[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_196":197,"_198":199},"routes/_app.x.$handle.$name",{"_11":12,"_13":14,"_15":16,"_17":18,"_38":-7,"_121":-5,"_19":122,"_123":124,"_125":126,"_127":128,"_129":130,"_21":131,"_132":-5,"_133":-5,"_134":24,"_135":136,"_137":83,"_138":-5,"_139":-5,"_140":141},"routes/_app.x.$handle.$name.branch.$branch.history_.$",{"_11":12,"_13":14,"_15":16,"_17":18,"_19":20,"_21":22,"_23":24,"_25":26,"_27":28,"_29":24,"_30":-5},"handle","basrqtb","name","ai-chat-agent","branchName","main","path","index.html","project",{"_32":70,"_13":14,"_36":71,"_72":-5,"_73":24,"_74":75,"_76":75,"_77":78,"_79":-5,"_30":66,"_80":81,"_82":83,"_84":-5,"_85":-5,"_86":26,"_40":41,"_87":-5,"_88":-5,"_89":-5,"_90":-5,"_91":92,"_93":92,"_94":95,"_96":95,"_97":95,"_98":92,"_99":92,"_52":100,"_101":102,"_103":104,"_105":24},"file",{"_62":63,"_13":18,"_64":21,"_65":24,"_69":-5},"isDirectory",false,"page",0,"history",[31],"hasNext","userId",{"_32":33,"_34":35,"_36":37,"_38":39,"_40":41,"_42":43,"_44":-5,"_45":46,"_47":-5,"_48":-5,"_49":-5,"_50":-5,"_51":-5,"_52":53,"_21":54,"_55":-5,"_56":-5,"_57":-5,"_58":-5,"_59":60},"id","399f0b6e-8733-414e-81c5-fc6f14bebe7c","branchId","7a7a0053-e8f2-4964-a4a4-1db233b0b299","createdAt",["D",1784461658944],"version",3,"orgId","019f7a31-a489-7235-90c1-4a4d8a3b724c","parentId","b733a7ff-8c50-4692-83ed-b4ffc0dc1590","incomingCommitId","aiChatMessageId","acc548d8-5eed-42da-b018-43f5b5e05dfa","aiChatThreadId","aiChatThreadUserId","aiChatThreadTitle","source","clientLabel","user",{"_32":66,"_40":41,"_11":12,"_67":68},{"_62":63,"_13":18,"_64":21,"_65":24},"pull","merge","revert","multiple","commits",[61],{"_32":33,"_34":35,"_36":37,"_38":39,"_40":41,"_42":43,"_44":-5,"_45":46,"_47":-5,"_48":-5,"_49":-5,"_50":-5,"_51":-5,"_52":53,"_21":54,"_55":-5,"_56":-5,"_57":-5},"fileId","a487b20c-8367-11f1-94b0-1607ee4eb77e","fileType","deleted","8c7202a6-4188-4fc7-b215-4d5b8f57307f","avatar","https://lh3.googleusercontent.com/a/ACg8ocKZQdstIErW_XnlSP1xC4ueIqj3l1jSn7wnDZFoG0LRN0MNFQ=s96-c","deletedAtVersion","9266455e-67f2-4930-bd28-8a82ca5471c1",["D",1784461593494],"deletedAt","legacyEnvVars","privacy","public","httpPrivacy","description","Chat IA multi-agente con OpenRouter y Mistral","imageUrl","pinnedFileId","a80532f6-8367-11f1-992d-1607ee4eb77e","pinPreview",true,"pinMigrated","screenshotVersion","likeCount","tursoName","tursoHostname","tursoMeasuredAt","pinned","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm",{"_11":12,"_120":68},"org",{"_11":12,"_115":116,"_67":68,"_30":66,"_117":118,"_119":24},"branch",{"_38":106,"_13":16,"_32":35,"_107":70,"_108":-5,"_109":-5,"_110":-5,"_111":-5,"_112":-5,"_113":14,"_114":12},"isOwned",4,"projectId","forkedBranchId","forkedBranchVersion","forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","displayName","Basr Qtb","tier","free","canRestrictVals","profileImageUrl","adoptOrgId",{"_32":70,"_13":14,"_36":192,"_72":-5,"_73":24,"_74":75,"_76":75,"_77":78,"_79":-5,"_30":66,"_80":81,"_82":83,"_84":-5,"_85":-5,"_86":26,"_40":41,"_87":-5,"_88":-5,"_89":-5,"_90":-5,"_91":92,"_93":92,"_94":95,"_96":95,"_97":95,"_98":92,"_99":92,"_52":193,"_101":194,"_103":195,"_105":24},"tree",[161,162,163,164],"files",[161,162,163,164],"counts",{"_150":26,"_151":152,"_153":26,"_154":155,"_156":157},"contents",[],{"_32":142,"_13":18,"_62":63,"_42":-5,"_34":35,"_143":39,"_144":83,"_36":145,"_38":26,"_64":21,"_146":147,"_65":24,"_148":-5,"_17":149},"upstreamProject","upstreamBranch","emptyEnvVars","flags",[],"hasHttpFiles","grants","orgs","customDomains",{},"f375325b-107d-44a7-a48c-2a19225029ad","branchVersion","latest",["D",1784461658944],"content","\n\n\n \n \n Chat IA Multi-Agente\n \n\n\n

Chat IA Multi-Agente

Chat (OpenRouter) → Delegador (OpenRouter) → Sub-agentes Enanitos (Mistral)

\n \n \n

\n \n\n\n","label",[18],"pulls","branches",1,"forks","envVars",{"_158":26,"_160":26},"connections",{"_158":26,"_159":26},"count","unfilledCount","emptyCount",{"_32":185,"_13":186,"_62":187,"_42":-5,"_34":35,"_143":152,"_144":83,"_36":188,"_38":26,"_64":170,"_146":189,"_65":24,"_148":-5,"_172":190,"_17":191},{"_32":142,"_13":18,"_62":63,"_42":-5,"_34":35,"_143":39,"_144":83,"_36":182,"_38":26,"_64":21,"_146":147,"_65":24,"_148":-5,"_172":183,"_17":184},{"_32":175,"_13":176,"_62":81,"_42":-5,"_34":35,"_143":106,"_144":83,"_36":177,"_38":26,"_64":178,"_146":179,"_65":24,"_148":-5,"_172":180,"_17":181},{"_32":165,"_13":166,"_62":167,"_42":-5,"_34":35,"_143":168,"_144":83,"_36":169,"_38":26,"_64":170,"_146":171,"_65":24,"_148":-5,"_172":173,"_17":174},"bbe0ff1d-0578-4850-b80d-c3b200618c67","prompts.ts","a48038c4-8367-11f1-916c-1607ee4eb77e",2,["D",1784461658940],"script","// System prompts para los 3 niveles de agentes\n\nexport const CHAT_SYSTEM_PROMPT = `[REGLAS DE COMANDOS]\n1. Recordatorios -> Escribe exactamente: RECOR-DATORIO [motivo] a las [HH:MM] el día [día y mes]\n2. Alarmas -> Escribe exactamente: ALAR-MA \"[HH:MM]\"\n3. Macros (Comida + peso) -> Calcula sin decimales y responde exactamente con este formato:\n ( MAC-ROS\n carbohidratos: X\n proteinas: Y\n grasas: Z\n kcal: A )\n4. Imágenes -> Escribe exactamente: IM-AGEN [descripción]\n5. Rutas/Cómo llegar -> Escribe exactamente: CO-MOLL [ubicación]\n6. Si el usuario te pide enviar un email, responde únicamente con este formato, sin textos extra:\n CO-RREO |<>| Aquí el mensaje que quiere ser enviado\n7. Si el usuario te pide crear, generar o escribir un archivo (como JSON, HTML, TXT, etc.), responde únicamente con este formato, sin textos extra ni formato Markdown:\n AR-CHIVO |<>| Aquí va el código puro y limpio del archivo sin las comillas triples de Markdown |<>|nombre_del_archivo.extension|<>|\n8. Si el usuario te pide enviar un correo que incluya un archivo creado por ti o documento adjunto, responde únicamente con este formato, sin textos extra:\n CO-RREOAR |<>|Cuerpo o texto del mensaje|<>|nombre_del_archivo.extension|<>|\n9. Delegación (Tareas complejas/Múltiples acciones) -> Si el usuario te pide una tarea compleja que requiere usar varios comandos a la vez o varios pasos (como poner 2 alarmas a la vez, poner 2 recordatorios o mezclar funciones...), responde únicamente con este formato, redactando en el mensaje un prompt perfecto en primera persona (como si fueras el usuario pidiendo la tarea detallada paso a paso):\n AG-ENTE |<>| [Aquí el prompt optimizado y masticado paso a paso para el siguiente agente]\n10. Si te pide más de una función a la vez, ejecutarás el punto 9.`;\n\nexport const DELEGATOR_SYSTEM_PROMPT = `[REGLAS DE COMANDOS]\n1. Recordatorios -> Escribe exactamente: RECOR-DATORIO [motivo] a las [HH:MM] el día [día y mes]\n2. Alarmas -> Escribe exactamente: ALAR-MA \"[HH:MM]\"\n3. Macros (Comida + peso) -> Calcula sin decimales y responde exactamente con este formato:\n ( MAC-ROS\n carbohidratos: X\n proteinas: Y\n grasas: Z\n kcal: A )\n4. Imágenes -> Escribe exactamente: IM-AGEN [descripción]\n5. Rutas/Cómo llegar -> Escribe exactamente: CO-MOLL [ubicación]\n6. Si el usuario te pide enviar un email, responde únicamente con este formato, sin textos extra:\n CO-RREO |<>| Aquí el mensaje que quiere ser enviado\n7. Si el usuario te pide crear, generar o escribir un archivo (como JSON, HTML, TXT, etc.), responde únicamente con este formato, sin textos extra ni formato Markdown:\n AR-CHIVO |<>| Aquí va el código puro y limpio del archivo sin las comillas triples de Markdown |<>|nombre_del_archivo.extension|<>|\n8. Si el usuario te pide enviar un correo que incluya un archivo creado por ti o documento adjunto, responde únicamente con este formato, sin textos extra:\n CO-RREOAR |<>|Cuerpo o texto del mensaje|<>|nombre_del_archivo.extension|<>|\n\nReparte las tareas que te piden de esta manera, tu solo repartes las tareas (sean alarmas, recordatorios, crear archivos... y lo explicas).\nSiempre deberás intentar comprimir 3 tareas en cada, es decir, si solo hay 3 las meterás todas en el primero. Explícales bien y detalladamente lo que hacer, explícales independientemente todo teniendo en cuenta que no contienen toda la información.\n\nEN-ANITO1 |><| tarea 1, 2 y 3 |><|\nEN-ANITO2 |><| tarea 4, 5 y 6 |><|...\n\nEjemplo de respuesta: pon una alarma a las 6\nEN-ANITO1 |><| Alarma a las 6 |><|\n\nEn caso de que te haga falta una enanito más potente (como para crear un HTML complejo) usarás:\nEN-ANITOPO //\\\\ [tarea] siempre al final`;\n\nexport const SUBAGENT_SYSTEM_PROMPT = `Divide la tarea en 3 enanitos básicos de esta manera:\nEN-ANITO1 |<>| aquí el comando |<>|\nEN-ANITO2 |<>| aquí el comando |<>|\nEN-ANITO3 |<>| aquí el comando |<>|\nSi la tarea no necesita las 3, dejarás vacía las que no hagan falta.\n\n[REGLAS DE COMANDOS]\n1. Recordatorios -> Escribe exactamente: RECOR-DATORIO [motivo] a las [HH:MM] el día [día y mes]\n2. Alarmas -> Escribe exactamente: ALAR-MA \"[HH:MM]\"\n3. Macros (Comida + peso) -> Calcula sin decimales y responde exactamente con este formato:\n ( MAC-ROS\n carbohidratos: X\n proteinas: Y\n grasas: Z\n kcal: A )\n4. Imágenes -> Escribe exactamente: IM-AGEN [descripción]\n5. Rutas/Cómo llegar -> Escribe exactamente: CO-MOLL [ubicación]\n6. Si el usuario te pide enviar un email, responde únicamente con este formato, sin textos extra:\n CO-RREO |<>| Aquí el mensaje que quiere ser enviado\n7. Si el usuario te pide crear, generar o escribir un archivo (como JSON, HTML, TXT, etc.), responde únicamente con este formato, sin textos extra ni formato Markdown:\n AR-CHIVO |<>| Aquí va el código puro y limpio del archivo sin las comillas triples de Markdown |<>|nombre_del_archivo.extension|<>|\n8. Si el usuario te pide enviar un correo que incluya un archivo creado por ti o documento adjunto, responde únicamente con este formato, sin textos extra:\n CO-RREOAR |<>|Cuerpo o texto del mensaje|<>|nombre_del_archivo.extension|<>|\n\nInstrucción que hay que repartir:`;\n","children",[],[166],"c3d3b0e6-6839-42f5-b8ee-73e5680dea47","index.ts",["D",1784461664890],"http","import { serveFile } from \"https://esm.town/v/std/utils/index.ts\";\nimport { orchestrate } from \"./agents.ts\";\n\nexport default async function(req: Request): Promise {\n const url = new URL(req.url);\n\n // API: POST ?api=chat\n if (url.searchParams.get(\"api\") === \"chat\" && req.method === \"POST\") {\n try {\n const body = await req.json();\n const message = body.message;\n if (!message) {\n return Response.json({ error: \"Falta el campo 'message'\" }, { status: 400 });\n }\n\n const result = await orchestrate(message);\n\n return Response.json({ output: result.finalOutput });\n } catch (err) {\n return Response.json(\n { error: err.message || \"Error interno\" },\n { status: 500 },\n );\n }\n }\n\n // Servir la UI\n return await serveFile(req, \"index.html\");\n}\n",[],[176],["D",1784461658944],[],[18],"133b9bd5-e99f-4d74-abd6-ee0df6022592","agents.ts","a4789362-8367-11f1-b4c6-1607ee4eb77e",["D",1784461658933],"import {\n CHAT_SYSTEM_PROMPT,\n DELEGATOR_SYSTEM_PROMPT,\n SUBAGENT_SYSTEM_PROMPT,\n} from \"./prompts.ts\";\n\nconst OPENROUTER_URL = \"https://openrouter.ai/api/v1/chat/completions\";\nconst MISTRAL_URL = \"https://api.mistral.ai/v1/chat/completions\";\nconst OPENROUTER_MODEL = \"openai/gpt-4o-mini\";\nconst MISTRAL_MODEL = \"mistral-small-latest\";\n\n// Llama a la API de OpenRouter\nasync function callOpenRouter(\n systemPrompt: string,\n userMessage: string,\n): Promise {\n const apiKey = Deno.env.get(\"OPENROUTER_API_KEY\");\n if (!apiKey) throw new Error(\"Falta OPENROUTER_API_KEY\");\n\n const res = await fetch(OPENROUTER_URL, {\n method: \"POST\",\n headers: {\n \"Authorization\": `Bearer ${apiKey}`,\n \"Content-Type\": \"application/json\",\n },\n body: JSON.stringify({\n model: OPENROUTER_MODEL,\n messages: [\n { role: \"system\", content: systemPrompt },\n { role: \"user\", content: userMessage },\n ],\n }),\n });\n\n if (!res.ok) {\n const err = await res.text();\n throw new Error(`OpenRouter error ${res.status}: ${err}`);\n }\n\n const data = await res.json();\n return data.choices[0].message.content;\n}\n\n// Llama a la API de Mistral\nasync function callMistral(\n systemPrompt: string,\n userMessage: string,\n): Promise {\n const apiKey = Deno.env.get(\"MISTRAL_API_KEY\");\n if (!apiKey) throw new Error(\"Falta MISTRAL_API_KEY\");\n\n const res = await fetch(MISTRAL_URL, {\n method: \"POST\",\n headers: {\n \"Authorization\": `Bearer ${apiKey}`,\n \"Content-Type\": \"application/json\",\n },\n body: JSON.stringify({\n model: MISTRAL_MODEL,\n messages: [\n { role: \"system\", content: systemPrompt },\n { role: \"user\", content: userMessage },\n ],\n }),\n });\n\n if (!res.ok) {\n const err = await res.text();\n throw new Error(`Mistral error ${res.status}: ${err}`);\n }\n\n const data = await res.json();\n return data.choices[0].message.content;\n}\n\n// Extrae el prompt de delegación de la respuesta del chat\nfunction extractDelegationPrompt(chatResponse: string): string | null {\n const match = chatResponse.match(/AG-ENTE\\s*\\|<>\\|\\s*(.*)/s);\n return match ? match[1].trim() : null;\n}\n\n// Extrae las tareas de los enanitos de la respuesta del delegador\ninterface EnanitoTask {\n id: string;\n task: string;\n isPowerful: boolean;\n}\n\nfunction extractEnanitoTasks(delegatorResponse: string): EnanitoTask[] {\n const tasks: EnanitoTask[] = [];\n\n // EN-ANITO1 |><| tarea |><|, EN-ANITO2 |><| tarea |><|, etc.\n const normalRegex = /EN-ANITO(\\d+)\\s*\\|><\\|\\s*(.*?)\\s*\\|><\\|/gs;\n let match;\n while ((match = normalRegex.exec(delegatorResponse)) !== null) {\n tasks.push({\n id: match[1],\n task: match[2].trim(),\n isPowerful: false,\n });\n }\n\n // EN-ANITOPO //\\ [tarea] (enanito potente, siempre al final)\n const powerfulRegex = /EN-ANITOPO\\s*\\/\\/\\\\\\s*(.*?)(?:\\n|$)/s;\n const powerfulMatch = powerfulRegex.exec(delegatorResponse);\n if (powerfulMatch) {\n tasks.push({\n id: \"PO\",\n task: powerfulMatch[1].trim(),\n isPowerful: true,\n });\n }\n\n return tasks;\n}\n\n// Ejecuta un sub-agente (enanito) con Mistral\nasync function runSubAgent(task: EnanitoTask): Promise {\n const userMessage = `${SUBAGENT_SYSTEM_PROMPT}\\n\\n${task.task}`;\n const systemMessage = task.isPowerful\n ? \"Eres un enanito potente, especializado en tareas complejas como crear HTML, archivos elaborados, etc. Ejecuta la tarea con precisión.\"\n : \"Eres un enanito básico. Ejecuta las tareas asignadas siguiendo las reglas de comandos exactamente.\";\n\n return await callMistral(systemMessage, task.task);\n}\n\n// Orquesta todo el flujo: chat -> delegador -> sub-agentes\nexport async function orchestrate(userMessage: string): Promise<{\n chatResponse: string;\n delegatorResponse: string | null;\n subagentResponses: { id: string; response: string }[];\n finalOutput: string;\n}> {\n // 1. Chat normal (OpenRouter)\n const chatResponse = await callOpenRouter(CHAT_SYSTEM_PROMPT, userMessage);\n\n // 2. ¿Necesita delegación?\n const delegationPrompt = extractDelegationPrompt(chatResponse);\n\n if (!delegationPrompt) {\n // No hay delegación, devolver solo la respuesta del chat\n return {\n chatResponse,\n delegatorResponse: null,\n subagentResponses: [],\n finalOutput: chatResponse,\n };\n }\n\n // 3. Delegador (OpenRouter)\n const delegatorResponse = await callOpenRouter(\n DELEGATOR_SYSTEM_PROMPT,\n delegationPrompt,\n );\n\n // 4. Extraer tareas de enanitos\n const enanitoTasks = extractEnanitoTasks(delegatorResponse);\n\n // 5. Ejecutar sub-agentes en paralelo (Mistral)\n const subagentResults = await Promise.all(\n enanitoTasks.map(async (task) => ({\n id: task.id,\n response: await runSubAgent(task),\n })),\n );\n\n // 6. Componer la respuesta final\n let finalOutput = `Respuesta del chat\\n${chatResponse}\\n{~~}\\n`;\n\n finalOutput += `Respuesta del delegador\\n${delegatorResponse}\\n{~~}\\n`;\n\n finalOutput += \"Respuesta de los subagentes\\n\";\n for (const result of subagentResults) {\n finalOutput += `SUB${result.id}\\n${result.response}\\nSUB${result.id}\\n`;\n }\n\n return {\n chatResponse,\n delegatorResponse,\n subagentResponses: subagentResults,\n finalOutput,\n };\n}\n",[],[186],["D",1784461593494],{"_11":12,"_120":68},{"_11":12,"_115":116,"_67":68,"_30":66,"_117":118,"_119":24},{"_38":106,"_13":16,"_32":35,"_107":70,"_108":-5,"_109":-5,"_110":-5,"_111":-5,"_112":-5,"_113":14,"_114":12},"ctx",{"_52":-5,"_139":229,"_230":231,"_232":233,"_234":24,"_235":24,"_236":24,"_237":83,"_238":-5,"_239":-7,"_240":-7,"_241":26,"_242":24,"_243":24,"_244":-5,"_245":24,"_246":24,"_247":24,"_248":-5},"ENV",{"_200":201,"_202":203,"_204":16,"_205":206,"_207":208,"_209":210,"_211":212,"_213":214,"_215":216,"_217":218,"_219":220,"_221":222,"_223":224,"_225":226,"_227":228},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","RENDER_GIT_COMMIT","e87eb04e707cf64e7e532c8e8ec395f209e17f15","RENDER_GIT_BRANCH","POSTHOG_API_TOKEN","phc_nmoyuAkStEZuUp69rYlg1DZgrIt19Gx0Y4Hwf8tGgJP","POSTHOG_API_HOST","https://wart.val.town","WEB_API_TEMPLATE","https://{val}.web.val.run","APP_URL","https://www.val.town","MODULE_URL","https://esm.town","VANITY_WEB_API_TEMPLATE","https://{label}.val.run/","DENOLS_URL","https://lsp.val.town","NODE_ENV","production","CLOUDFLARE_TURNSTILE_SITE_KEY","0x4AAAAAACtaIZJH7s9rBMZ6","GOOGLE_CLIENT_ID","875989696899-jtsqkhlbdv5jrlsk8631p7ipjg3c6o62.apps.googleusercontent.com","STRIPE_PUBLISHABLE_KEY","pk_live_51N2zvWEpZTTLipIkVEwXrUG9AT0yeOOEYxszarZT1jpsZV58Ly6wlg3SltZpHN26CUalOBmUSY7adQRQiSkCZz2500qSXNkf8Y","API_URL","https://api.val.town",[],"preferencesKeybindings","default","preferencesTheme","light","preferencesFixers","preferencesCopilot","preferencesSelectionAi","preferencesCloseBrackets","preferencesChatPanelState","preferencesDprint","preferencesDenofmt","notificationCount","flagDenols","isBusinessOrgMember","businessOrgHandle","isBusinessOrgAdmin","isHigherTownieTier","navCollapsed","currentOrgId"]

History · index.html