3570,3571],{"class":924},"last_service",[112,3573,2106],{"class":843},[112,3575,844],{"class":843},[112,3577,2100],{"class":843},[112,3579,3580],{"class":847},"2024-01-12",[112,3582,2106],{"class":843},[112,3584,2109],{"class":843},[112,3586,3587,3589,3592,3594,3596,3598,3601],{"class":114,"line":240},[112,3588,3299],{"class":843},[112,3590,3591],{"class":924},"next_service_due",[112,3593,2106],{"class":843},[112,3595,844],{"class":843},[112,3597,2100],{"class":843},[112,3599,3600],{"class":847},"2025-01-12",[112,3602,3445],{"class":843},[112,3604,3605],{"class":114,"line":246},[112,3606,3455],{"class":843},[112,3608,3609],{"class":114,"line":252},[112,3610,2214],{"class":843},[112,3612,3613],{"class":114,"line":258},[112,3614,2219],{"class":843},[73,3616,3617,3620],{},[76,3618,3619],{},"Aperçu de la Structure de Données",": Le système utilise des fichiers JSON pour simuler une vraie base de données, avec des données structurées pour les clients, chaudières, planning de maintenance et commandes. Chaque enregistrement inclut des informations complètes nécessaires pour les interactions du chatbot.",[68,3622,3624,3630],{"className":3623},[332,333,334],[89,3625,3627],{"id":3626},"_5-tests-et-assurance-qualité",[76,3628,3629],{},"5. Tests et Assurance Qualité",[73,3631,3632],{},[344,3633,3634],{"href":346},[76,3635,349],{},[68,3637,3639],{"className":3638},[71],[73,3640,3641],{},"Le projet MCP-in-Chat inclut une suite de tests complète qui assure la fiabilité et la fonctionnalité dans tous les composants.",[809,3643,3645],{"id":3644},"couverture-de-tests","Couverture de Tests",[363,3647,3648,3654,3660,3666,3672],{},[366,3649,3650,3653],{},[76,3651,3652],{},"Outils MCP",": Tous les 4 outils testés avec divers scénarios d'entrée",[366,3655,3656,3659],{},[76,3657,3658],{},"Endpoints API",": Tests complets d'API REST avec authentification",[366,3661,3662,3665],{},[76,3663,3664],{},"Gestion d'Erreurs",": Validation des réponses d'erreur et cas limites",[366,3667,3668,3671],{},[76,3669,3670],{},"Intégration",": Tests de flux de travail de bout en bout",[366,3673,3674,3677],{},[76,3675,3676],{},"Performance",": Tests de temps de réponse et charge",[809,3679,3681],{"id":3680},"résumé-des-résultats-de-tests","Résumé des Résultats de Tests",[102,3683,3686],{"className":3684,"code":3685,"language":2877},[2875],"📊 RÉSUMÉ DES TESTS\n============================================================\nTotal des Tests: 5\nRéussis: 5 ✅\nÉchoués: 0 ❌\nTaux de Succès: 100.0%\n\n📋 RÉSULTATS DÉTAILLÉS\n------------------------------------------------------------\n1. boiler-info: ✅ RÉUSSI\n2. maintenance-get: ✅ RÉUSSI\n3. maintenance-schedule: ✅ RÉUSSI\n4. purchase: ✅ RÉUSSI\n5. email: ✅ RÉUSSI\n\n🎉 Tous les tests ont réussi! Les outils MCP fonctionnent correctement.\n",[109,3687,3685],{"__ignoreMap":107},[809,3689,3691],{"id":3690},"commandes-de-test-rapide","Commandes de Test Rapide",[102,3693,3697],{"className":3694,"code":3695,"language":3696,"meta":107,"style":107},"language-bash shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","# Exécuter tous les tests d'outils MCP\nnpm test\n\n# Tester des composants spécifiques\nnpm run test:api # Tests API backend\nnpm run test:mcp # Intégration MCP Cursor\nnpm run test:protocol # Validation protocole MCP\nnpm run test:compiled # Test version compilée\nnpm run test:connection # Test connexion\nnpm run test:quick # Vérification rapide santé\n","bash",[109,3698,3699,3704,3712,3716,3721,3734,3746,3758,3770,3782],{"__ignoreMap":107},[112,3700,3701],{"class":114,"line":115},[112,3702,3703],{"class":2073},"# Exécuter tous les tests d'outils MCP\n",[112,3705,3706,3709],{"class":114,"line":121},[112,3707,3708],{"class":3287},"npm",[112,3710,3711],{"class":847}," test\n",[112,3713,3714],{"class":114,"line":127},[112,3715,292],{"emptyLinePlaceholder":291},[112,3717,3718],{"class":114,"line":132},[112,3719,3720],{"class":2073},"# Tester des composants spécifiques\n",[112,3722,3723,3725,3728,3731],{"class":114,"line":138},[112,3724,3708],{"class":3287},[112,3726,3727],{"class":847}," run",[112,3729,3730],{"class":847}," test:api",[112,3732,3733],{"class":2073}," # Tests API backend\n",[112,3735,3736,3738,3740,3743],{"class":114,"line":144},[112,3737,3708],{"class":3287},[112,3739,3727],{"class":847},[112,3741,3742],{"class":847}," test:mcp",[112,3744,3745],{"class":2073}," # Intégration MCP Cursor\n",[112,3747,3748,3750,3752,3755],{"class":114,"line":150},[112,3749,3708],{"class":3287},[112,3751,3727],{"class":847},[112,3753,3754],{"class":847}," test:protocol",[112,3756,3757],{"class":2073}," # Validation protocole MCP\n",[112,3759,3760,3762,3764,3767],{"class":114,"line":156},[112,3761,3708],{"class":3287},[112,3763,3727],{"class":847},[112,3765,3766],{"class":847}," test:compiled",[112,3768,3769],{"class":2073}," # Test version compilée\n",[112,3771,3772,3774,3776,3779],{"class":114,"line":162},[112,3773,3708],{"class":3287},[112,3775,3727],{"class":847},[112,3777,3778],{"class":847}," test:connection",[112,3780,3781],{"class":2073}," # Test connexion\n",[112,3783,3784,3786,3788,3791],{"class":114,"line":168},[112,3785,3708],{"class":3287},[112,3787,3727],{"class":847},[112,3789,3790],{"class":847}," test:quick",[112,3792,3793],{"class":2073}," # Vérification rapide santé\n",[68,3795,3797,3803],{"className":3796},[332,333,334],[89,3798,3800],{"id":3799},"_6-commencer",[76,3801,3802],{},"6. Commencer",[73,3804,3805],{},[344,3806,3807],{"href":346},[76,3808,349],{},[68,3810,3812],{"className":3811},[71],[73,3813,3814],{},"Le projet MCP-in-Chat est conçu pour être facilement accessible aux développeurs qui veulent comprendre et implémenter des serveurs MCP pour leurs propres cas d'usage.",[809,3816,3818],{"id":3817},"prérequis","Prérequis",[363,3820,3821,3824,3827,3830],{},[366,3822,3823],{},"Node.js 18+ et npm",[366,3825,3826],{},"Connaissance de TypeScript",[366,3828,3829],{},"Compréhension de base du protocole MCP",[366,3831,3832],{},"Accès aux chatbots externes (Claude, ChatGPT, Cursor)",[809,3834,3836],{"id":3835},"installation","Installation",[102,3838,3840],{"className":3694,"code":3839,"language":3696,"meta":107,"style":107},"# Cloner le repository\ngit clone https://github.com/michaelwybraniec/mcp-in-chat.git\ncd mcp-in-chat\n\n# Installer les dépendances\nnpm install\n\n# Construire le projet\nnpm run build\n\n# Démarrer le serveur API\nnpm run start:api\n\n# Démarrer le serveur MCP (dans un autre terminal)\nnpm run start:mcp\n",[109,3841,3842,3847,3858,3867,3871,3876,3883,3887,3892,3901,3905,3910,3919,3923,3928],{"__ignoreMap":107},[112,3843,3844],{"class":114,"line":115},[112,3845,3846],{"class":2073},"# Cloner le repository\n",[112,3848,3849,3852,3855],{"class":114,"line":121},[112,3850,3851],{"class":3287},"git",[112,3853,3854],{"class":847}," clone",[112,3856,3857],{"class":847}," https://github.com/michaelwybraniec/mcp-in-chat.git\n",[112,3859,3860,3864],{"class":114,"line":127},[112,3861,3863],{"class":3862},"s2Zo4","cd",[112,3865,3866],{"class":847}," mcp-in-chat\n",[112,3868,3869],{"class":114,"line":132},[112,3870,292],{"emptyLinePlaceholder":291},[112,3872,3873],{"class":114,"line":138},[112,3874,3875],{"class":2073},"# Installer les dépendances\n",[112,3877,3878,3880],{"class":114,"line":144},[112,3879,3708],{"class":3287},[112,3881,3882],{"class":847}," install\n",[112,3884,3885],{"class":114,"line":150},[112,3886,292],{"emptyLinePlaceholder":291},[112,3888,3889],{"class":114,"line":156},[112,3890,3891],{"class":2073},"# Construire le projet\n",[112,3893,3894,3896,3898],{"class":114,"line":162},[112,3895,3708],{"class":3287},[112,3897,3727],{"class":847},[112,3899,3900],{"class":847}," build\n",[112,3902,3903],{"class":114,"line":168},[112,3904,292],{"emptyLinePlaceholder":291},[112,3906,3907],{"class":114,"line":174},[112,3908,3909],{"class":2073},"# Démarrer le serveur API\n",[112,3911,3912,3914,3916],{"class":114,"line":180},[112,3913,3708],{"class":3287},[112,3915,3727],{"class":847},[112,3917,3918],{"class":847}," start:api\n",[112,3920,3921],{"class":114,"line":186},[112,3922,292],{"emptyLinePlaceholder":291},[112,3924,3925],{"class":114,"line":192},[112,3926,3927],{"class":2073},"# Démarrer le serveur MCP (dans un autre terminal)\n",[112,3929,3930,3932,3934],{"class":114,"line":198},[112,3931,3708],{"class":3287},[112,3933,3727],{"class":847},[112,3935,3936],{"class":847}," start:mcp\n",[809,3938,3940],{"id":3939},"intégration-avec-chatbots-externes","Intégration avec Chatbots Externes",[73,3942,3943],{},"Le serveur MCP est conçu pour fonctionner avec n'importe quel chatbot compatible MCP:",[1433,3945,3946,3952,3958],{},[366,3947,3948,3951],{},[76,3949,3950],{},"Claude Desktop",": Ajouter le serveur MCP à la configuration de Claude",[366,3953,3954,3957],{},[76,3955,3956],{},"ChatGPT",": Utiliser le serveur MCP comme outil personnalisé",[366,3959,3960,3963],{},[76,3961,3962],{},"Cursor IDE",": Intégrer directement avec l'environnement de développement",[809,3965,3967],{"id":3966},"personnalisation","Personnalisation",[73,3969,3970],{},"Le projet est conçu pour être facilement personnalisable pour différents domaines métier:",[1433,3972,3973,3979,3985,3991],{},[366,3974,3975,3978],{},[76,3976,3977],{},"Remplacer les Services Simulés",": Connecter aux APIs et bases de données réelles",[366,3980,3981,3984],{},[76,3982,3983],{},"Ajouter de Nouveaux Outils",": Étendre le serveur MCP avec fonctionnalité supplémentaire",[366,3986,3987,3990],{},[76,3988,3989],{},"Modifier la Logique Métier",": Adapter les services pour votre cas d'usage spécifique",[366,3992,3993,3996],{},[76,3994,3995],{},"Améliorer la Sécurité",": Ajouter des couches supplémentaires d'authentification et autorisation",[68,3998,4000,4006],{"className":3999},[332,333,334],[89,4001,4003],{"id":4002},"_7-leçons-apprises",[76,4004,4005],{},"7. Leçons Apprises",[73,4007,4008],{},[344,4009,4010],{"href":346},[76,4011,349],{},[68,4013,4015],{"className":4014},[71],[73,4016,4017],{},"Le projet MCP-in-Chat fournit des insights précieux sur la construction de serveurs MCP prêts pour la production et leur intégration avec des chatbots externes.",[809,4019,4021],{"id":4020},"facteurs-clés-de-succès","Facteurs Clés de Succès",[1433,4023,4024,4030,4036,4042,4047],{},[366,4025,4026,4029],{},[76,4027,4028],{},"Architecture en Couches",": Séparation claire entre outils MCP, couche API et logique métier",[366,4031,4032,4035],{},[76,4033,4034],{},"Tests Complets",": 100% de couverture de tests assure la fiabilité",[366,4037,4038,4041],{},[76,4039,4040],{},"Sécurité en Premier",": Authentification, validation et limitation de débit dès le début",[366,4043,4044,4046],{},[76,4045,1420],{},": Documentation complète pour une intégration facile",[366,4048,4049,4052],{},[76,4050,4051],{},"Services Simulés",": Simulation de données réaliste pour développement et tests",[809,4054,4056],{"id":4055},"défis-communs-et-solutions","Défis Communs et Solutions",[1433,4058,4059,4078,4095,4112],{},[366,4060,4061,4064],{},[76,4062,4063],{},"Complexité d'Intégration Externe",[363,4065,4066,4072],{},[366,4067,4068,4071],{},[76,4069,4070],{},"Défi",": Tester avec des chatbots externes peut être difficile",[366,4073,4074,4077],{},[76,4075,4076],{},"Solution",": Suite de tests locaux complète avec scénarios réalistes",[366,4079,4080,4083],{},[76,4081,4082],{},"Implémentation de Sécurité",[363,4084,4085,4090],{},[366,4086,4087,4089],{},[76,4088,4070],{},": Assurer une communication sécurisée entre composants",[366,4091,4092,4094],{},[76,4093,4076],{},": Sécurité en couches avec authentification à chaque niveau",[366,4096,4097,4100],{},[76,4098,4099],{},"Gestion des Données",[363,4101,4102,4107],{},[366,4103,4104,4106],{},[76,4105,4070],{},": Maintenir un état cohérent entre services",[366,4108,4109,4111],{},[76,4110,4076],{},": Couche de données centralisée avec contrôles d'accès",[366,4113,4114,4116],{},[76,4115,3664],{},[363,4117,4118,4123],{},[366,4119,4120,4122],{},[76,4121,4070],{},": Gestion élégante des erreurs dans les interfaces conversationnelles",[366,4124,4125,4127],{},[76,4126,4076],{},": Réponses d'erreur structurées avec orientation actionnable",[809,4129,4131],{"id":4130},"considérations-de-performance","Considérations de Performance",[363,4133,4134,4140,4146,4152],{},[366,4135,4136,4139],{},[76,4137,4138],{},"Temps de Réponse",": Les outils MCP doivent répondre en 2-3 secondes",[366,4141,4142,4145],{},[76,4143,4144],{},"Limitation de Débit",": Implémenter des limites de débit appropriées pour prévenir l'abus",[366,4147,4148,4151],{},[76,4149,4150],{},"Cache",": Mettre en cache les données fréquemment accédées pour améliorer les performances",[366,4153,4154,4157],{},[76,4155,4156],{},"Monitoring",": Implémenter logging et monitoring complets",[68,4159,4161,4167],{"className":4160},[332,333,334],[89,4162,4164],{"id":4163},"_8-améliorations-futures",[76,4165,4166],{},"8. Améliorations Futures",[73,4168,4169],{},[344,4170,4171],{"href":346},[76,4172,349],{},[68,4174,4176],{"className":4175},[71],[73,4177,4178],{},"Le projet MCP-in-Chat sert de base pour des implémentations plus sophistiquées de serveurs MCP et fournit une feuille de route pour le développement futur.",[809,4180,4182],{"id":4181},"améliorations-planifiées","Améliorations Planifiées",[1433,4184,4185,4204,4223,4242],{},[366,4186,4187,4190],{},[76,4188,4189],{},"Intégration de Services Réels",[363,4191,4192,4195,4198,4201],{},[366,4193,4194],{},"Remplacer les services simulés par des APIs réelles",[366,4196,4197],{},"Intégrer avec des passerelles de paiement",[366,4199,4200],{},"Connecter aux bases de données clients",[366,4202,4203],{},"Implémenter des services email réels",[366,4205,4206,4209],{},[76,4207,4208],{},"Fonctionnalités Avancées",[363,4210,4211,4214,4217,4220],{},[366,4212,4213],{},"Support multilingue",[366,4215,4216],{},"Intégration vocale",[366,4218,4219],{},"Intégration d'application mobile",[366,4221,4222],{},"Analyses avancées et rapports",[366,4224,4225,4228],{},[76,4226,4227],{},"Améliorations d'Évolutivité",[363,4229,4230,4233,4236,4239],{},[366,4231,4232],{},"Architecture de microservices",[366,4234,4235],{},"Équilibrage de charge",[366,4237,4238],{},"Optimisation de base de données",[366,4240,4241],{},"Stratégies de cache",[366,4243,4244,4247],{},[76,4245,4246],{},"Améliorations de Sécurité",[363,4248,4249,4252,4255,4258],{},[366,4250,4251],{},"Intégration OAuth 2.0",[366,4253,4254],{},"Contrôle d'accès basé sur les rôles",[366,4256,4257],{},"Logging d'audit",[366,4259,4260],{},"Fonctionnalités de conformité",[809,4262,4264],{"id":4263},"applications-dindustrie","Applications d'Industrie",[73,4266,4267],{},"L'architecture MCP-in-Chat peut être adaptée pour diverses industries:",[1433,4269,4270,4276,4282,4288,4294],{},[366,4271,4272,4275],{},[76,4273,4274],{},"Santé",": Planification de rendez-vous patients et accès aux dossiers médicaux",[366,4277,4278,4281],{},[76,4279,4280],{},"Finance",": Gestion de comptes et traitement de transactions",[366,4283,4284,4287],{},[76,4285,4286],{},"Éducation",": Inscription aux cours et services étudiants",[366,4289,4290,4293],{},[76,4291,4292],{},"Immobilier",": Recherche de propriétés et gestion de réservations",[366,4295,4296,4299],{},[76,4297,4298],{},"Voyage",": Réservation de vols et gestion d'itinéraires",[68,4301,4303,4309],{"className":4302},[332,333,334],[89,4304,4306],{"id":4305},"_9-conclusion",[76,4307,4308],{},"9. Conclusion",[73,4310,4311],{},[344,4312,4313],{"href":346},[76,4314,349],{},[68,4316,4318],{"className":4317},[71],[73,4319,4320],{},"Le projet MCP-in-Chat démontre le pouvoir et le potentiel des serveurs du Protocole de Contexte de Modèle dans la création de systèmes d'IA conversationnelle sophistiqués. En fournissant un exemple complet et prêt pour la production, il sert à la fois de ressource d'apprentissage et de base pour des implémentations du monde réel.",[809,4322,4324],{"id":4323},"points-clés","Points Clés",[1433,4326,4327,4333,4339,4345,4351],{},[366,4328,4329,4332],{},[76,4330,4331],{},"Les Serveurs MCP Permettent une Valeur Commerciale Réelle",": Le projet montre comment les serveurs MCP peuvent connecter l'IA conversationnelle aux systèmes d'entreprise",[366,4334,4335,4338],{},[76,4336,4337],{},"Le Développement Systématique Fonctionne",": La méthodologie AWP a assuré un développement cohérent et de haute qualité",[366,4340,4341,4344],{},[76,4342,4343],{},"La Sécurité est Primordiale",": L'architecture de sécurité en couches est essentielle pour les systèmes de production",[366,4346,4347,4350],{},[76,4348,4349],{},"Les Tests sont Critiques",": Les tests complets assurent la fiabilité et la fonctionnalité",[366,4352,4353,4356],{},[76,4354,4355],{},"La Documentation Compte",": La documentation complète permet une adoption et personnalisation faciles",[809,4358,4359],{"id":800},"Prochaines Étapes",[73,4361,4362],{},"Pour les développeurs intéressés par les serveurs MCP:",[1433,4364,4365,4376,4382,4388],{},[366,4366,4367,4370,4371,4375],{},[76,4368,4369],{},"Étudier le Code",": Examiner le ",[344,4372,4374],{"href":1338,"rel":4373},[1276],"repository MCP-in-Chat"," pour comprendre l'implémentation",[366,4377,4378,4381],{},[76,4379,4380],{},"Exécuter le Demo",": Configurer le projet localement et tester avec des chatbots externes",[366,4383,4384,4387],{},[76,4385,4386],{},"Personnaliser",": Adapter l'architecture pour votre cas d'usage spécifique",[366,4389,4390,4393],{},[76,4391,4392],{},"Contribuer",": Partager les améliorations et améliorations avec la communauté",[73,4395,4396],{},"Le projet MCP-in-Chat représente un pas significatif vers l'avant dans la réalisation des serveurs MCP accessibles et pratiques pour les applications du monde réel. En fournissant un exemple complet et fonctionnel, il réduit la barrière d'entrée pour les développeurs qui veulent intégrer l'IA conversationnelle avec leurs systèmes métier.",[4398,4399],"hr",{},[73,4401,4402,844],{},[76,4403,4404],{},"Ressources",[363,4406,4407,4413,4420],{},[366,4408,4409],{},[344,4410,4412],{"href":1338,"rel":4411},[1276],"Repository MCP-in-Chat",[366,4414,4415],{},[344,4416,4419],{"href":4417,"rel":4418},"https://modelcontextprotocol.io/",[1276],"Documentation du Protocole de Contexte de Modèle",[366,4421,4422],{},[344,4423,1364],{"href":4424,"rel":4425},"https://github.com/michaelwybraniec/mcp-agentic-sldc",[1276],[1293,4427,4428],{},"html .light .shiki span {color: var(--shiki-light);background: var(--shiki-light-bg);font-style: var(--shiki-light-font-style);font-weight: var(--shiki-light-font-weight);text-decoration: var(--shiki-light-text-decoration);}html.light .shiki span {color: var(--shiki-light);background: var(--shiki-light-bg);font-style: var(--shiki-light-font-style);font-weight: var(--shiki-light-font-weight);text-decoration: var(--shiki-light-text-decoration);}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html pre.shiki code .sHwdD, html code.shiki .sHwdD{--shiki-light:#90A4AE;--shiki-light-font-style:italic;--shiki-default:#546E7A;--shiki-default-font-style:italic;--shiki-dark:#676E95;--shiki-dark-font-style:italic}html pre.shiki code .spNyl, html code.shiki .spNyl{--shiki-light:#9C3EDA;--shiki-default:#C792EA;--shiki-dark:#C792EA}html pre.shiki code .sTEyZ, html code.shiki .sTEyZ{--shiki-light:#90A4AE;--shiki-default:#EEFFFF;--shiki-dark:#BABED8}html pre.shiki code .sMK4o, html code.shiki .sMK4o{--shiki-light:#39ADB5;--shiki-default:#89DDFF;--shiki-dark:#89DDFF}html pre.shiki code .swJcz, html code.shiki .swJcz{--shiki-light:#E53935;--shiki-default:#F07178;--shiki-dark:#F07178}html pre.shiki code .sfazB, html code.shiki .sfazB{--shiki-light:#91B859;--shiki-default:#C3E88D;--shiki-dark:#C3E88D}html pre.shiki code .sBMFI, html code.shiki .sBMFI{--shiki-light:#E2931D;--shiki-default:#FFCB6B;--shiki-dark:#FFCB6B}html pre.shiki code .sbssI, html code.shiki .sbssI{--shiki-light:#F76D47;--shiki-default:#F78C6C;--shiki-dark:#F78C6C}html pre.shiki code .s2Zo4, html code.shiki .s2Zo4{--shiki-light:#6182B8;--shiki-default:#82AAFF;--shiki-dark:#82AAFF}",{"title":107,"searchDepth":121,"depth":121,"links":4430},[4431,4439,4445,4450,4455,4460,4466,4471,4475],{"id":1347,"depth":121,"text":1350,"children":4432},[4433,4434,4435,4436,4437,4438],{"id":1375,"depth":127,"text":1376},{"id":1394,"depth":127,"text":1395},{"id":1430,"depth":127,"text":1431},{"id":1519,"depth":127,"text":1520},{"id":1536,"depth":127,"text":1537},{"id":1578,"depth":127,"text":1579},{"id":2042,"depth":121,"text":2045,"children":4440},[4441,4442,4443,4444],{"id":2060,"depth":127,"text":2061},{"id":2228,"depth":127,"text":2229},{"id":2448,"depth":127,"text":2449},{"id":2637,"depth":127,"text":2638},{"id":2841,"depth":121,"text":2844,"children":4446},[4447,4448,4449],{"id":2859,"depth":127,"text":2860},{"id":2882,"depth":127,"text":2883},{"id":2901,"depth":127,"text":2902},{"id":2923,"depth":121,"text":2926,"children":4451},[4452,4453,4454],{"id":2941,"depth":127,"text":2942},{"id":3195,"depth":127,"text":3196},{"id":3252,"depth":127,"text":3253},{"id":3626,"depth":121,"text":3629,"children":4456},[4457,4458,4459],{"id":3644,"depth":127,"text":3645},{"id":3680,"depth":127,"text":3681},{"id":3690,"depth":127,"text":3691},{"id":3799,"depth":121,"text":3802,"children":4461},[4462,4463,4464,4465],{"id":3817,"depth":127,"text":3818},{"id":3835,"depth":127,"text":3836},{"id":3939,"depth":127,"text":3940},{"id":3966,"depth":127,"text":3967},{"id":4002,"depth":121,"text":4005,"children":4467},[4468,4469,4470],{"id":4020,"depth":127,"text":4021},{"id":4055,"depth":127,"text":4056},{"id":4130,"depth":127,"text":4131},{"id":4163,"depth":121,"text":4166,"children":4472},[4473,4474],{"id":4181,"depth":127,"text":4182},{"id":4263,"depth":127,"text":4264},{"id":4305,"depth":121,"text":4308,"children":4476},[4477,4478],{"id":4323,"depth":127,"text":4324},{"id":800,"depth":127,"text":4359},"2025-08-06","Une démonstration complète de comment construire un serveur MCP pour l\\'intégration de chatbot e-commerce, avec services de maintenance de chaudières, accès aux données en temps réel et flux de travail automatisés utilisant AWP.",{"src":1315},{},{"title":35,"description":4480},"P8LC4lAwNb3A1h50RgXfwGod9QuANRbeqN8tqrTsmTY",{"id":4486,"title":27,"authors":4487,"badge":4491,"body":4493,"date":6386,"description":6387,"extension":1313,"image":6388,"meta":6389,"navigation":291,"path":28,"seo":6390,"stem":29,"__hash__":6391},"posts_fr/fr/blog/5. mcp-tiny-agents-on-premises.md",[4488],{"name":55,"description":4489,"to":57,"avatar":4490},"Freelance, Serveurs MCP, Développeur Full-Stack, Architecture",{"src":59},{"label":4492,"color":1328,"variant":63},"MCP, On-Premises",{"type":65,"value":4494,"toc":6369},[4495,4499,4502,4802,4808,4810,4823,4870,4883,5583,6213,6345,6366],[89,4496,4498],{"id":4497},"vue-densemble-de-larchitecture","Vue d'Ensemble de l'Architecture",[73,4500,4501],{},"La beauté des Agents MCP Tiny réside dans leur simplicité architecturale. Qu'ils soient déployés dans le cloud ou on-premises, les composants principaux restent les mêmes : un agent léger, un client MCP et des outils connectés. Voici comment l'architecture complète on-premises se compare aux alternatives cloud :",[102,4503,4505],{"className":104,"code":4504,"language":106,"meta":107,"style":107},"graph TB\n subgraph \"Infrastructure On-Premises\"\n subgraph \"Stack IA Local\"\n Agent[\"Agent Tiny\u003Cbr/>(~50 lignes)\"]\n LocalLLM[\"LLM Local\u003Cbr/>Ollama/LM Studio\u003Cbr/>Qwen2.5-32B\"]\n MCPClient[\"Client MCP\u003Cbr/>Gestionnaire d'Outils\"]\n end\n \n subgraph \"Serveurs MCP Locaux\"\n FileServer[\"Système de Fichiers\u003Cbr/>Serveur MCP\"]\n WebServer[\"Playwright\u003Cbr/>Serveur MCP\"]\n BusinessAPI[\"API Métier\u003Cbr/>Serveur MCP\"]\n DatabaseServer[\"Base de Données\u003Cbr/>Serveur MCP\"]\n end\n \n subgraph \"Couche Matérielle\"\n GPU[\"GPU/CPU\u003Cbr/>16-140GB VRAM\"]\n Storage[\"Stockage Modèles\u003Cbr/>GGUF/Safetensors\"]\n end\n end\n \n subgraph \"Alternative Cloud (Article HF)\"\n CloudAgent[\"Agent Tiny\u003Cbr/>(Même Code)\"]\n CloudAPI[\"Nebius/Cohere\u003Cbr/>Qwen2.5-72B\"]\n CloudMCP[\"Client MCP Cloud\"]\n end\n \n subgraph \"Architecture Hybride\"\n Router[\"Routeur Intelligent\u003Cbr/>Classification des Données\"]\n LocalPath[\"Données Sensibles → Local\"]\n CloudPath[\"Tâches Complexes → Cloud\"]\n end\n \n Agent -->|\"Requêtes d'Outils\"| MCPClient\n MCPClient -->|\"Appels de Fonctions\"| LocalLLM\n LocalLLM -->|\"Inférence\"| GPU\n GPU -->|\"Chargement Modèles\"| Storage\n \n MCPClient -->|\"Exécution d'Outils\"| FileServer\n MCPClient -->|\"Navigation Web\"| WebServer\n MCPClient -->|\"Logique Métier\"| BusinessAPI\n MCPClient -->|\"Requêtes de Données\"| DatabaseServer\n \n CloudAgent -->|\"Appels API\"| CloudMCP\n CloudMCP -->|\"Inférence\"| CloudAPI\n \n Router -->|\"Décision de Route\"| LocalPath\n Router -->|\"Décision de Route\"| CloudPath\n LocalPath -->|\"Exécuter en Local\"| Agent\n CloudPath -->|\"Exécuter dans le Cloud\"| CloudAgent\n \n Agent -.->|\"Boucle While\u003Cbr/>Jusqu'à Completion\"| Agent\n \n classDef localInfra stroke:#0277bd,stroke-width:2px\n classDef cloudInfra stroke:#f57c00,stroke-width:2px\n classDef hybridInfra stroke:#7b1fa2,stroke-width:2px\n classDef hardware stroke:#388e3c,stroke-width:2px\n \n class Agent,LocalLLM,MCPClient,FileServer,WebServer,BusinessAPI,DatabaseServer localInfra\n class CloudAgent,CloudAPI,CloudMCP cloudInfra\n class Router,LocalPath,CloudPath hybridInfra\n class GPU,Storage hardware\n",[109,4506,4507,4512,4517,4522,4527,4532,4537,4542,4547,4552,4557,4562,4567,4572,4576,4580,4585,4590,4595,4599,4603,4607,4612,4617,4622,4627,4631,4635,4640,4645,4650,4655,4659,4663,4668,4673,4678,4683,4687,4692,4697,4702,4707,4711,4716,4721,4725,4730,4735,4740,4745,4749,4754,4758,4763,4768,4773,4778,4782,4787,4792,4797],{"__ignoreMap":107},[112,4508,4509],{"class":114,"line":115},[112,4510,4511],{},"graph TB\n",[112,4513,4514],{"class":114,"line":121},[112,4515,4516],{}," subgraph \"Infrastructure On-Premises\"\n",[112,4518,4519],{"class":114,"line":127},[112,4520,4521],{}," subgraph \"Stack IA Local\"\n",[112,4523,4524],{"class":114,"line":132},[112,4525,4526],{}," Agent[\"Agent Tiny\u003Cbr/>(~50 lignes)\"]\n",[112,4528,4529],{"class":114,"line":138},[112,4530,4531],{}," LocalLLM[\"LLM Local\u003Cbr/>Ollama/LM Studio\u003Cbr/>Qwen2.5-32B\"]\n",[112,4533,4534],{"class":114,"line":144},[112,4535,4536],{}," MCPClient[\"Client MCP\u003Cbr/>Gestionnaire d'Outils\"]\n",[112,4538,4539],{"class":114,"line":150},[112,4540,4541],{}," end\n",[112,4543,4544],{"class":114,"line":156},[112,4545,4546],{}," \n",[112,4548,4549],{"class":114,"line":162},[112,4550,4551],{}," subgraph \"Serveurs MCP Locaux\"\n",[112,4553,4554],{"class":114,"line":168},[112,4555,4556],{}," FileServer[\"Système de Fichiers\u003Cbr/>Serveur MCP\"]\n",[112,4558,4559],{"class":114,"line":174},[112,4560,4561],{}," WebServer[\"Playwright\u003Cbr/>Serveur MCP\"]\n",[112,4563,4564],{"class":114,"line":180},[112,4565,4566],{}," BusinessAPI[\"API Métier\u003Cbr/>Serveur MCP\"]\n",[112,4568,4569],{"class":114,"line":186},[112,4570,4571],{}," DatabaseServer[\"Base de Données\u003Cbr/>Serveur MCP\"]\n",[112,4573,4574],{"class":114,"line":192},[112,4575,4541],{},[112,4577,4578],{"class":114,"line":198},[112,4579,4546],{},[112,4581,4582],{"class":114,"line":204},[112,4583,4584],{}," subgraph \"Couche Matérielle\"\n",[112,4586,4587],{"class":114,"line":210},[112,4588,4589],{}," GPU[\"GPU/CPU\u003Cbr/>16-140GB VRAM\"]\n",[112,4591,4592],{"class":114,"line":216},[112,4593,4594],{}," Storage[\"Stockage Modèles\u003Cbr/>GGUF/Safetensors\"]\n",[112,4596,4597],{"class":114,"line":222},[112,4598,4541],{},[112,4600,4601],{"class":114,"line":228},[112,4602,480],{},[112,4604,4605],{"class":114,"line":234},[112,4606,485],{},[112,4608,4609],{"class":114,"line":240},[112,4610,4611],{}," subgraph \"Alternative Cloud (Article HF)\"\n",[112,4613,4614],{"class":114,"line":246},[112,4615,4616],{}," CloudAgent[\"Agent Tiny\u003Cbr/>(Même Code)\"]\n",[112,4618,4619],{"class":114,"line":252},[112,4620,4621],{}," CloudAPI[\"Nebius/Cohere\u003Cbr/>Qwen2.5-72B\"]\n",[112,4623,4624],{"class":114,"line":258},[112,4625,4626],{}," CloudMCP[\"Client MCP Cloud\"]\n",[112,4628,4629],{"class":114,"line":264},[112,4630,480],{},[112,4632,4633],{"class":114,"line":270},[112,4634,485],{},[112,4636,4637],{"class":114,"line":276},[112,4638,4639],{}," subgraph \"Architecture Hybride\"\n",[112,4641,4642],{"class":114,"line":282},[112,4643,4644],{}," Router[\"Routeur Intelligent\u003Cbr/>Classification des Données\"]\n",[112,4646,4647],{"class":114,"line":288},[112,4648,4649],{}," LocalPath[\"Données Sensibles → Local\"]\n",[112,4651,4652],{"class":114,"line":295},[112,4653,4654],{}," CloudPath[\"Tâches Complexes → Cloud\"]\n",[112,4656,4657],{"class":114,"line":301},[112,4658,480],{},[112,4660,4661],{"class":114,"line":307},[112,4662,485],{},[112,4664,4665],{"class":114,"line":313},[112,4666,4667],{}," Agent -->|\"Requêtes d'Outils\"| MCPClient\n",[112,4669,4670],{"class":114,"line":319},[112,4671,4672],{}," MCPClient -->|\"Appels de Fonctions\"| LocalLLM\n",[112,4674,4675],{"class":114,"line":325},[112,4676,4677],{}," LocalLLM -->|\"Inférence\"| GPU\n",[112,4679,4680],{"class":114,"line":1758},[112,4681,4682],{}," GPU -->|\"Chargement Modèles\"| Storage\n",[112,4684,4685],{"class":114,"line":1764},[112,4686,485],{},[112,4688,4689],{"class":114,"line":1770},[112,4690,4691],{}," MCPClient -->|\"Exécution d'Outils\"| FileServer\n",[112,4693,4694],{"class":114,"line":1776},[112,4695,4696],{}," MCPClient -->|\"Navigation Web\"| WebServer\n",[112,4698,4699],{"class":114,"line":1782},[112,4700,4701],{}," MCPClient -->|\"Logique Métier\"| BusinessAPI\n",[112,4703,4704],{"class":114,"line":1788},[112,4705,4706],{}," MCPClient -->|\"Requêtes de Données\"| DatabaseServer\n",[112,4708,4709],{"class":114,"line":1793},[112,4710,485],{},[112,4712,4713],{"class":114,"line":1798},[112,4714,4715],{}," CloudAgent -->|\"Appels API\"| CloudMCP\n",[112,4717,4718],{"class":114,"line":1804},[112,4719,4720],{}," CloudMCP -->|\"Inférence\"| CloudAPI\n",[112,4722,4723],{"class":114,"line":1810},[112,4724,485],{},[112,4726,4727],{"class":114,"line":1816},[112,4728,4729],{}," Router -->|\"Décision de Route\"| LocalPath\n",[112,4731,4732],{"class":114,"line":1822},[112,4733,4734],{}," Router -->|\"Décision de Route\"| CloudPath\n",[112,4736,4737],{"class":114,"line":1828},[112,4738,4739],{}," LocalPath -->|\"Exécuter en Local\"| Agent\n",[112,4741,4742],{"class":114,"line":1833},[112,4743,4744],{}," CloudPath -->|\"Exécuter dans le Cloud\"| CloudAgent\n",[112,4746,4747],{"class":114,"line":1838},[112,4748,485],{},[112,4750,4751],{"class":114,"line":1844},[112,4752,4753],{}," Agent -.->|\"Boucle While\u003Cbr/>Jusqu'à Completion\"| Agent\n",[112,4755,4756],{"class":114,"line":1850},[112,4757,485],{},[112,4759,4760],{"class":114,"line":1856},[112,4761,4762],{}," classDef localInfra stroke:#0277bd,stroke-width:2px\n",[112,4764,4765],{"class":114,"line":1862},[112,4766,4767],{}," classDef cloudInfra stroke:#f57c00,stroke-width:2px\n",[112,4769,4770],{"class":114,"line":1868},[112,4771,4772],{}," classDef hybridInfra stroke:#7b1fa2,stroke-width:2px\n",[112,4774,4775],{"class":114,"line":1874},[112,4776,4777],{}," classDef hardware stroke:#388e3c,stroke-width:2px\n",[112,4779,4780],{"class":114,"line":1880},[112,4781,485],{},[112,4783,4784],{"class":114,"line":1886},[112,4785,4786],{}," class Agent,LocalLLM,MCPClient,FileServer,WebServer,BusinessAPI,DatabaseServer localInfra\n",[112,4788,4789],{"class":114,"line":1892},[112,4790,4791],{}," class CloudAgent,CloudAPI,CloudMCP cloudInfra\n",[112,4793,4794],{"class":114,"line":1898},[112,4795,4796],{}," class Router,LocalPath,CloudPath hybridInfra\n",[112,4798,4799],{"class":114,"line":1903},[112,4800,4801],{}," class GPU,Storage hardware\n",[73,4803,4804,4807],{},[76,4805,4806],{},"Idée Clé"," : Le même code d'agent fonctionne sur tous les modèles de déploiement. La puissance des APIs standardisées signifie que votre investissement dans les outils MCP et la logique d'agent reste portable, que vous choisissiez la commodité du cloud, le contrôle on-premises ou une approche hybride stratégique.",[4398,4809],{},[68,4811,4813,4817],{"className":4812},[332,333,334],[89,4814,4816],{"id":4815},"le-piège-de-la-commodité-cloud","Le Piège de la Commodité Cloud",[73,4818,4819],{},[344,4820,4821],{"href":346},[76,4822,349],{},[68,4824,4826,4832,4835,4867],{"className":4825},[71],[73,4827,4828,4829],{},"L'implémentation de Hugging Face illustre l'élégance des architectures IA modernes. Avec seulement quelques lignes de TypeScript, vous pouvez créer un agent qui se connecte à plusieurs serveurs MCP (système de fichiers, navigation web via Playwright) et exploite des modèles puissants comme Qwen/Qwen2.5-72B-Instruct. L'idée fondamentale est profonde : ",[76,4830,4831],{},"\"Une fois que vous avez un Client MCP, un Agent n'est littéralement qu'une boucle while au-dessus.\"",[73,4833,4834],{},"Mais cette commodité s'accompagne de dépendances :",[363,4836,4837,4843,4849,4855,4861],{},[366,4838,4839,4842],{},[76,4840,4841],{},"Confidentialité des Données"," : Chaque requête, chaque appel d'outil, chaque contexte métier transite par des APIs externes",[366,4844,4845,4848],{},[76,4846,4847],{},"Imprévisibilité des Coûts"," : La tarification basée sur les tokens peut s'envoler avec des interactions d'agents complexes",[366,4850,4851,4854],{},[76,4852,4853],{},"Contraintes de Latence"," : Les allers-retours réseau ajoutent du délai à chaque étape d'inférence",[366,4856,4857,4860],{},[76,4858,4859],{},"Dépendance au Fournisseur"," : Changer de fournisseur nécessite des modifications de code et une revalidation",[366,4862,4863,4866],{},[76,4864,4865],{},"Problèmes de Conformité"," : Les industries réglementées peuvent interdire l'envoi de données vers des services externes",[73,4868,4869],{},"La question devient : Pouvons-nous maintenir la simplicité des Tiny Agents tout en obtenant un contrôle complet on-premises ?",[68,4871,4873,4877],{"className":4872},[332,333,334],[89,4874,4876],{"id":4875},"anatomie-dun-tiny-agent-local","Anatomie d'un Tiny Agent Local",[73,4878,4879],{},[344,4880,4881],{"href":346},[76,4882,349],{},[68,4884,4886,4889,4895,4898,4930,4933,4939,4942,4948,4951,4977,4980,5045,5048,5054,5057,5379,5392],{"className":4885},[71],[73,4887,4888],{},"La réponse est oui, mais avec des compromis importants. Analysons ce qui change lors du passage du cloud à l'on-premises :",[809,4890,4892],{"id":4891},"sélection-du-modèle-et-moteur-dinférence",[76,4893,4894],{},"Sélection du Modèle et Moteur d'Inférence",[73,4896,4897],{},"Au lieu d'appeler des APIs externes, nous avons besoin d'inférence locale. Les options se sont considérablement améliorées :",[363,4899,4900,4906,4912,4918,4924],{},[366,4901,4902,4905],{},[76,4903,4904],{},"Ollama"," : Déploiement le plus simple, supporte Qwen2.5, Llama 3.1 et d'autres modèles instruits",[366,4907,4908,4911],{},[76,4909,4910],{},"llama.cpp"," : Exécution directe de modèles avec inférence optimisée",[366,4913,4914,4917],{},[76,4915,4916],{},"LM Studio"," : Interface conviviale avec compatibilité API",[366,4919,4920,4923],{},[76,4921,4922],{},"vLLM"," : Service de niveau production avec endpoints compatibles OpenAI",[366,4925,4926,4929],{},[76,4927,4928],{},"LocalAI"," : Compatibilité complète avec l'API OpenAI pour les modèles locaux",[73,4931,4932],{},"L'idée clé de l'article HF s'applique ici : les LLMs modernes ont un support natif des appels de fonctions. Des modèles comme Qwen2.5-32B-Instruct, Llama 3.1-70B-Instruct, et même des variantes plus petites peuvent gérer efficacement l'utilisation d'outils.",[809,4934,4936],{"id":4935},"larchitecture-du-serveur-mcp-reste-inchangée",[76,4937,4938],{},"L'Architecture du Serveur MCP Reste Inchangée",[73,4940,4941],{},"C'est ici que brille le protocole MCP. Vos serveurs MCP existants—qu'ils exposent des systèmes de fichiers, des bases de données ou des APIs métier personnalisées—continuent de fonctionner sans modification. L'abstraction du protocole signifie que vos outils restent portables entre les déploiements cloud et on-premises.",[809,4943,4945],{"id":4944},"implémentation-modifiée-de-lagent",[76,4946,4947],{},"Implémentation Modifiée de l'Agent",[73,4949,4950],{},"La logique centrale de l'agent change à peine. Au lieu de :",[102,4952,4954],{"className":2064,"code":4953,"language":2066,"meta":107,"style":107},"const client = new InferenceClient(apiKey);\n",[109,4955,4956],{"__ignoreMap":107},[112,4957,4958,4960,4963,4965,4968,4971,4974],{"class":114,"line":115},[112,4959,2080],{"class":2079},[112,4961,4962],{"class":2083}," client ",[112,4964,2087],{"class":843},[112,4966,4967],{"class":843}," new",[112,4969,4970],{"class":3862}," InferenceClient",[112,4972,4973],{"class":2083},"(apiKey)",[112,4975,4976],{"class":843},";\n",[73,4978,4979],{},"Vous vous connectez à votre endpoint local :",[102,4981,4983],{"className":2064,"code":4982,"language":2066,"meta":107,"style":107},"const client = new InferenceClient({\n baseUrl: \"http://localhost:1234/v1\", // LM Studio\n apiKey: \"not-needed-for-local\"\n});\n",[109,4984,4985,5002,5021,5035],{"__ignoreMap":107},[112,4986,4987,4989,4991,4993,4995,4997,5000],{"class":114,"line":115},[112,4988,2080],{"class":2079},[112,4990,4962],{"class":2083},[112,4992,2087],{"class":843},[112,4994,4967],{"class":843},[112,4996,4970],{"class":3862},[112,4998,4999],{"class":2083},"(",[112,5001,3265],{"class":843},[112,5003,5004,5007,5009,5011,5014,5016,5018],{"class":114,"line":121},[112,5005,5006],{"class":839}," baseUrl",[112,5008,844],{"class":843},[112,5010,2100],{"class":843},[112,5012,5013],{"class":847},"http://localhost:1234/v1",[112,5015,2106],{"class":843},[112,5017,2353],{"class":843},[112,5019,5020],{"class":2073}," // LM Studio\n",[112,5022,5023,5026,5028,5030,5033],{"class":114,"line":127},[112,5024,5025],{"class":839}," apiKey",[112,5027,844],{"class":843},[112,5029,2100],{"class":843},[112,5031,5032],{"class":847},"not-needed-for-local",[112,5034,3445],{"class":843},[112,5036,5037,5040,5043],{"class":114,"line":132},[112,5038,5039],{"class":843},"}",[112,5041,5042],{"class":2083},")",[112,5044,4976],{"class":843},[73,5046,5047],{},"La boucle while, l'appel d'outils et l'intégration MCP restent identiques. C'est la puissance des APIs standardisées—l'agent ne se soucie pas d'où se produit l'inférence.",[809,5049,5051],{"id":5050},"la-boucle-while-en-action",[76,5052,5053],{},"La Boucle While en Action",[73,5055,5056],{},"Rappelez-vous l'idée centrale de l'article HF : \"un Agent n'est littéralement qu'une boucle while.\" Voici comment cela se déroule en pratique :",[102,5058,5060],{"className":104,"code":5059,"language":106,"meta":107,"style":107},"flowchart TD\n Start([\"Requête Utilisateur\u003Cbr/>Obtenir météo et sauvegarder\"])\n \n subgraph \"Boucle While du Tiny Agent (On-Premises)\"\n Initialize[\"Initialiser Agent\u003Cbr/>Charger LLM Local\u003Cbr/>Connecter Serveurs MCP\"]\n \n subgraph \"Boucle Principale\"\n ParseIntent[\"LLM Analyse Intention\u003Cbr/>Local Qwen2.5-32B\"]\n ToolDecision{\"Outils\u003Cbr/>Nécessaires?\"}\n \n subgraph \"Phase 1 d'Exécution\"\n CallTool[\"Appeler Outil MCP\u003Cbr/>get_weather(lat, lng)\"]\n ExecuteTool[\"Exécuter Outil\u003Cbr/>Récupérer Données Météo\"]\n ToolResult[\"Résultat\u003Cbr/>Température: 72°F\"]\n end\n \n FeedResult[\"Alimenter Résultat au LLM\u003Cbr/>Poursuivre Raisonnement\"]\n \n subgraph \"Phase 2 d'Exécution\"\n CallTool2[\"Appeler Autre Outil\u003Cbr/>write_file(weather.txt)\"]\n ExecuteTool2[\"Exécuter Écriture\u003Cbr/>Sauvegarder Données Météo\"]\n ToolResult2[\"Fichier Sauvegardé\u003Cbr/>Desktop/weather.txt\"]\n end\n \n Complete{\"Tâche Terminée?\"}\n Response[\"Générer Réponse\u003Cbr/>Météo sauvegardée avec succès\"]\n end\n end\n \n End([\"Tâche Accomplie\"])\n \n %% Connexions de flux\n Start --> Initialize\n Initialize --> ParseIntent\n ParseIntent --> ToolDecision\n \n ToolDecision -->|\"Oui - Besoin Météo\"| CallTool\n CallTool --> ExecuteTool\n ExecuteTool --> ToolResult\n ToolResult --> FeedResult\n \n FeedResult --> ToolDecision\n ToolDecision -->|\"Oui - Besoin Sauvegarde\"| CallTool2\n CallTool2 --> ExecuteTool2\n ExecuteTool2 --> ToolResult2\n ToolResult2 --> FeedResult\n \n ToolDecision -->|\"Plus d'Outils\"| Complete\n Complete -->|\"Oui\"| Response\n Complete -->|\"Non - Continuer\"| ParseIntent\n \n Response --> End\n \n %% Annotation idée clé\n LoopNote[\"Idée Centrale:\u003Cbr/>Agent = Boucle While\u003Cbr/>+ Client MCP\u003Cbr/>+ LLM Local\"]\n LoopNote -.-> ParseIntent\n \n %% Styles\n classDef agent stroke:#1976d2,stroke-width:2px\n classDef tool stroke:#388e3c,stroke-width:2px\n classDef decision stroke:#f57c00,stroke-width:2px\n classDef insight stroke:#c2185b,stroke-width:2px\n \n class Initialize,ParseIntent,FeedResult,Response agent\n class CallTool,ExecuteTool,ToolResult,CallTool2,ExecuteTool2,ToolResult2 tool\n class ToolDecision,Complete decision\n class LoopNote insight\n",[109,5061,5062,5066,5071,5075,5080,5085,5089,5094,5099,5104,5109,5114,5119,5124,5129,5134,5138,5143,5147,5152,5157,5162,5167,5171,5175,5180,5185,5189,5193,5197,5202,5206,5211,5216,5221,5226,5230,5235,5240,5245,5250,5254,5259,5264,5269,5274,5279,5283,5288,5293,5298,5302,5307,5311,5316,5321,5326,5330,5335,5340,5345,5350,5355,5359,5364,5369,5374],{"__ignoreMap":107},[112,5063,5064],{"class":114,"line":115},[112,5065,697],{},[112,5067,5068],{"class":114,"line":121},[112,5069,5070],{}," Start([\"Requête Utilisateur\u003Cbr/>Obtenir météo et sauvegarder\"])\n",[112,5072,5073],{"class":114,"line":127},[112,5074,485],{},[112,5076,5077],{"class":114,"line":132},[112,5078,5079],{}," subgraph \"Boucle While du Tiny Agent (On-Premises)\"\n",[112,5081,5082],{"class":114,"line":138},[112,5083,5084],{}," Initialize[\"Initialiser Agent\u003Cbr/>Charger LLM Local\u003Cbr/>Connecter Serveurs MCP\"]\n",[112,5086,5087],{"class":114,"line":144},[112,5088,4546],{},[112,5090,5091],{"class":114,"line":150},[112,5092,5093],{}," subgraph \"Boucle Principale\"\n",[112,5095,5096],{"class":114,"line":156},[112,5097,5098],{}," ParseIntent[\"LLM Analyse Intention\u003Cbr/>Local Qwen2.5-32B\"]\n",[112,5100,5101],{"class":114,"line":162},[112,5102,5103],{}," ToolDecision{\"Outils\u003Cbr/>Nécessaires?\"}\n",[112,5105,5106],{"class":114,"line":168},[112,5107,5108],{}," \n",[112,5110,5111],{"class":114,"line":174},[112,5112,5113],{}," subgraph \"Phase 1 d'Exécution\"\n",[112,5115,5116],{"class":114,"line":180},[112,5117,5118],{}," CallTool[\"Appeler Outil MCP\u003Cbr/>get_weather(lat, lng)\"]\n",[112,5120,5121],{"class":114,"line":186},[112,5122,5123],{}," ExecuteTool[\"Exécuter Outil\u003Cbr/>Récupérer Données Météo\"]\n",[112,5125,5126],{"class":114,"line":192},[112,5127,5128],{}," ToolResult[\"Résultat\u003Cbr/>Température: 72°F\"]\n",[112,5130,5131],{"class":114,"line":198},[112,5132,5133],{}," end\n",[112,5135,5136],{"class":114,"line":204},[112,5137,5108],{},[112,5139,5140],{"class":114,"line":210},[112,5141,5142],{}," FeedResult[\"Alimenter Résultat au LLM\u003Cbr/>Poursuivre Raisonnement\"]\n",[112,5144,5145],{"class":114,"line":216},[112,5146,5108],{},[112,5148,5149],{"class":114,"line":222},[112,5150,5151],{}," subgraph \"Phase 2 d'Exécution\"\n",[112,5153,5154],{"class":114,"line":228},[112,5155,5156],{}," CallTool2[\"Appeler Autre Outil\u003Cbr/>write_file(weather.txt)\"]\n",[112,5158,5159],{"class":114,"line":234},[112,5160,5161],{}," ExecuteTool2[\"Exécuter Écriture\u003Cbr/>Sauvegarder Données Météo\"]\n",[112,5163,5164],{"class":114,"line":240},[112,5165,5166],{}," ToolResult2[\"Fichier Sauvegardé\u003Cbr/>Desktop/weather.txt\"]\n",[112,5168,5169],{"class":114,"line":246},[112,5170,5133],{},[112,5172,5173],{"class":114,"line":252},[112,5174,5108],{},[112,5176,5177],{"class":114,"line":258},[112,5178,5179],{}," Complete{\"Tâche Terminée?\"}\n",[112,5181,5182],{"class":114,"line":264},[112,5183,5184],{}," Response[\"Générer Réponse\u003Cbr/>Météo sauvegardée avec succès\"]\n",[112,5186,5187],{"class":114,"line":270},[112,5188,4541],{},[112,5190,5191],{"class":114,"line":276},[112,5192,480],{},[112,5194,5195],{"class":114,"line":282},[112,5196,485],{},[112,5198,5199],{"class":114,"line":288},[112,5200,5201],{}," End([\"Tâche Accomplie\"])\n",[112,5203,5204],{"class":114,"line":295},[112,5205,485],{},[112,5207,5208],{"class":114,"line":301},[112,5209,5210],{}," %% Connexions de flux\n",[112,5212,5213],{"class":114,"line":307},[112,5214,5215],{}," Start --> Initialize\n",[112,5217,5218],{"class":114,"line":313},[112,5219,5220],{}," Initialize --> ParseIntent\n",[112,5222,5223],{"class":114,"line":319},[112,5224,5225],{}," ParseIntent --> ToolDecision\n",[112,5227,5228],{"class":114,"line":325},[112,5229,485],{},[112,5231,5232],{"class":114,"line":1758},[112,5233,5234],{}," ToolDecision -->|\"Oui - Besoin Météo\"| CallTool\n",[112,5236,5237],{"class":114,"line":1764},[112,5238,5239],{}," CallTool --> ExecuteTool\n",[112,5241,5242],{"class":114,"line":1770},[112,5243,5244],{}," ExecuteTool --> ToolResult\n",[112,5246,5247],{"class":114,"line":1776},[112,5248,5249],{}," ToolResult --> FeedResult\n",[112,5251,5252],{"class":114,"line":1782},[112,5253,485],{},[112,5255,5256],{"class":114,"line":1788},[112,5257,5258],{}," FeedResult --> ToolDecision\n",[112,5260,5261],{"class":114,"line":1793},[112,5262,5263],{}," ToolDecision -->|\"Oui - Besoin Sauvegarde\"| CallTool2\n",[112,5265,5266],{"class":114,"line":1798},[112,5267,5268],{}," CallTool2 --> ExecuteTool2\n",[112,5270,5271],{"class":114,"line":1804},[112,5272,5273],{}," ExecuteTool2 --> ToolResult2\n",[112,5275,5276],{"class":114,"line":1810},[112,5277,5278],{}," ToolResult2 --> FeedResult\n",[112,5280,5281],{"class":114,"line":1816},[112,5282,485],{},[112,5284,5285],{"class":114,"line":1822},[112,5286,5287],{}," ToolDecision -->|\"Plus d'Outils\"| Complete\n",[112,5289,5290],{"class":114,"line":1828},[112,5291,5292],{}," Complete -->|\"Oui\"| Response\n",[112,5294,5295],{"class":114,"line":1833},[112,5296,5297],{}," Complete -->|\"Non - Continuer\"| ParseIntent\n",[112,5299,5300],{"class":114,"line":1838},[112,5301,485],{},[112,5303,5304],{"class":114,"line":1844},[112,5305,5306],{}," Response --> End\n",[112,5308,5309],{"class":114,"line":1850},[112,5310,485],{},[112,5312,5313],{"class":114,"line":1856},[112,5314,5315],{}," %% Annotation idée clé\n",[112,5317,5318],{"class":114,"line":1862},[112,5319,5320],{}," LoopNote[\"Idée Centrale:\u003Cbr/>Agent = Boucle While\u003Cbr/>+ Client MCP\u003Cbr/>+ LLM Local\"]\n",[112,5322,5323],{"class":114,"line":1868},[112,5324,5325],{}," LoopNote -.-> ParseIntent\n",[112,5327,5328],{"class":114,"line":1874},[112,5329,485],{},[112,5331,5332],{"class":114,"line":1880},[112,5333,5334],{}," %% Styles\n",[112,5336,5337],{"class":114,"line":1886},[112,5338,5339],{}," classDef agent stroke:#1976d2,stroke-width:2px\n",[112,5341,5342],{"class":114,"line":1892},[112,5343,5344],{}," classDef tool stroke:#388e3c,stroke-width:2px\n",[112,5346,5347],{"class":114,"line":1898},[112,5348,5349],{}," classDef decision stroke:#f57c00,stroke-width:2px\n",[112,5351,5352],{"class":114,"line":1903},[112,5353,5354],{}," classDef insight stroke:#c2185b,stroke-width:2px\n",[112,5356,5357],{"class":114,"line":1909},[112,5358,485],{},[112,5360,5361],{"class":114,"line":1915},[112,5362,5363],{}," class Initialize,ParseIntent,FeedResult,Response agent\n",[112,5365,5366],{"class":114,"line":1921},[112,5367,5368],{}," class CallTool,ExecuteTool,ToolResult,CallTool2,ExecuteTool2,ToolResult2 tool\n",[112,5370,5371],{"class":114,"line":1927},[112,5372,5373],{}," class ToolDecision,Complete decision\n",[112,5375,5376],{"class":114,"line":1933},[112,5377,5378],{}," class LoopNote insight\n",[68,5380,5382,5386],{"className":5381},[332,333,334],[89,5383,5385],{"id":5384},"défis-dimplémentation","Défis d'Implémentation",[73,5387,5388],{},[344,5389,5390],{"href":346},[76,5391,349],{},[68,5393,5395,5398,5404,5407,5427,5433,5439,5442,5462,5468,5474,5477,5503,5509,5535,5541,5544,5570],{"className":5394},[71],[73,5396,5397],{},"Le passage à l'on-premises n'est pas sans défis. Voici les considérations clés :",[809,5399,5401],{"id":5400},"exigences-matérielles",[76,5402,5403],{},"Exigences Matérielles",[73,5405,5406],{},"Contrairement aux fournisseurs cloud avec leurs clusters massifs de GPUs, vous êtes limité par le matériel local :",[363,5408,5409,5415,5421],{},[366,5410,5411,5414],{},[76,5412,5413],{},"Mémoire"," : Les modèles 70B nécessitent ~140GB VRAM pour une inférence confortable",[366,5416,5417,5420],{},[76,5418,5419],{},"Modèles Plus Petits"," : Les modèles 7B-13B peuvent fonctionner sur des GPUs grand public avec 16-24GB VRAM",[366,5422,5423,5426],{},[76,5424,5425],{},"Inférence CPU"," : Possible mais significativement plus lente, particulièrement pour l'utilisation complexe d'outils",[73,5428,5429,5432],{},[76,5430,5431],{},"Approche Pratique"," : Commencez avec des modèles quantifiés (format GGUF) qui peuvent fonctionner sur le matériel disponible. Un modèle 32B bien quantifié surpasse souvent un modèle 70B mal configuré.",[809,5434,5436],{"id":5435},"compromis-de-performance",[76,5437,5438],{},"Compromis de Performance",[73,5440,5441],{},"L'inférence locale introduit une latence que les fournisseurs cloud ont optimisée :",[363,5443,5444,5450,5456],{},[366,5445,5446,5449],{},[76,5447,5448],{},"Latence du Premier Token"," : Les modèles locaux nécessitent un temps d'initialisation",[366,5451,5452,5455],{},[76,5453,5454],{},"Débit"," : Les configurations mono-GPU ne peuvent pas égaler l'inférence distribuée dans le cloud",[366,5457,5458,5461],{},[76,5459,5460],{},"Concurrence"," : Plusieurs sessions d'agents se disputent les mêmes ressources locales",[73,5463,5464,5467],{},[76,5465,5466],{},"Stratégie d'Atténuation"," : Gardez les modèles chargés en mémoire entre les requêtes, utilisez la mise en cache des modèles et envisagez d'exécuter plusieurs petits modèles plutôt qu'un grand.",[809,5469,5471],{"id":5470},"critères-de-sélection-des-modèles",[76,5472,5473],{},"Critères de Sélection des Modèles",[73,5475,5476],{},"Tous les modèles ne sont pas égaux pour le déploiement on-premises :",[363,5478,5479,5485,5491,5497],{},[366,5480,5481,5484],{},[76,5482,5483],{},"Qualité des Appels de Fonction"," : Testez extensivement avec vos outils MCP spécifiques",[366,5486,5487,5490],{},[76,5488,5489],{},"Longueur de Contexte"," : Des contextes plus longs permettent des conversations d'agents plus sophistiquées",[366,5492,5493,5496],{},[76,5494,5495],{},"Tolérance à la Quantification"," : Certains modèles se dégradent significativement une fois quantifiés",[366,5498,5499,5502],{},[76,5500,5501],{},"Licences"," : Assurez-vous des droits d'utilisation commerciale pour les déploiements en entreprise",[73,5504,5505,5508],{},[76,5506,5507],{},"Modèles Recommandés pour l'On-Premises"," :",[363,5510,5511,5517,5523,5529],{},[366,5512,5513,5516],{},[76,5514,5515],{},"Qwen2.5-32B-Instruct"," : Excellent pour les appels de fonction, exigences matérielles raisonnables",[366,5518,5519,5522],{},[76,5520,5521],{},"Llama 3.1-70B-Instruct"," : Si vous avez le matériel, performances exceptionnelles",[366,5524,5525,5528],{},[76,5526,5527],{},"Mistral-Small-3.1-24B"," : Optimisé spécifiquement pour les appels de fonction",[366,5530,5531,5534],{},[76,5532,5533],{},"Gemma 3 27B"," : Bon équilibre entre capacité et efficacité",[809,5536,5538],{"id":5537},"complexité-dintégration",[76,5539,5540],{},"Complexité d'Intégration",[73,5542,5543],{},"Les fournisseurs cloud gèrent la compatibilité des APIs, mais les configurations locales nécessitent plus de configuration :",[363,5545,5546,5552,5558,5564],{},[366,5547,5548,5551],{},[76,5549,5550],{},"Passerelle API"," : Assurer des endpoints compatibles avec OpenAI",[366,5553,5554,5557],{},[76,5555,5556],{},"Équilibrage de Charge"," : Distribuer les requêtes entre plusieurs instances de modèles",[366,5559,5560,5563],{},[76,5561,5562],{},"Surveillance"," : Suivi des performances, utilisation des ressources et taux d'erreur",[366,5565,5566,5569],{},[76,5567,5568],{},"Mises à Jour"," : Gestion des mises à jour de modèles et contrôle de version",[68,5571,5573,5577],{"className":5572},[332,333,334],[89,5574,5576],{"id":5575},"cadre-de-décision-entreprise","Cadre de Décision Entreprise",[73,5578,5579],{},[344,5580,5581],{"href":346},[76,5582,349],{},[68,5584,5586,5589,5595,5598,5870,5876,5881,5892,5897,5908,5914,5920,5923,5949,5952,5972,5978,5981,6005,6018],{"className":5585},[71],[73,5587,5588],{},"La décision entre les agents MCP cloud et on-premises n'est pas purement technique—elle est stratégique. Comprendre les compromis est essentiel pour prendre des décisions architecturales éclairées.",[809,5590,5592],{"id":5591},"comparaison-cloud-vs-on-premises",[76,5593,5594],{},"Comparaison Cloud vs On-Premises",[73,5596,5597],{},"Voici une comparaison complète pour guider votre choix :",[102,5599,5601],{"className":104,"code":5600,"language":106,"meta":107,"style":107},"graph TD\n subgraph Cloud [\"Agents MCP Cloud\"]\n CloudAdvantages[\"Avantages\u003Cbr/>• Modèles puissants (70B+)\u003Cbr/>• Calcul illimité\u003Cbr/>• Pas d'investissement matériel\u003Cbr/>• Mise à l'échelle instantanée\u003Cbr/>• Infrastructure gérée\"]\n \n CloudRisks[\"Préoccupations de Sécurité\u003Cbr/>• Données hors site\u003Cbr/>• Dépendance fournisseur\u003Cbr/>• Coûts imprévisibles\u003Cbr/>• Défis de conformité\u003Cbr/>• Dépendances API\"]\n \n CloudCosts[\"Modèle de Coûts\u003Cbr/>• Paiement par token\u003Cbr/>• 2 000-10 000€/mois\u003Cbr/>• Mise à l'échelle variable\u003Cbr/>• Pas d'investissement initial\"]\n end\n \n subgraph OnPrem [\"Agents MCP On-Premises\"]\n OnPremAdvantages[\"Bénéfices Sécurité\u003Cbr/>• Souveraineté totale des données\u003Cbr/>• Contrôle total des audits\u003Cbr/>• Conforme aux réglementations\u003Cbr/>• Sans dépendance fournisseur\u003Cbr/>• Fonctionnement hors ligne\"]\n \n OnPremChallenges[\"Défis d'Implémentation\u003Cbr/>• Investissement matériel requis\u003Cbr/>• Limites de performance modèles\u003Cbr/>• Complexité opérationnelle\u003Cbr/>• Contraintes d'échelle\u003Cbr/>• Mises à jour manuelles\"]\n \n OnPremCosts[\"Structure de Coûts\u003Cbr/>• 10 000-50 000€ initial\u003Cbr/>• Rentabilité 6-18 mois\u003Cbr/>• Coûts opérationnels fixes\u003Cbr/>• Mise à l'échelle prévisible\"]\n end\n \n subgraph Decision [\"Facteurs de Décision\"]\n DataSensitivity[\"Sensibilité des Données\u003Cbr/>Haute sensibilité → On-Premises\u003Cbr/>Faible sensibilité → Cloud\"]\n \n Compliance[\"Exigences de Conformité\u003Cbr/>Réglementation stricte → On-Premises\u003Cbr/>Conformité standard → Cloud\"]\n \n TechnicalCapacity[\"Ressources Techniques\u003Cbr/>Équipe IA/ML forte → On-Premises\u003Cbr/>Ressources limitées → Cloud\"]\n \n CostModel[\"Préférences de Coûts\u003Cbr/>Coûts prévisibles → On-Premises\u003Cbr/>Coûts variables → Cloud\"]\n end\n \n subgraph Hybrid [\"Architecture Hybride\"]\n HybridBenefits[\"Combinaison Stratégique\u003Cbr/>• Router données sensibles en local\u003Cbr/>• Utiliser cloud pour tâches complexes\u003Cbr/>• Optimiser coûts dynamiquement\u003Cbr/>• Distribuer risque opérationnel\"]\n end\n \n %% Flux de décision\n DataSensitivity --> OnPremAdvantages\n DataSensitivity --> CloudAdvantages\n \n Compliance --> OnPremAdvantages\n Compliance --> CloudAdvantages\n \n TechnicalCapacity --> OnPremChallenges\n TechnicalCapacity --> CloudRisks\n \n CostModel --> OnPremCosts\n CostModel --> CloudCosts\n \n %% Connexions hybrides\n OnPremAdvantages -.-> HybridBenefits\n CloudAdvantages -.-> HybridBenefits\n \n %% Styles\n classDef cloudStyle stroke:#1976d2,stroke-width:2px\n classDef onpremStyle stroke:#388e3c,stroke-width:2px\n classDef decisionStyle stroke:#f57c00,stroke-width:2px\n classDef hybridStyle stroke:#7b1fa2,stroke-width:2px\n \n class Cloud,CloudAdvantages,CloudRisks,CloudCosts cloudStyle\n class OnPrem,OnPremAdvantages,OnPremChallenges,OnPremCosts onpremStyle\n class Decision,DataSensitivity,Compliance,TechnicalCapacity,CostModel decisionStyle\n class Hybrid,HybridBenefits hybridStyle\n",[109,5602,5603,5607,5612,5617,5621,5626,5630,5635,5639,5643,5648,5653,5657,5662,5666,5671,5675,5679,5684,5689,5693,5698,5702,5707,5711,5716,5720,5724,5729,5734,5738,5742,5747,5752,5757,5761,5766,5771,5775,5780,5785,5789,5794,5799,5803,5808,5813,5818,5822,5826,5831,5836,5841,5846,5850,5855,5860,5865],{"__ignoreMap":107},[112,5604,5605],{"class":114,"line":115},[112,5606,465],{},[112,5608,5609],{"class":114,"line":121},[112,5610,5611],{}," subgraph Cloud [\"Agents MCP Cloud\"]\n",[112,5613,5614],{"class":114,"line":127},[112,5615,5616],{}," CloudAdvantages[\"Avantages\u003Cbr/>• Modèles puissants (70B+)\u003Cbr/>• Calcul illimité\u003Cbr/>• Pas d'investissement matériel\u003Cbr/>• Mise à l'échelle instantanée\u003Cbr/>• Infrastructure gérée\"]\n",[112,5618,5619],{"class":114,"line":132},[112,5620,4546],{},[112,5622,5623],{"class":114,"line":138},[112,5624,5625],{}," CloudRisks[\"Préoccupations de Sécurité\u003Cbr/>• Données hors site\u003Cbr/>• Dépendance fournisseur\u003Cbr/>• Coûts imprévisibles\u003Cbr/>• Défis de conformité\u003Cbr/>• Dépendances API\"]\n",[112,5627,5628],{"class":114,"line":144},[112,5629,4546],{},[112,5631,5632],{"class":114,"line":150},[112,5633,5634],{}," CloudCosts[\"Modèle de Coûts\u003Cbr/>• Paiement par token\u003Cbr/>• 2 000-10 000€/mois\u003Cbr/>• Mise à l'échelle variable\u003Cbr/>• Pas d'investissement initial\"]\n",[112,5636,5637],{"class":114,"line":156},[112,5638,480],{},[112,5640,5641],{"class":114,"line":162},[112,5642,485],{},[112,5644,5645],{"class":114,"line":168},[112,5646,5647],{}," subgraph OnPrem [\"Agents MCP On-Premises\"]\n",[112,5649,5650],{"class":114,"line":174},[112,5651,5652],{}," OnPremAdvantages[\"Bénéfices Sécurité\u003Cbr/>• Souveraineté totale des données\u003Cbr/>• Contrôle total des audits\u003Cbr/>• Conforme aux réglementations\u003Cbr/>• Sans dépendance fournisseur\u003Cbr/>• Fonctionnement hors ligne\"]\n",[112,5654,5655],{"class":114,"line":180},[112,5656,4546],{},[112,5658,5659],{"class":114,"line":186},[112,5660,5661],{}," OnPremChallenges[\"Défis d'Implémentation\u003Cbr/>• Investissement matériel requis\u003Cbr/>• Limites de performance modèles\u003Cbr/>• Complexité opérationnelle\u003Cbr/>• Contraintes d'échelle\u003Cbr/>• Mises à jour manuelles\"]\n",[112,5663,5664],{"class":114,"line":192},[112,5665,4546],{},[112,5667,5668],{"class":114,"line":198},[112,5669,5670],{}," OnPremCosts[\"Structure de Coûts\u003Cbr/>• 10 000-50 000€ initial\u003Cbr/>• Rentabilité 6-18 mois\u003Cbr/>• Coûts opérationnels fixes\u003Cbr/>• Mise à l'échelle prévisible\"]\n",[112,5672,5673],{"class":114,"line":204},[112,5674,480],{},[112,5676,5677],{"class":114,"line":210},[112,5678,485],{},[112,5680,5681],{"class":114,"line":216},[112,5682,5683],{}," subgraph Decision [\"Facteurs de Décision\"]\n",[112,5685,5686],{"class":114,"line":222},[112,5687,5688],{}," DataSensitivity[\"Sensibilité des Données\u003Cbr/>Haute sensibilité → On-Premises\u003Cbr/>Faible sensibilité → Cloud\"]\n",[112,5690,5691],{"class":114,"line":228},[112,5692,4546],{},[112,5694,5695],{"class":114,"line":234},[112,5696,5697],{}," Compliance[\"Exigences de Conformité\u003Cbr/>Réglementation stricte → On-Premises\u003Cbr/>Conformité standard → Cloud\"]\n",[112,5699,5700],{"class":114,"line":240},[112,5701,4546],{},[112,5703,5704],{"class":114,"line":246},[112,5705,5706],{}," TechnicalCapacity[\"Ressources Techniques\u003Cbr/>Équipe IA/ML forte → On-Premises\u003Cbr/>Ressources limitées → Cloud\"]\n",[112,5708,5709],{"class":114,"line":252},[112,5710,4546],{},[112,5712,5713],{"class":114,"line":258},[112,5714,5715],{}," CostModel[\"Préférences de Coûts\u003Cbr/>Coûts prévisibles → On-Premises\u003Cbr/>Coûts variables → Cloud\"]\n",[112,5717,5718],{"class":114,"line":264},[112,5719,480],{},[112,5721,5722],{"class":114,"line":270},[112,5723,485],{},[112,5725,5726],{"class":114,"line":276},[112,5727,5728],{}," subgraph Hybrid [\"Architecture Hybride\"]\n",[112,5730,5731],{"class":114,"line":282},[112,5732,5733],{}," HybridBenefits[\"Combinaison Stratégique\u003Cbr/>• Router données sensibles en local\u003Cbr/>• Utiliser cloud pour tâches complexes\u003Cbr/>• Optimiser coûts dynamiquement\u003Cbr/>• Distribuer risque opérationnel\"]\n",[112,5735,5736],{"class":114,"line":288},[112,5737,480],{},[112,5739,5740],{"class":114,"line":295},[112,5741,485],{},[112,5743,5744],{"class":114,"line":301},[112,5745,5746],{}," %% Flux de décision\n",[112,5748,5749],{"class":114,"line":307},[112,5750,5751],{}," DataSensitivity --> OnPremAdvantages\n",[112,5753,5754],{"class":114,"line":313},[112,5755,5756],{}," DataSensitivity --> CloudAdvantages\n",[112,5758,5759],{"class":114,"line":319},[112,5760,485],{},[112,5762,5763],{"class":114,"line":325},[112,5764,5765],{}," Compliance --> OnPremAdvantages\n",[112,5767,5768],{"class":114,"line":1758},[112,5769,5770],{}," Compliance --> CloudAdvantages\n",[112,5772,5773],{"class":114,"line":1764},[112,5774,485],{},[112,5776,5777],{"class":114,"line":1770},[112,5778,5779],{}," TechnicalCapacity --> OnPremChallenges\n",[112,5781,5782],{"class":114,"line":1776},[112,5783,5784],{}," TechnicalCapacity --> CloudRisks\n",[112,5786,5787],{"class":114,"line":1782},[112,5788,485],{},[112,5790,5791],{"class":114,"line":1788},[112,5792,5793],{}," CostModel --> OnPremCosts\n",[112,5795,5796],{"class":114,"line":1793},[112,5797,5798],{}," CostModel --> CloudCosts\n",[112,5800,5801],{"class":114,"line":1798},[112,5802,485],{},[112,5804,5805],{"class":114,"line":1804},[112,5806,5807],{}," %% Connexions hybrides\n",[112,5809,5810],{"class":114,"line":1810},[112,5811,5812],{}," OnPremAdvantages -.-> HybridBenefits\n",[112,5814,5815],{"class":114,"line":1816},[112,5816,5817],{}," CloudAdvantages -.-> HybridBenefits\n",[112,5819,5820],{"class":114,"line":1822},[112,5821,485],{},[112,5823,5824],{"class":114,"line":1828},[112,5825,5334],{},[112,5827,5828],{"class":114,"line":1833},[112,5829,5830],{}," classDef cloudStyle stroke:#1976d2,stroke-width:2px\n",[112,5832,5833],{"class":114,"line":1838},[112,5834,5835],{}," classDef onpremStyle stroke:#388e3c,stroke-width:2px\n",[112,5837,5838],{"class":114,"line":1844},[112,5839,5840],{}," classDef decisionStyle stroke:#f57c00,stroke-width:2px\n",[112,5842,5843],{"class":114,"line":1850},[112,5844,5845],{}," classDef hybridStyle stroke:#7b1fa2,stroke-width:2px\n",[112,5847,5848],{"class":114,"line":1856},[112,5849,485],{},[112,5851,5852],{"class":114,"line":1862},[112,5853,5854],{}," class Cloud,CloudAdvantages,CloudRisks,CloudCosts cloudStyle\n",[112,5856,5857],{"class":114,"line":1868},[112,5858,5859],{}," class OnPrem,OnPremAdvantages,OnPremChallenges,OnPremCosts onpremStyle\n",[112,5861,5862],{"class":114,"line":1874},[112,5863,5864],{}," class Decision,DataSensitivity,Compliance,TechnicalCapacity,CostModel decisionStyle\n",[112,5866,5867],{"class":114,"line":1880},[112,5868,5869],{}," class Hybrid,HybridBenefits hybridStyle\n",[809,5871,5873],{"id":5872},"analyse-des-coûts",[76,5874,5875],{},"Analyse des Coûts",[73,5877,5878,5508],{},[76,5879,5880],{},"Coûts Cloud (Estimés)",[363,5882,5883,5886,5889],{},[366,5884,5885],{},"Interactions complexes d'agents : 50-200 tokens par appel d'outil",[366,5887,5888],{},"Usage entreprise : 10,000+ interactions d'agents quotidiennes",[366,5890,5891],{},"Coûts mensuels : $2,000-$10,000+ selon le modèle et l'usage",[73,5893,5894,5508],{},[76,5895,5896],{},"Coûts On-Premises",[363,5898,5899,5902,5905],{},[366,5900,5901],{},"Matériel : $10,000-$50,000 d'investissement initial",[366,5903,5904],{},"Maintenance : Frais opérationnels continus",[366,5906,5907],{},"Point d'équilibre : Généralement 6-18 mois selon l'usage",[73,5909,5910,5913],{},[76,5911,5912],{},"Approche Hybride"," : Utilisez l'on-premises pour les données sensibles, le cloud pour les pics de charge ou les tâches spécialisées.",[809,5915,5917],{"id":5916},"sécurité-et-conformité",[76,5918,5919],{},"Sécurité et Conformité",[73,5921,5922],{},"L'on-premises offre des avantages significatifs :",[363,5924,5925,5931,5937,5943],{},[366,5926,5927,5930],{},[76,5928,5929],{},"Souveraineté des Données"," : Tout le traitement se fait dans votre infrastructure",[366,5932,5933,5936],{},[76,5934,5935],{},"Pistes d'Audit"," : Visibilité complète sur les actions des agents et les flux de données",[366,5938,5939,5942],{},[76,5940,5941],{},"Conformité"," : Plus facile de respecter GDPR, HIPAA, SOC2",[366,5944,5945,5948],{},[76,5946,5947],{},"Sécurité Personnalisée"," : Intégration avec l'infrastructure de sécurité existante",[73,5950,5951],{},"Mais introduit aussi des responsabilités :",[363,5953,5954,5960,5966],{},[366,5955,5956,5959],{},[76,5957,5958],{},"Sécurité des Modèles"," : S'assurer que les modèles ne sont pas compromis ou biaisés",[366,5961,5962,5965],{},[76,5963,5964],{},"Sécurité de l'Infrastructure"," : Protéger l'infrastructure IA elle-même",[366,5967,5968,5971],{},[76,5969,5970],{},"Contrôle d'Accès"," : Gérer qui peut déployer et modifier les agents",[809,5973,5975],{"id":5974},"maturité-opérationnelle",[76,5976,5977],{},"Maturité Opérationnelle",[73,5979,5980],{},"L'exécution d'IA on-premises nécessite des capacités organisationnelles :",[363,5982,5983,5989,5994,6000],{},[366,5984,5985,5988],{},[76,5986,5987],{},"DevOps pour l'IA"," : Pipelines CI/CD pour le déploiement des modèles",[366,5990,5991,5993],{},[76,5992,5562],{}," : Compréhension des métriques spécifiques à l'IA et des modes de défaillance",[366,5995,5996,5999],{},[76,5997,5998],{},"Mise à l'Échelle"," : Gestion des ressources à mesure que l'utilisation des agents croît",[366,6001,6002,6004],{},[76,6003,5568],{}," : Maintenir les modèles et l'infrastructure à jour",[68,6006,6008,6012],{"className":6007},[332,333,334],[89,6009,6011],{"id":6010},"stratégie-darchitecture-hybride","Stratégie d'Architecture Hybride",[73,6013,6014],{},[344,6015,6016],{"href":346},[76,6017,349],{},[68,6019,6021,6024,6030,6162,6168,6200],{"className":6020},[71],[73,6022,6023],{},"L'approche la plus pragmatique combine souvent les déploiements cloud et on-premises. Une architecture hybride permet aux organisations d'optimiser à la fois pour la sécurité et la capacité tout en maintenant la flexibilité opérationnelle.",[809,6025,6027],{"id":6026},"implémentation-du-routage-intelligent",[76,6028,6029],{},"Implémentation du Routage Intelligent",[102,6031,6033],{"className":2064,"code":6032,"language":2066,"meta":107,"style":107},"const agent = new Agent({\n // Local pour les opérations sensibles\n localProvider: \"http://localhost:1234/v1\",\n localModel: \"qwen2.5-32b-instruct\",\n \n // Cloud pour les tâches complexes\n cloudProvider: \"nebius\",\n cloudModel: \"Qwen/Qwen2.5-72B-Instruct\",\n \n // Routage basé sur la sensibilité des tâches\n routingStrategy: \"data-classification\"\n});\n",[109,6034,6035,6053,6058,6073,6089,6094,6099,6115,6131,6135,6140,6154],{"__ignoreMap":107},[112,6036,6037,6039,6042,6044,6046,6049,6051],{"class":114,"line":115},[112,6038,2080],{"class":2079},[112,6040,6041],{"class":2083}," agent ",[112,6043,2087],{"class":843},[112,6045,4967],{"class":843},[112,6047,6048],{"class":3862}," Agent",[112,6050,4999],{"class":2083},[112,6052,3265],{"class":843},[112,6054,6055],{"class":114,"line":121},[112,6056,6057],{"class":2073}," // Local pour les opérations sensibles\n",[112,6059,6060,6063,6065,6067,6069,6071],{"class":114,"line":127},[112,6061,6062],{"class":839}," localProvider",[112,6064,844],{"class":843},[112,6066,2100],{"class":843},[112,6068,5013],{"class":847},[112,6070,2106],{"class":843},[112,6072,2109],{"class":843},[112,6074,6075,6078,6080,6082,6085,6087],{"class":114,"line":132},[112,6076,6077],{"class":839}," localModel",[112,6079,844],{"class":843},[112,6081,2100],{"class":843},[112,6083,6084],{"class":847},"qwen2.5-32b-instruct",[112,6086,2106],{"class":843},[112,6088,2109],{"class":843},[112,6090,6091],{"class":114,"line":138},[112,6092,6093],{"class":2083}," \n",[112,6095,6096],{"class":114,"line":144},[112,6097,6098],{"class":2073}," // Cloud pour les tâches complexes\n",[112,6100,6101,6104,6106,6108,6111,6113],{"class":114,"line":150},[112,6102,6103],{"class":839}," cloudProvider",[112,6105,844],{"class":843},[112,6107,2100],{"class":843},[112,6109,6110],{"class":847},"nebius",[112,6112,2106],{"class":843},[112,6114,2109],{"class":843},[112,6116,6117,6120,6122,6124,6127,6129],{"class":114,"line":156},[112,6118,6119],{"class":839}," cloudModel",[112,6121,844],{"class":843},[112,6123,2100],{"class":843},[112,6125,6126],{"class":847},"Qwen/Qwen2.5-72B-Instruct",[112,6128,2106],{"class":843},[112,6130,2109],{"class":843},[112,6132,6133],{"class":114,"line":162},[112,6134,6093],{"class":2083},[112,6136,6137],{"class":114,"line":168},[112,6138,6139],{"class":2073}," // Routage basé sur la sensibilité des tâches\n",[112,6141,6142,6145,6147,6149,6152],{"class":114,"line":174},[112,6143,6144],{"class":839}," routingStrategy",[112,6146,844],{"class":843},[112,6148,2100],{"class":843},[112,6150,6151],{"class":847},"data-classification",[112,6153,3445],{"class":843},[112,6155,6156,6158,6160],{"class":114,"line":180},[112,6157,5039],{"class":843},[112,6159,5042],{"class":2083},[112,6161,4976],{"class":843},[809,6163,6165],{"id":6164},"avantages-hybrides",[76,6166,6167],{},"Avantages Hybrides",[363,6169,6170,6176,6182,6188,6194],{},[366,6171,6172,6175],{},[76,6173,6174],{},"Classification des Données"," : Route automatiquement les données sensibles vers le traitement local",[366,6177,6178,6181],{},[76,6179,6180],{},"Optimisation des Performances"," : Utilise les ressources cloud pour les tâches intensives en calcul",[366,6183,6184,6187],{},[76,6185,6186],{},"Gestion des Coûts"," : Équilibre les coûts fixes on-premises avec l'usage variable du cloud",[366,6189,6190,6193],{},[76,6191,6192],{},"Distribution des Risques"," : Évite les points uniques de défaillance dans chaque modèle de déploiement",[366,6195,6196,6199],{},[76,6197,6198],{},"Migration Progressive"," : Commence en local et étend l'usage cloud à mesure que la confiance augmente",[68,6201,6203,6207],{"className":6202},[332,333,334],[89,6204,6206],{"id":6205},"conclusion-et-prochaines-étapes","Conclusion et Prochaines Étapes",[73,6208,6209],{},[344,6210,6211],{"href":346},[76,6212,349],{},[68,6214,6216,6219,6223,6255,6261,6266,6280,6285,6299,6305,6342],{"className":6215},[71],[73,6217,6218],{},"L'exploration des Agents MCP Tiny On-Premises révèle une vérité convaincante : l'élégance du concept d'\"agent en 50 lignes\" de Hugging Face n'est pas diminuée par le déploiement local—elle est améliorée par les avantages de contrôle et de sécurité que fournit l'infrastructure on-premises.",[809,6220,6221],{"id":4323},[76,6222,4324],{},[1433,6224,6225,6231,6237,6243,6249],{},[366,6226,6227,6230],{},[76,6228,6229],{},"Faisabilité Technique"," : L'architecture de l'agent reste identique—seul l'endpoint d'inférence change",[366,6232,6233,6236],{},[76,6234,6235],{},"Puissance du Protocole MCP"," : Vos investissements en outils sont entièrement portables entre cloud et on-premises",[366,6238,6239,6242],{},[76,6240,6241],{},"Avantages Stratégiques"," : Les déploiements on-premises offrent souveraineté des données, avantages de conformité et prévisibilité des coûts",[366,6244,6245,6248],{},[76,6246,6247],{},"Réalité d'Implémentation"," : Les contraintes matérielles nécessitent une sélection soigneuse des modèles, mais des solutions capables existent",[366,6250,6251,6254],{},[76,6252,6253],{},"Optimisation Hybride"," : L'approche la plus pratique combine les deux modèles de déploiement selon la sensibilité des données",[809,6256,6258],{"id":6257},"cadre-de-décision",[76,6259,6260],{},"Cadre de Décision",[73,6262,6263,5508],{},[76,6264,6265],{},"Choisissez l'On-Premises Quand",[363,6267,6268,6271,6274,6277],{},[366,6269,6270],{},"La sensibilité des données est élevée (finance, santé, juridique)",[366,6272,6273],{},"Les exigences de conformité sont strictes (GDPR, HIPAA, SOC2)",[366,6275,6276],{},"Les coûts prévisibles sont préférés aux prix variables",[366,6278,6279],{},"Une équipe technique solide est disponible pour l'implémentation",[73,6281,6282,5508],{},[76,6283,6284],{},"Choisissez le Cloud Quand",[363,6286,6287,6290,6293,6296],{},[366,6288,6289],{},"La mise à l'échelle rapide est essentielle",[366,6291,6292],{},"Les dernières capacités des modèles sont requises",[366,6294,6295],{},"Les ressources techniques sont limitées",[366,6297,6298],{},"Les charges de travail variables rendent l'économie cloud favorable",[809,6300,6302],{"id":6301},"chemin-dimplémentation-recommandé",[76,6303,6304],{},"Chemin d'Implémentation Recommandé",[1433,6306,6307,6313,6319,6325,6330,6336],{},[366,6308,6309,6312],{},[76,6310,6311],{},"Phase d'Évaluation"," : Évaluez votre sensibilité aux données, besoins de conformité et capacités techniques",[366,6314,6315,6318],{},[76,6316,6317],{},"Déploiement Pilote"," : Commencez avec une petite configuration on-premises utilisant des modèles quantifiés (Qwen2.5-32B)",[366,6320,6321,6324],{},[76,6322,6323],{},"Évaluation des Performances"," : Comparez les performances locales vs. cloud pour vos cas d'usage spécifiques",[366,6326,6327,6329],{},[76,6328,5875],{}," : Calculez les points d'équilibre et le coût total de possession",[366,6331,6332,6335],{},[76,6333,6334],{},"Architecture Hybride"," : Concevez un routage intelligent basé sur la classification des données et la complexité des tâches",[366,6337,6338,6341],{},[76,6339,6340],{},"Mise à l'Échelle Progressive"," : Étendez les modèles réussis tout en maintenant les limites de sécurité",[73,6343,6344],{},"L'avenir de l'IA d'entreprise ne consiste pas à choisir entre la commodité du cloud et le contrôle on-premises—il s'agit d'architecturer des systèmes qui combinent intelligemment les deux approches. Les Agents MCP Tiny rendent cette vision pratique, fournissant la simplicité et la portabilité nécessaires pour des déploiements d'IA durables sur toute infrastructure.",[68,6346,6349,6355],{"className":6347},[1282,6348,1284,1285],"text-end",[73,6350,6351],{},[344,6352,6353],{"href":346},[76,6354,349],{},[68,6356,6358],{"className":6357},[6348],[73,6359,6360],{},[344,6361,6363],{"href":6362},"/contact",[76,6364,6365],{},"Prêt à explorer les agents IA on-premises ?",[1293,6367,6368],{},"html .light .shiki span {color: var(--shiki-light);background: var(--shiki-light-bg);font-style: var(--shiki-light-font-style);font-weight: var(--shiki-light-font-weight);text-decoration: var(--shiki-light-text-decoration);}html.light .shiki span {color: var(--shiki-light);background: var(--shiki-light-bg);font-style: var(--shiki-light-font-style);font-weight: var(--shiki-light-font-weight);text-decoration: var(--shiki-light-text-decoration);}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html pre.shiki code .spNyl, html code.shiki .spNyl{--shiki-light:#9C3EDA;--shiki-default:#C792EA;--shiki-dark:#C792EA}html pre.shiki code .sTEyZ, html code.shiki .sTEyZ{--shiki-light:#90A4AE;--shiki-default:#EEFFFF;--shiki-dark:#BABED8}html pre.shiki code .sMK4o, html code.shiki .sMK4o{--shiki-light:#39ADB5;--shiki-default:#89DDFF;--shiki-dark:#89DDFF}html pre.shiki code .s2Zo4, html code.shiki .s2Zo4{--shiki-light:#6182B8;--shiki-default:#82AAFF;--shiki-dark:#82AAFF}html pre.shiki code .swJcz, html code.shiki .swJcz{--shiki-light:#E53935;--shiki-default:#F07178;--shiki-dark:#F07178}html pre.shiki code .sfazB, html code.shiki .sfazB{--shiki-light:#91B859;--shiki-default:#C3E88D;--shiki-dark:#C3E88D}html pre.shiki code .sHwdD, html code.shiki .sHwdD{--shiki-light:#90A4AE;--shiki-light-font-style:italic;--shiki-default:#546E7A;--shiki-default-font-style:italic;--shiki-dark:#676E95;--shiki-dark-font-style:italic}",{"title":107,"searchDepth":121,"depth":121,"links":6370},[6371,6372,6373],{"id":4497,"depth":121,"text":4498},{"id":4815,"depth":121,"text":4816},{"id":4875,"depth":121,"text":4876,"children":6374},[6375,6376,6377,6378,6379,6380,6381,6382,6383,6384,6385],{"id":4891,"depth":127,"text":4894},{"id":4935,"depth":127,"text":4938},{"id":4944,"depth":127,"text":4947},{"id":5050,"depth":127,"text":5053},{"id":5591,"depth":127,"text":5594},{"id":5872,"depth":127,"text":5875},{"id":5916,"depth":127,"text":5919},{"id":5974,"depth":127,"text":5977},{"id":4323,"depth":127,"text":4324},{"id":6257,"depth":127,"text":6260},{"id":6301,"depth":127,"text":6304},"2025-06-28","Explorer comment exécuter des agents basés sur MCP entièrement on-premises en utilisant des LLMs locaux, en examinant les compromis entre la commodité du cloud et le contrôle local pour les déploiements d'IA en entreprise.",{"src":1315},{},{"title":27,"description":6387},"epMsG1Ni4vE1sze094ppzqcTx_LgZOojqXuq61W9IDg",1779118049364]