[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8},"actionData","errors","root",{"_106":107,"_108":109},"routes/embed.x.$handle.$name.$",{"_9":10,"_11":12,"_13":14,"_15":16,"_17":18,"_19":20,"_21":-5},"handle","jeffvincent","name","mibbs-bot","path","llm.ts","project",{"_22":45,"_11":12,"_33":46,"_47":-5,"_48":42,"_49":50,"_51":50,"_52":53,"_54":-5,"_55":56,"_57":58,"_59":32,"_60":-5,"_61":-5,"_62":63,"_64":65,"_66":67,"_68":69,"_70":71,"_72":-5,"_73":74,"_75":74,"_76":77,"_78":77,"_79":77,"_80":74,"_81":74,"_82":83,"_84":85,"_86":87,"_88":42},"file",{"_22":23,"_11":14,"_24":25,"_26":-5,"_27":28,"_29":30,"_31":32,"_33":34,"_35":36,"_37":38,"_39":40,"_41":42,"_43":-5,"_13":44},"permalink","https://www.val.town/x/jeffvincent/mibbs-bot/code/llm.ts","extension","id","218fcda1-6cff-4e69-b7d8-3af030b5fe9e","fileId","e4a87a9e-ab92-11f1-bee4-1607ee4eb77e","parentId","branchId","05facaa4-90f3-48b5-bf13-4bf8572bc454","branchVersion",103,"latest",true,"createdAt",["D",1788970373165],"version",2,"fileType","script","content","import Anthropic from \"npm:@anthropic-ai/sdk\";\nimport OpenAI from \"npm:openai\";\n\nexport const MODEL = Deno.env.get(\"MODEL\") ?? \"claude-sonnet-4-6\";\nexport const anthropic = new Anthropic({ apiKey: Deno.env.get(\"ANTHROPIC_API_KEY\")! });\nexport const openai = new OpenAI({ apiKey: Deno.env.get(\"OPENAI_API_KEY\") ?? \"not-set\" });\n\nexport const isGptModel = (model: string) => model.startsWith(\"gpt-\");\n\n// \"low\" gives some reasoning for tool routing without medium/high latency; \"none\" disables it entirely\nexport const REASONING_EFFORT = \"low\" as const;\n\n// One-shot message (no tools, no history) — used by cron jobs and simple callers\nexport async function createMessage(opts: {\n prompt: string;\n system?: string;\n maxTokens?: number;\n}): Promise {\n const { prompt, system = \"\", maxTokens = 512 } = opts;\n\n if (isGptModel(MODEL)) {\n const params: any = {\n model: MODEL,\n input: [{ role: \"user\", content: prompt }],\n max_output_tokens: maxTokens,\n reasoning: { effort: REASONING_EFFORT },\n };\n if (system) params.instructions = system;\n\n const response = await (openai as any).responses.create(params);\n const msgItem = (response.output as any[]).find((o: any) => o.type === \"message\");\n return msgItem?.content?.find((c: any) => c.type === \"output_text\")?.text ?? response.output_text ?? \"\";\n }\n\n const response = await anthropic.messages.create({\n model: MODEL,\n max_tokens: maxTokens,\n ...(system ? { system } : {}),\n messages: [{ role: \"user\", content: prompt }],\n });\n return response.content.find((b: any) => b.type === \"text\")?.text ?? \"\";\n}\n\n// ── Agentic loop ───────────────────────────────────────────────────────────────\n\n// Provider-neutral content block types. History messages from the DB are plain\n// strings; the current user message may include an image block for vision.\nexport type AgentContentBlock =\n | { type: \"text\"; text: string }\n | { type: \"image\"; base64: string; mediaType: string };\n\nexport type AgentContent = string | AgentContentBlock[];\n\nexport interface AgentMessage {\n role: \"user\" | \"assistant\";\n content: AgentContent;\n}\n\n// Single dual-provider agentic tool-use loop. Tool definitions use Anthropic's\n// schema (the canonical format); the GPT-path conversion happens internally.\n// Each caller passes its own system prompt, messages, and executeTool callback.\nexport async function runAgent(opts: {\n system: string;\n messages: AgentMessage[];\n tools: Anthropic.Tool[];\n executeTool: (name: string, input: any) => Promise;\n maxTokens?: number;\n maxIterations?: number;\n onAssistantText?: (text: string) => Promise;\n}): Promise {\n const {\n system,\n messages,\n tools,\n executeTool,\n maxTokens = 1024,\n maxIterations = 10,\n onAssistantText,\n } = opts;\n\n if (isGptModel(MODEL)) {\n return _runAgentGpt({ system, messages, tools, executeTool, maxTokens, maxIterations, onAssistantText });\n }\n return _runAgentAnthropic({ system, messages, tools, executeTool, maxTokens, maxIterations, onAssistantText });\n}\n\n// ── GPT Responses API path ─────────────────────────────────────────────────────\n\nfunction _toGptContent(content: AgentContent): any {\n if (typeof content === \"string\") return content;\n return content.map((block) => {\n if (block.type === \"text\") return { type: \"input_text\", text: block.text };\n return { type: \"input_image\", image_url: `data:${block.mediaType};base64,${block.base64}` };\n });\n}\n\nasync function _runAgentGpt(opts: {\n system: string;\n messages: AgentMessage[];\n tools: Anthropic.Tool[];\n executeTool: (name: string, input: any) => Promise;\n maxTokens: number;\n maxIterations: number;\n onAssistantText?: (text: string) => Promise;\n}): Promise {\n const { system, messages, tools, executeTool, maxTokens, maxIterations, onAssistantText } = opts;\n\n const oaiTools = tools.map((t) => ({\n type: \"function\" as const,\n name: t.name,\n description: t.description,\n parameters: t.input_schema as any,\n }));\n\n const initialInput = messages.map((m) => ({\n role: m.role,\n content: _toGptContent(m.content),\n }));\n\n let responseId: string | null = null;\n let toolResults: any[] = [];\n\n for (let i = 0; i < maxIterations; i++) {\n const params: any = {\n model: MODEL,\n input: i === 0 ? initialInput : toolResults,\n tools: oaiTools,\n max_output_tokens: maxTokens,\n reasoning: { effort: REASONING_EFFORT },\n };\n if (i === 0) {\n params.instructions = system;\n } else {\n params.previous_response_id = responseId;\n }\n\n const response = await (openai as any).responses.create(params);\n responseId = response.id;\n\n const toolCalls = (response.output as any[]).filter((o: any) => o.type === \"function_call\");\n\n if (!toolCalls.length) {\n const msgItem = (response.output as any[]).find((o: any) => o.type === \"message\");\n const reply = msgItem?.content?.find((c: any) => c.type === \"output_text\")?.text ?? response.output_text ?? \"Done.\";\n if (onAssistantText) await onAssistantText(reply);\n return reply;\n }\n\n toolResults = [];\n for (const tc of toolCalls) {\n const result = await executeTool(tc.name, JSON.parse(tc.arguments));\n toolResults.push({ type: \"function_call_output\", call_id: tc.call_id, output: result });\n }\n }\n\n return \"Sorry, I couldn't complete that.\";\n}\n\n// ── Anthropic path ─────────────────────────────────────────────────────────────\n\nfunction _toAnthropicContent(content: AgentContent): Anthropic.ContentBlockParam[] | string {\n if (typeof content === \"string\") return content;\n return content.map((block): Anthropic.ContentBlockParam => {\n if (block.type === \"text\") return { type: \"text\", text: block.text };\n return {\n type: \"image\",\n source: {\n type: \"base64\",\n media_type: block.mediaType as \"image/jpeg\" | \"image/png\" | \"image/gif\" | \"image/webp\",\n data: block.base64,\n },\n };\n });\n}\n\nasync function _runAgentAnthropic(opts: {\n system: string;\n messages: AgentMessage[];\n tools: Anthropic.Tool[];\n executeTool: (name: string, input: any) => Promise;\n maxTokens: number;\n maxIterations: number;\n onAssistantText?: (text: string) => Promise;\n}): Promise {\n const { system, messages, tools, executeTool, maxTokens, maxIterations, onAssistantText } = opts;\n\n const anthropicMessages: Anthropic.MessageParam[] = messages.map((m) => ({\n role: m.role,\n content: _toAnthropicContent(m.content) as any,\n }));\n\n for (let i = 0; i < maxIterations; i++) {\n const response = await anthropic.messages.create({\n model: MODEL,\n max_tokens: maxTokens,\n system,\n tools,\n messages: anthropicMessages,\n });\n\n if (response.stop_reason === \"end_turn\") {\n for (const block of response.content) {\n if (block.type === \"text\") {\n if (onAssistantText) await onAssistantText(block.text);\n return block.text;\n }\n }\n return \"Done.\";\n }\n\n if (response.stop_reason === \"tool_use\") {\n anthropicMessages.push({ role: \"assistant\", content: response.content });\n const results: Anthropic.ToolResultBlockParam[] = [];\n for (const block of response.content) {\n if (block.type === \"tool_use\") {\n const result = await executeTool(block.name, block.input as any);\n results.push({ type: \"tool_result\", tool_use_id: block.id, content: result });\n }\n }\n anthropicMessages.push({ role: \"user\", content: results });\n } else {\n break;\n }\n }\n\n return \"Sorry, I couldn't complete that.\";\n}\n","deleted",false,"label",[14],"ea90a44b-e5c3-4abb-8fb5-1fdf1a363f60",["D",1788733325122],"deletedAt","legacyEnvVars","privacy","public","httpPrivacy","description","mibbs bot code","imageUrl","userId","e8e1fec0-46e7-11f0-a801-9b9c2b15cc6a","pinnedFileId","63f1ffa8-aa41-11f1-aa27-1607ee4eb77e","pinPreview","pinMigrated","screenshotVersion","likeCount",0,"orgId","7a1a376d-d8d2-47ee-b61d-1016a5b7e33d","tursoName","ea90a44be5c34abb8fb51fdf1a363f60","tursoHostname","ea90a44be5c34abb8fb51fdf1a363f60-val-town.aws-us-east-2.turso.io","tursoMeasuredAt",["D",1789522252250],"pinned","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm","user",{"_9":10,"_105":101},"org",{"_9":10,"_99":-5,"_100":101,"_55":56,"_102":103,"_104":42},"branch",{"_35":89,"_11":90,"_22":28,"_91":45,"_92":-5,"_93":-5,"_94":-5,"_95":-5,"_96":-5,"_97":12,"_98":10},"isOwned",130,"main","projectId","forkedBranchId","forkedBranchVersion","forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","displayName","avatar","https://img.clerk.com/eyJ0eXBlIjoicHJveHkiLCJzcmMiOiJodHRwczovL2ltYWdlcy5jbGVyay5kZXYvb2F1dGhfZ2l0aHViL2ltZ18yeU41V0tQaE1XdnpOME5pR05pMVFRNkh0bUIifQ","tier","free","canRestrictVals","profileImageUrl","ctx",{"_82":-5,"_139":140,"_141":142,"_143":144,"_145":42,"_146":42,"_147":42,"_148":32,"_149":-5,"_150":-7,"_151":-7,"_152":63,"_153":42,"_154":42,"_155":-5,"_156":42,"_157":42,"_158":42,"_159":-5},"ENV",{"_110":111,"_112":113,"_114":90,"_115":116,"_117":118,"_119":120,"_121":122,"_123":124,"_125":126,"_127":128,"_129":130,"_131":132,"_133":134,"_135":136,"_137":138},"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","orgs",[],"preferencesKeybindings","default","preferencesTheme","light","preferencesFixers","preferencesCopilot","preferencesSelectionAi","preferencesCloseBrackets","preferencesChatPanelState","preferencesDprint","preferencesDenofmt","notificationCount","flagDenols","isBusinessOrgMember","businessOrgHandle","isBusinessOrgAdmin","isHigherTownieTier","navCollapsed","currentOrgId"]