AI Biz Asst

[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_180":181,"_182":183},"routes/_app.x.$handle.$name",{"_99":98,"_32":45,"_107":85,"_27":108,"_13":14,"_109":-5,"_11":110,"_111":112,"_113":114,"_115":116,"_117":118,"_20":-5,"_119":120,"_121":122,"_123":25,"_124":125,"_126":57,"_127":-5,"_128":-5,"_129":130},"routes/_app.x.$handle.$name.branch.$branch.version.$version.diff",{"_11":12,"_13":14,"_15":16,"_17":-5,"_18":-5},"project",{"_43":44,"_32":45,"_38":46,"_47":-5,"_48":25,"_49":50,"_51":50,"_52":-5,"_53":-5,"_18":54,"_55":31,"_56":57,"_58":57,"_59":-5,"_60":61,"_62":63,"_64":-5,"_65":-5,"_66":-5,"_67":-5,"_68":69,"_70":69,"_71":72,"_73":72,"_74":72,"_75":69,"_76":69,"_77":78,"_79":80,"_81":82,"_83":25},"version",8,"changes",[19],"aiChatInfo","userId",{"_20":21,"_22":23,"_24":25,"_26":25,"_27":28,"_29":28},"file",{"_30":31,"_32":28,"_33":34,"_35":37,"_38":39,"_40":14,"_41":42,"_24":25},"previous",{"_30":31,"_32":28,"_33":34,"_35":36},"deleted",false,"deletedHere","path","Main.tsx","latestPath","fileId","add9a364-0e85-44f4-b7a6-eb8196e45ea7","name","fileType","http","content","import { fetch } from \"https://esm.town/v/std/fetch\";\nimport { OpenAI } from \"https://esm.town/v/std/openai\";\nimport { z } from \"npm:zod\";\n\n// --- Types (Minimal & Core) ---\ntype TskIn = { mid: string; tid: string; p: T };\ntype TskOut = { mid: string; cid: string; p: T; e?: string };\ntype Lvl = \"DB\" | \"IN\" | \"WN\" | \"ER\"; // Abbreviated log levels\ntype LogFn = (lvl: Lvl, comp: string, msg: string, dt?: any, mid?: string, tid?: string) => void;\ntype LogEntry = { ts: string; mid: string; tid?: string; lvl: Lvl; comp: string; msg: string; dt?: any };\ntype BaseIn = { bType?: string; sDesc?: string; tAud?: string; bTone?: string; goals?: string };\n\n// Task-specific inputs (reusing types from before, minimal change needed)\ntype CntIn = BaseIn & { cType: string; plat: string; specInstr?: string };\ntype StratIn = BaseIn & { plats: string[]; budgetLvl?: string; duration?: string };\ntype LogoIn = BaseIn & { bName: string; stylePrefs?: string; colorPrefs?: string };\ntype BrandIn = BaseIn & { bName: string; coreVals?: string };\ntype SimIn = BaseIn & { simType: string; specParams: Record | string };\ntype MetaIn = { pInput: string };\n\n// Consolidated User Input Schema (reusing schema, essential for validation)\nconst UserInSchema = z.object({\n task: z.enum([\n \"create_content\",\n \"develop_strategy\",\n \"generate_logo_concept\",\n \"build_brand_identity\",\n \"run_simulation\",\n \"meta_prompt\",\n ]),\n data: z.record(z.any()),\n}).refine((input) => {\n const d = input.data;\n const has = (key: string) => d && d[key] != null && d[key] !== \"\";\n const hasArr = (key: string) => Array.isArray(d?.[key]) && d[key].length > 0;\n switch (input.task) {\n case \"create_content\":\n return has(\"businessType\") && has(\"serviceDescription\") && has(\"targetAudience\") && has(\"contentType\")\n && has(\"platform\");\n case \"develop_strategy\":\n return has(\"businessType\") && has(\"serviceDescription\") && has(\"targetAudience\") && hasArr(\"platforms\");\n case \"generate_logo_concept\":\n return has(\"businessType\") && has(\"serviceDescription\") && has(\"targetAudience\") && has(\"businessName\");\n case \"build_brand_identity\":\n return has(\"businessType\") && has(\"serviceDescription\") && has(\"targetAudience\") && has(\"businessName\");\n case \"run_simulation\":\n return has(\"businessType\") && has(\"serviceDescription\") && has(\"targetAudience\") && has(\"simulationType\")\n && has(\"specificParams\") && (typeof d?.specificParams === \"object\" || typeof d?.specificParams === \"string\");\n case \"meta_prompt\":\n return has(\"prompt_input\");\n default:\n return false;\n }\n}, { message: \"Missing required fields for the selected task or invalid data format.\" });\ntype UserIn = z.infer;\n\n// Task-specific outputs (MODIFIED LogoOut)\ntype CntOut = { content: string; suggestions?: string };\ntype StratOut = { strategyMarkdown: string; advertisingIdeas?: string[]; audienceInsights?: string[] }; // Renamed fields for clarity\n// MODIFIED LogoOut for better JSON structure\ntype LogoConceptDetail = { title: string; description: string; aiPrompt: string };\ntype LogoOut = { concepts: LogoConceptDetail[]; svgExample?: string };\n\ntype BrandOut = {\n brandVoiceAndPersonality: string;\n keyMessages: string[];\n colorPaletteSuggestions: string; // Detailed string, potentially with hex and justification\n typographySuggestions: string; // Detailed string\n};\ntype SimOut = { simulationResult: string; analysis?: string };\ntype MetaOut = { systemPrompt: string };\n\ntype FinalRes = { res?: any; err?: string; details?: string; dbgLog: LogEntry[]; mid: string };\n\n// Questionnaire Types (reusing types)\ntype QState = {\n sid: string;\n task: UserIn[\"task\"];\n reqFlds: string[];\n answers: Partial>;\n qHist: { qid: string; field: string; text: string }[];\n isDone: boolean;\n};\ntype Quest = {\n id: string;\n field: string;\n text: string;\n type: \"text\" | \"textarea\" | \"select\";\n opts?: { lbl: string; val: string }[];\n};\ntype QResp = { sid: string; nextQ?: Quest; isDone: boolean; data?: UserIn; err?: string; state: QState | null };\n\n// --- Data Structure for Questionnaire Fields (Simplified) ---\nconst Q_FIELDS: {\n [key: string]: {\n txt: string;\n typ?: Quest[\"type\"];\n opts?: Quest[\"opts\"];\n opt?: true;\n tasks?: Array;\n ph?: string;\n map?: string;\n };\n} = {\n bType: { txt: \"Type of business?\", tasks: [\"base\"], ph: \"e.g., SaaS, Bakery\", map: \"businessType\" },\n sDesc: {\n txt: \"Describe products/services.\",\n typ: \"textarea\",\n tasks: [\"base\"],\n ph: \"Briefly explain what you offer\",\n map: \"serviceDescription\",\n },\n tAud: {\n txt: \"Ideal customer? (Specific)\",\n typ: \"textarea\",\n tasks: [\"base\"],\n ph: \"e.g., Small business owners in marketing\",\n map: \"targetAudience\",\n },\n bTone: {\n txt: \"Desired tone?\",\n typ: \"select\",\n opt: true,\n tasks: [\"base\"],\n opts: [\n { lbl: \"-- Tone --\", val: \"\" },\n { lbl: \"Professional\", val: \"professional\" },\n { lbl: \"Friendly\", val: \"friendly\" },\n { lbl: \"Casual\", val: \"casual\" },\n { lbl: \"Humorous\", val: \"humorous\" },\n { lbl: \"Informative\", val: \"informative\" },\n ],\n map: \"brandTone\",\n },\n goals: { txt: \"Main goals for this task?\", typ: \"textarea\", opt: true, tasks: [\"base\"], map: \"goals\" },\n bName: { txt: \"Exact business name?\", tasks: [\"generate_logo_concept\", \"build_brand_identity\"], map: \"businessName\" },\n cType: {\n txt: \"Content type needed?\",\n typ: \"select\",\n tasks: [\"create_content\"],\n opts: [\n { lbl: \"-- Type --\", val: \"\" },\n { lbl: \"Promo Post\", val: \"promo_post\" },\n { lbl: \"Blog Ideas\", val: \"blog_idea\" },\n { lbl: \"Blog Draft\", val: \"blog_draft\" },\n { lbl: \"Ad Copy\", val: \"ad_copy\" },\n ],\n map: \"contentType\",\n },\n plat: {\n txt: \"Platform for content?\",\n typ: \"select\",\n tasks: [\"create_content\"],\n opts: [\n { lbl: \"-- Platform --\", val: \"\" },\n { lbl: \"Generic\", val: \"generic\" },\n { lbl: \"Facebook\", val: \"facebook\" },\n { lbl: \"Instagram\", val: \"instagram\" },\n { lbl: \"LinkedIn\", val: \"linkedin\" },\n { lbl: \"Blog\", val: \"blog\" },\n ],\n map: \"platform\",\n },\n specInstr: {\n txt: \"Specific instructions/keywords?\",\n typ: \"textarea\",\n opt: true,\n tasks: [\"create_content\"],\n map: \"specificInstructions\",\n },\n plats: {\n txt: \"Strategy platforms? (comma separated)\",\n tasks: [\"develop_strategy\"],\n ph: \"e.g., LinkedIn, SEO\",\n map: \"platforms\",\n },\n budgetLvl: {\n txt: \"Approx budget level?\",\n typ: \"select\",\n opt: true,\n tasks: [\"develop_strategy\"],\n opts: [{ lbl: \"-- Budget --\", val: \"\" }, { lbl: \"Low (Organic)\", val: \"low\" }, {\n lbl: \"Medium (Mod Ads)\",\n val: \"medium\",\n }, { lbl: \"High (Sig Ads)\", val: \"high\" }],\n map: \"budgetLevel\",\n },\n duration: { txt: \"Strategy duration?\", opt: true, tasks: [\"develop_strategy\"], map: \"duration\" },\n stylePrefs: { txt: \"Logo style preferences?\", opt: true, tasks: [\"generate_logo_concept\"], map: \"stylePreferences\" },\n colorPrefs: { txt: \"Logo color preferences?\", opt: true, tasks: [\"generate_logo_concept\"], map: \"colorPreferences\" },\n coreVals: {\n txt: \"Core values/mission?\",\n typ: \"textarea\",\n opt: true,\n tasks: [\"build_brand_identity\"],\n map: \"coreValues\",\n },\n simType: {\n txt: \"Simulation type?\",\n typ: \"select\",\n tasks: [\"run_simulation\"],\n opts: [{ lbl: \"-- Sim --\", val: \"\" }, { lbl: \"Ad Response Est.\", val: \"ad_response_estimate\" }, {\n lbl: \"Customer Interaction\",\n val: \"customer_interaction\",\n }, { lbl: \"Other\", val: \"other\" }],\n map: \"simulationType\",\n },\n specParams: {\n txt: \"Sim details/params?\",\n typ: \"textarea\",\n tasks: [\"run_simulation\"],\n ph: \"Ad: audience, platform, offer. Cust: scenario, issue. Other: describe.\",\n map: \"specificParams\",\n },\n pInput: { txt: \"AI task desc or prompt to refine?\", typ: \"textarea\", tasks: [\"meta_prompt\"], map: \"prompt_input\" },\n};\n\n// --- LogAgent Class (Minor Abbreviation) ---\nclass LogAgent {\n logs: LogEntry[];\n midCtx?: string;\n constructor(mid?: string) {\n this.logs = [];\n this.midCtx = mid;\n }\n log(lvl: Lvl, comp: string, msg: string, dt?: any, mid?: string, tid?: string): void {\n const effMid = mid || this.midCtx || \"N/A\";\n let safeDt: any;\n if (dt != null) {\n try {\n safeDt = typeof structuredClone === \"function\"\n ? structuredClone(dt)\n : JSON.parse(JSON.stringify(dt, this.getCircReplacer()));\n } catch (e) {\n safeDt = `[Log Err: ${(e as Error).message}]`;\n }\n } else safeDt = dt;\n const entry: LogEntry = { ts: new Date().toISOString(), mid: effMid, tid, lvl, comp, msg, dt: safeDt };\n this.logs.push(entry);\n const dtStr = safeDt != null ? `| D: ${JSON.stringify(safeDt)}` : \"\";\n console.log(\n `[${entry.ts}] [${entry.mid}] [${entry.tid || \"---\"}] [${entry.lvl}] [${entry.comp}] ${entry.msg} ${dtStr}`,\n );\n }\n private getCircReplacer = () => {\n const s = new WeakSet();\n return (_k: string, v: any) => {\n if (typeof v === \"object\" && v !== null) { // Corrected 'o' to 'object'\n if (s.has(v)) return \"[C]\";\n s.add(v);\n }\n return v;\n };\n };\n getLogs(): LogEntry[] {\n try {\n return typeof structuredClone === \"function\"\n ? structuredClone(this.logs)\n : JSON.parse(JSON.stringify(this.logs, this.getCircReplacer()));\n } catch (e) {\n console.error(\"[LogAgent] Fail copy:\", e);\n return [...this.logs];\n }\n }\n clearLogs(): void {\n this.logs = [];\n }\n setMidCtx(mid: string): void {\n this.midCtx = mid;\n }\n}\n\n// --- OpenAI Call Wrapper (MODIFIED to enable JSON mode) ---\nasync function callOpenAI(\n sysP: string,\n userP: string,\n mid: string,\n tid: string,\n log: LogFn,\n): Promise {\n log(\"DB\", \"oai\", `Call tid=${tid}`, { spL: sysP.length, upL: userP.length }, mid, tid);\n try {\n const oai = new OpenAI();\n const comp = await oai.chat.completions.create({\n model: \"gpt-4o-mini\",\n messages: [{ role: \"system\", content: sysP }, { role: \"user\", content: userP }],\n temperature: 0.7,\n max_tokens: 2048,\n response_format: { type: \"json_object\" }, // Enable JSON mode\n });\n const resT = comp.choices[0]?.message?.content;\n const usage = comp.usage;\n if (!resT) {\n log(\"WN\", \"oai\", `No text tid=${tid}.`, { usage, fin: comp.choices[0]?.finish_reason }, mid, tid);\n return null;\n }\n log(\"IN\", \"oai\", `OK tid=${tid}`, { resL: resT.length, usage, fin: comp.choices[0]?.finish_reason }, mid, tid);\n return resT.trim(); // The result should be a JSON string\n } catch (err: any) {\n const d = { m: err.message, n: err.name, s: err.stack, t: err.type, c: err.code }; // Abbreviated error details\n log(\"ER\", \"oai\", `Fail tid=${tid}`, { e: d }, mid, tid);\n throw new Error(`API fail: ${err.message}` + (err.code ? ` (Code: ${err.code})` : \"\"));\n }\n}\n\n// --- QAgent Class (Refactored & Concise) ---\nclass QAgent {\n lg: LogFn;\n fDefs = Q_FIELDS;\n constructor(lgFn: LogFn) {\n this.lg = lgFn;\n }\n\n private getTaskFields(task: UserIn[\"task\"]) {\n const fields = Object.keys(this.fDefs).filter(k =>\n this.fDefs[k].tasks?.includes(task) || (this.fDefs[k].tasks?.includes(\"base\") && task !== \"meta_prompt\")\n );\n const req = fields.filter(k => !this.fDefs[k].opt);\n const opt = fields.filter(k => this.fDefs[k].opt);\n return { req, opt };\n }\n\n private isOptional(f: string): boolean {\n return this.fDefs[f]?.opt === true;\n }\n\n private selectNextField(req: string[], opt: string[], ans: Record): string | null {\n for (const f of req)\n if (!(f in ans) || (ans[f] === \"\" && !this.isOptional(f))) {\n this.lg(\"DB\", \"QA\", `Sel REQ: ${f}`, { ak: Object.keys(ans) });\n return f;\n }\n for (const f of opt)\n if (!(f in ans)) {\n this.lg(\"DB\", \"QA\", `Sel OPT: ${f}`, { ak: Object.keys(ans) });\n return f;\n }\n this.lg(\"DB\", \"QA\", `All fields done.`, { ak: Object.keys(ans) });\n return null;\n }\n\n private formatQuest(f: string, st: QState): Quest {\n const id = `${f}-${Date.now()}`;\n const fC = this.fDefs[f];\n let txt = fC?.txt || `Provide '${f}'.`;\n const typ = fC?.typ || \"text\";\n const opts = fC?.opts;\n const isOpt = fC?.opt || false;\n const bN = st?.answers?.bName || \"your business\";\n if (f === \"bType\") txt = `Type of business for ${bN}?`;\n if (f === \"sDesc\") txt = `Describe products/services for ${bN}.`;\n if (f === \"tAud\") txt = `Ideal customer for ${bN}? (Specific)`;\n if (f === \"coreVals\") txt = `Core values/mission of ${bN}?${isOpt ? \" (Optional)\" : \"\"}`;\n txt += isOpt && !txt.includes(\"(Optional)\") ? \" (Optional)\" : \"\";\n return { id, field: f, text: txt, type: typ, opts };\n }\n\n async nextStep(\n sid: string | null,\n task: UserIn[\"task\"],\n lastAns: { field: string; value: any } | null,\n curSt?: QState | null,\n ): Promise {\n let st: QState;\n const sessId = sid || `s-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`;\n const { req: reqFlds, opt: optFlds } = this.getTaskFields(task);\n const allFlds = [...reqFlds, ...optFlds];\n\n if (!curSt || !sid || curSt.sid !== sessId || curSt.task !== task) {\n this.lg(\"IN\", \"QA\", `Init/Reset state: ${task}`, { os: curSt?.sid, ot: curSt?.task, ns: sessId }, sessId);\n st = { sid: sessId, task, reqFlds: allFlds, answers: {}, qHist: [], isDone: false };\n if (lastAns?.field && allFlds.includes(lastAns.field)) {\n const v = typeof lastAns.value === \"string\" ? lastAns.value.trim() : lastAns.value;\n if (v !== \"\" || this.isOptional(lastAns.field)) st.answers[lastAns.field] = v;\n else this.lg(\"DB\", \"QA\", `Ign empty req init: ${lastAns.field}`, { v }, sessId);\n }\n } else {\n st = { ...curSt, answers: { ...curSt.answers }, qHist: [...curSt.qHist] };\n if (lastAns?.field && allFlds.includes(lastAns.field)) {\n this.lg(\"DB\", \"QA\", `Proc ans: ${lastAns.field}`, { v: lastAns.value }, st.sid);\n const v = typeof lastAns.value === \"string\" ? lastAns.value.trim() : lastAns.value;\n if (v !== \"\" || !reqFlds.includes(lastAns.field)) st.answers[lastAns.field] = v;\n else {\n this.lg(\"WN\", \"QA\", `Empty req ans: ${lastAns.field}. Re-ask.`, undefined, st.sid);\n delete st.answers[lastAns.field];\n }\n const lastQ = st.qHist[st.qHist.length - 1];\n if (lastQ && lastQ.field !== lastAns.field)\n this.lg(\"WN\", \"QA\", `Ans for ${lastAns.field}, last Q was ${lastQ?.field}. Sync issue?`, undefined, st.sid);\n }\n }\n\n const nextF = this.selectNextField(reqFlds, optFlds, st.answers);\n\n if (!nextF) {\n this.lg(\"IN\", \"QA\", `All fields for ${st.task}. Validate.`, { g: Object.keys(st.answers) }, st.sid);\n const finalD = this.finalizeData(st.answers, st.task, st.sid);\n try {\n UserInSchema.parse(finalD);\n this.lg(\"IN\", \"QA\", \"Valid OK.\", undefined, st.sid);\n st.isDone = true;\n return { sid: st.sid, isDone: true, data: finalD, state: st };\n } catch (validErr) {\n if (validErr instanceof z.ZodError) {\n const errMsgs = validErr.issues.map(i => `${i.path.join(\".\")}: ${i.message}`).join(\"; \");\n this.lg(\"ER\", \"QA\", \"Valid FAIL.\", { e: validErr.issues }, st.sid);\n st.isDone = false;\n const firstFailPath = validErr.issues[0]?.path;\n let fToReask: string | null = null;\n if (firstFailPath && firstFailPath.length > 1) {\n const schemaF = firstFailPath[1].toString();\n fToReask = Object.keys(this.fDefs).find(k => this.fDefs[k].map === schemaF) || null;\n if (fToReask && !allFlds.includes(fToReask)) fToReask = null;\n }\n if (fToReask) {\n delete st.answers[fToReask];\n const lastQIdx = st.qHist.findIndex(q => q.field === fToReask);\n if (lastQIdx !== -1) st.qHist.splice(lastQIdx, 1);\n const nextQ = this.formatQuest(fToReask, st);\n this.lg(\"WN\", \"QA\", `Valid fail on ${fToReask}. Re-ask.`, { e: errMsgs }, st.sid);\n st.qHist.push({ qid: nextQ.id, field: nextQ.field, text: nextQ.text });\n return {\n sid: st.sid,\n isDone: false,\n nextQ,\n state: st,\n err: `Check input for '${nextQ.text.replace(/\\s+\$Optional\$\\s*$/, \"\")}': ${errMsgs}`,\n };\n } else {\n this.lg(\"ER\", \"QA\", \"Valid fail, no field to re-ask.\", { e: errMsgs }, st.sid);\n return { sid: st.sid, isDone: false, err: `Valid fail: ${errMsgs}. Cannot proceed.`, state: st };\n }\n } else {\n this.lg(\"ER\", \"QA\", \"Unexpected valid error.\", { e: (validErr as Error).message }, st.sid);\n return {\n sid: st.sid,\n isDone: false,\n err: `Unexpected valid error: ${(validErr as Error).message}`,\n state: st,\n };\n }\n }\n } else {\n const nextQ = this.formatQuest(nextF, st);\n st.qHist.push({ qid: nextQ.id, field: nextQ.field, text: nextQ.text });\n this.lg(\"IN\", \"QA\", `Ask: ${nextF}`, { qT: nextQ.text }, st.sid);\n return { sid: st.sid, isDone: false, nextQ, state: st };\n }\n }\n\n async goBack(sid: string, task: UserIn[\"task\"], curSt: QState): Promise {\n this.lg(\"IN\", \"QA\", `Back action: ${task}`, { sessId: sid });\n const st = { ...curSt, answers: { ...curSt.answers }, qHist: [...curSt.qHist] };\n if (st.qHist.length < 1) {\n this.lg(\"WN\", \"QA\", \"Back fail: no history.\", undefined, st.sid);\n return { sid, isDone: false, err: \"Cannot go back further.\", state: st };\n }\n const lastQ = st.qHist.pop();\n if (lastQ && st.answers) {\n delete st.answers[lastQ.field];\n this.lg(\"DB\", \"QA\", `Back: removed ans for '${lastQ.field}'.`, { qid: lastQ.qid }, st.sid);\n }\n const QToReask = st.qHist[st.qHist.length - 1];\n if (QToReask) {\n const nextQ = this.formatQuest(QToReask.field, st);\n st.qHist[st.qHist.length - 1] = { qid: nextQ.id, field: nextQ.field, text: nextQ.text };\n this.lg(\"IN\", \"QA\", `Back to Q: ${QToReask.field}`, { qT: nextQ.text }, st.sid);\n return { sid, isDone: false, nextQ, state: st };\n } else {\n this.lg(\"WN\", \"QA\", \"Back fail: history empty after pop. Restarting.\", undefined, st.sid);\n return this.nextStep(sid, task, null, null);\n }\n }\n\n async restart(sid: string, task: UserIn[\"task\"]): Promise {\n this.lg(\"IN\", \"QA\", `Restart action: ${task}`, { sessId: sid });\n return this.nextStep(sid, task, null, null);\n }\n\n private finalizeData(gatheredAnswers: Record, task: UserIn[\"task\"], sid?: string): UserIn {\n let processedAnswers: Record = { ...gatheredAnswers };\n const { req: reqFlds, opt: optFlds } = this.getTaskFields(task);\n\n optFlds.forEach(internalKey => {\n if (\n processedAnswers[internalKey] == null\n || (typeof processedAnswers[internalKey] === \"string\" && processedAnswers[internalKey].trim() === \"\")\n ) {\n delete processedAnswers[internalKey];\n }\n });\n\n reqFlds.forEach(internalKey => {\n const fieldDef = this.fDefs[internalKey];\n const isFieldOptionalInDef = fieldDef?.opt === true;\n if (\n !isFieldOptionalInDef\n && (processedAnswers[internalKey] == null\n || (typeof processedAnswers[internalKey] === \"string\" && processedAnswers[internalKey].trim() === \"\"))\n ) {\n this.lg(\n \"WN\",\n \"finalize\",\n `Required internal field '${internalKey}' is empty/null before mapping. UserInSchema will validate mapped fields.`,\n { v: processedAnswers[internalKey] },\n sid,\n );\n }\n });\n\n const platsInternalKey = \"plats\";\n if (task === \"develop_strategy\" && processedAnswers[platsInternalKey] !== undefined) {\n if (typeof processedAnswers[platsInternalKey] === \"string\") {\n processedAnswers[platsInternalKey] = processedAnswers[platsInternalKey].split(\",\").map((p: string) => p.trim())\n .filter((p: string) => p.length > 0);\n } else if (!Array.isArray(processedAnswers[platsInternalKey])) {\n processedAnswers[platsInternalKey] = [];\n }\n processedAnswers[platsInternalKey] = (processedAnswers[platsInternalKey] as any[]).filter((p: any) =>\n typeof p === \"string\"\n );\n }\n\n const specParamsInternalKey = \"specParams\";\n if (task === \"run_simulation\" && processedAnswers[specParamsInternalKey] !== undefined) {\n const val = processedAnswers[specParamsInternalKey];\n if (typeof val === \"string\") {\n const pStr = val.trim();\n if (pStr.startsWith(\"{\") && pStr.endsWith(\"}\")) {\n try {\n const p = JSON.parse(pStr);\n processedAnswers[specParamsInternalKey] = typeof p === \"object\" && p !== null && !Array.isArray(p)\n ? p\n : { description: pStr };\n } catch (e) {\n processedAnswers[specParamsInternalKey] = { description: pStr };\n }\n } else {\n processedAnswers[specParamsInternalKey] = { description: pStr };\n }\n } else if (typeof val !== \"object\" || val === null || Array.isArray(val)) {\n try {\n processedAnswers[specParamsInternalKey] = { description: JSON.stringify(val) };\n } catch (e) {\n processedAnswers[specParamsInternalKey] = { description: \"[Error serializing specificParams]\" };\n }\n }\n }\n\n const schemaData: Record = {};\n for (const internalKey in processedAnswers) {\n if (Object.prototype.hasOwnProperty.call(processedAnswers, internalKey)) {\n const fieldDef = this.fDefs[internalKey];\n if (fieldDef && fieldDef.map) {\n schemaData[fieldDef.map] = processedAnswers[internalKey];\n } else if (fieldDef) {\n this.lg(\n \"WN\",\n \"finalize\",\n `Internal field '${internalKey}' defined in Q_FIELDS but has no 'map' property. It will NOT be added to schema data unless '${internalKey}' is a direct schema key.`,\n { val: processedAnswers[internalKey] },\n sid,\n );\n } else {\n this.lg(\n \"WN\",\n \"finalize\",\n `Field '${internalKey}' from answers is not defined in Q_FIELDS. It will be excluded from schema data.`,\n { val: processedAnswers[internalKey] },\n sid,\n );\n }\n }\n }\n this.lg(\"DB\", \"finalize\", \"Data mapped to schema keys for validation\", {\n fromAnswers: processedAnswers,\n toSchemaData: schemaData,\n }, sid);\n return { task, data: schemaData };\n }\n}\n\n// --- Generic Agent Execution Function & Configurations (MODIFIED) ---\ntype AgentConfig = {\n n: string; // Name\n p: (p: T_In) => { sys: string; user: string }; // Prompts\n ps: (rawJson: string) => T_Out; // Parse JSON output\n pp?: (out: T_Out, p: T_In) => T_Out; // Post-process\n};\n\nconst AGENT_CFGS: { [key in UserIn[\"task\"]]: AgentConfig } = {\n create_content: {\n n: \"Cnt\",\n p: (p: CntIn) => ({\n sys: `You are a helpful AI copywriter.\nContext: Business Type: ${p.bType || \"N/A\"}, Services: ${p.sDesc || \"N/A\"}, Target Audience: ${\n p.tAud || \"General Public\"\n }, Tone: ${p.bTone || \"Professional\"}, Goal: ${p.goals || \"Engage audience\"}.\nTask: Create content of type \"${p.cType}\" for platform \"${p.plat}\". Specific Instructions: ${p.specInstr || \"None\"}.\nPlatform Guidelines:\n- Social media posts (Facebook, Instagram, LinkedIn): Include relevant hashtags, calls to action, and aim for brevity and engagement.\n- Blog Ideas: Generate 3-5 distinct blog post titles or brief concepts.\n- Blog Draft: Write a 300-500 word draft in Markdown format.\n- Ad Copy: Create compelling ad copy with a clear headline, description, and call to action.\n- Generic: Create adaptable content suitable for general use.\n\nOutput ONLY a valid JSON object with the following structure:\n{\n \"content\": \"The generated content as a string. For 'Blog Ideas', this should be a list of titles/concepts separated by newlines.\",\n \"suggestions\": \"Optional brief suggestions for improving or using the content (e.g., 'Consider adding an image/video.', 'Research relevant keywords for blog posts.') as a string.\"\n}\nDo not include any explanatory text before or after the JSON object.`,\n user: `Generate the ${p.cType} based on the provided context.`,\n }),\n ps: (rawJson: string): CntOut => {\n const parsed = JSON.parse(rawJson);\n if (typeof parsed.content !== \"string\") {\n throw new Error(\"Parsed JSON for CntOut is missing or has invalid 'content' field.\");\n }\n // suggestions is optional\n return {\n content: parsed.content,\n suggestions: typeof parsed.suggestions === \"string\" ? parsed.suggestions : undefined,\n };\n },\n pp: (out, p) => { // Post-processing suggestions can still be useful\n let suggestions = out.suggestions || \"\";\n if (p.cType === \"promo_post\" || p.cType === \"ad_copy\") suggestions += \" Consider A/B testing different versions.\";\n else if (p.cType === \"blog_idea\" && (!suggestions || !suggestions.toLowerCase().includes(\"keyword\")))\n suggestions += \" Research keywords for each idea.\";\n else if (p.cType === \"blog_draft\") suggestions += \" Add internal/external links and a concluding paragraph.\";\n return { ...out, suggestions: suggestions.trim() || undefined };\n },\n },\n develop_strategy: {\n n: \"Strat\",\n p: (p: StratIn) => {\n const pS = p.plats?.join(\", \") || \"N/A\";\n return {\n sys: `You are an AI Marketing Strategist.\nContext: Business Type: ${p.bType || \"N/A\"}, Services: ${p.sDesc || \"N/A\"}, Target Audience: ${\n p.tAud || \"General Public\"\n }, Primary Goals: ${\n p.goals || \"Generate leads and increase brand awareness\"\n }, Key Platforms: ${pS}, Budget Level: ${p.budgetLvl || \"Low\"}, Strategy Duration: ${\n p.duration || \"Ongoing\"\n }, Desired Tone: ${p.bTone || \"Professional\"}.\n\nTask: Develop a concise marketing strategy.\n\nOutput ONLY a valid JSON object with the following structure:\n{\n \"strategyMarkdown\": \"A comprehensive strategy document in Markdown format. It should include sections like: Executive Summary, Target Audience Deep Dive (insights, behaviors, needs), Recommended Platforms (evaluation of ${pS} and any other suggestions, with justification), Content Themes & Key Messages (2-4 themes with example messages), Platform-Specific Tactics (2-3 actionable tactics per chosen platform, prioritized by budget), KPIs (3-5 key performance indicators like Traffic, Conversion Rate, Engagement Rate), and Next Steps.\",\n \"advertisingIdeas\": [\"A list of 2-3 distinct advertising ideas or campaign concepts as strings, if applicable based on budget and goals.\"],\n \"audienceInsights\": [\"A list of 2-3 key bullet-point insights about the target audience behavior or needs as strings.\"]\n}\nEnsure the 'strategyMarkdown' is well-structured. Do not include any explanatory text before or after the JSON object.`,\n user: `Develop the marketing strategy.`,\n };\n },\n ps: (rawJson: string): StratOut => {\n const parsed = JSON.parse(rawJson);\n if (typeof parsed.strategyMarkdown !== \"string\") {\n throw new Error(\"Parsed JSON for StratOut is missing or has invalid 'strategyMarkdown' field.\");\n }\n return {\n strategyMarkdown: parsed.strategyMarkdown,\n advertisingIdeas: Array.isArray(parsed.advertisingIdeas)\n ? parsed.advertisingIdeas.filter(s => typeof s === \"string\")\n : undefined,\n audienceInsights: Array.isArray(parsed.audienceInsights)\n ? parsed.audienceInsights.filter(s => typeof s === \"string\")\n : undefined,\n };\n },\n },\n generate_logo_concept: {\n n: \"Logo\",\n p: (p: LogoIn) => ({\n sys: `You are a Creative Design Assistant.\nContext: Business Name: \"${p.bName}\", Business Type: ${p.bType || \"N/A\"}, Services: ${\n p.sDesc || \"N/A\"\n }, Target Audience: ${p.tAud || \"General Public\"}, Style Preferences: ${\n p.stylePrefs || \"Versatile, modern\"\n }, Color Preferences: ${p.colorPrefs || \"Suggest based on business type and audience\"}, Desired Tone: ${\n p.bTone || \"Professional\"\n }.\n\nTask: Generate 2-3 distinct logo concepts. For each concept, provide a title, a detailed description (visuals, symbolism, font ideas, color palette considerations), and a detailed AI art generation prompt suitable for models like DALL-E or Midjourney. Optionally, provide a very basic SVG example if it's simple (e.g., an abstract shape or initial).\n\nOutput ONLY a valid JSON object with the following structure:\n{\n \"concepts\": [\n {\n \"title\": \"Concept Title 1 (e.g., Modern Initial Mark)\",\n \"description\": \"Detailed description of the concept, including visual elements, symbolism, potential font styles, and color palette ideas...\",\n \"aiPrompt\": \"A detailed AI art generation prompt for this concept. Example: 'Minimalist vector logo for [Business Name], letter [X], [style preferences e.g., sleek lines, geometric shapes], [color preferences e.g., blue and silver], white background, high resolution.'\"\n }\n // ... (include 1 or 2 more concept objects)\n ],\n \"svgExample\": \"Optional: A string containing very basic SVG code for one of the concepts, or leave empty if not applicable/too complex. Example: ''\"\n}\nDo not include any explanatory text before or after the JSON object.`,\n user: `Generate logo concepts for \"${p.bName}\".`,\n }),\n ps: (rawJson: string): LogoOut => {\n const parsed = JSON.parse(rawJson);\n if (\n !Array.isArray(parsed.concepts) || parsed.concepts.some(c =>\n typeof c.title !== \"string\" || typeof c.description !== \"string\" || typeof c.aiPrompt !== \"string\"\n )\n ) {\n throw new Error(\"Parsed JSON for LogoOut has invalid 'concepts' structure.\");\n }\n const concepts: LogoConceptDetail[] = parsed.concepts.map(c => ({\n title: c.title,\n description: c.description,\n aiPrompt: c.aiPrompt,\n }));\n return {\n concepts: concepts,\n svgExample: typeof parsed.svgExample === \"string\" && parsed.svgExample.trim() !== \"\"\n ? parsed.svgExample\n : undefined,\n };\n },\n },\n build_brand_identity: {\n n: \"Brand\",\n p: (p: BrandIn) => ({\n sys: `You are an AI Brand Strategist.\nContext: Business Name: \"${p.bName}\", Business Type: ${p.bType || \"N/A\"}, Services: ${\n p.sDesc || \"N/A\"\n }, Target Audience: ${p.tAud || \"General Public\"}, Core Values/Mission: ${p.coreVals || \"N/A\"}, Desired Tone: ${\n p.bTone || \"Professional\"\n }.\n\nTask: Define key elements of the brand identity.\n\nOutput ONLY a valid JSON object with the following structure:\n{\n \"brandVoiceAndPersonality\": \"A detailed description of the brand's voice, personality, and communication style. Should align with the desired tone: '${\n p.bTone || \"Professional\"\n }'.\",\n \"keyMessages\": [\n \"A list of 3-5 concise key messages (as strings) that communicate the brand's value proposition, targeted at the audience, and reflecting its core values.\"\n ],\n \"colorPaletteSuggestions\": \"A description of a suggested color palette, including primary, secondary, and accent colors. Mention example HEX codes and justify choices (e.g., 'Primary: #0A2A4E (Deep Blue) - conveys trust and professionalism. Secondary: #F1C40F (Gold) - adds a touch of quality and optimism.').\",\n \"typographySuggestions\": \"Suggestions for typography, including a headline font and a body text font (preferably from Google Fonts for web accessibility). Justify choices (e.g., 'Headline: Montserrat (Bold) - modern and strong. Body: Open Sans - highly readable and versatile.').\"\n}\nDo not include any explanatory text before or after the JSON object.`,\n user: `Develop brand identity guide for \"${p.bName}\".`,\n }),\n ps: (rawJson: string): BrandOut => {\n const parsed = JSON.parse(rawJson);\n if (\n typeof parsed.brandVoiceAndPersonality !== \"string\"\n || !Array.isArray(parsed.keyMessages) || parsed.keyMessages.some(m => typeof m !== \"string\")\n || typeof parsed.colorPaletteSuggestions !== \"string\"\n || typeof parsed.typographySuggestions !== \"string\"\n ) {\n throw new Error(\"Parsed JSON for BrandOut has invalid structure or missing fields.\");\n }\n return {\n brandVoiceAndPersonality: parsed.brandVoiceAndPersonality,\n keyMessages: parsed.keyMessages,\n colorPaletteSuggestions: parsed.colorPaletteSuggestions,\n typographySuggestions: parsed.typographySuggestions,\n };\n },\n },\n run_simulation: {\n n: \"Sim\",\n p: (p: SimIn) => {\n let pD = \"N/A\";\n try {\n pD = typeof p.specParams === \"string\" ? p.specParams : JSON.stringify(p.specParams);\n } catch (e) {\n pD = \"[Serialization failed for specificParams]\";\n }\n return {\n sys: `You are an AI Simulation Model.\nContext: Business Type: ${p.bType || \"N/A\"}, Services: ${p.sDesc || \"N/A\"}, Target Audience: ${\n p.tAud || \"General Public\"\n }, Desired Tone: ${p.bTone || \"Professional\"}.\nSimulation Task Type: \"${p.simType}\".\nSpecific Parameters/Scenario: ${pD}.\n\nTask: Execute the simulation based on the type and parameters.\n- 'ad_response_estimate': Qualitatively estimate potential ad effectiveness. Consider likelihoods, target audience alignment. Do not give financial guarantees.\n- 'customer_interaction': Role-play a customer interaction. Provide dialogue for both 'Customer:' and 'Agent:'. Aim for 3-5 exchanges.\n- 'other': Execute the simulation as described in the 'Specific Parameters'. Interpret the request and state any assumptions made.\n\nOutput ONLY a valid JSON object with the following structure:\n{\n \"simulationResult\": \"The direct output of the simulation. For 'customer_interaction', this should be the dialogue. For 'ad_response_estimate', a summary of the estimated response. For 'other', the result of the custom simulation.\",\n \"analysis\": \"Optional: A brief (2-4 sentences) analysis or key takeaways from the simulation outcome as a string.\"\n}\nDo not include any explanatory text before or after the JSON object.`,\n user: `Run simulation '${p.simType}'.`,\n };\n },\n ps: (rawJson: string): SimOut => {\n const parsed = JSON.parse(rawJson);\n if (typeof parsed.simulationResult !== \"string\") {\n throw new Error(\"Parsed JSON for SimOut is missing or has invalid 'simulationResult' field.\");\n }\n return {\n simulationResult: parsed.simulationResult,\n analysis: typeof parsed.analysis === \"string\" ? parsed.analysis : undefined,\n };\n },\n },\n meta_prompt: {\n n: \"Meta\",\n p: (p: MetaIn) => ({\n sys: `You are an AI Prompt Engineering Assistant.\nUser's Request for a System Prompt (this is the input you need to process):\n\\`\\`\\`\n${p.pInput}\n\\`\\`\\`\nTask: Based ONLY on the user's request above and general prompt engineering principles, generate or refine a SYSTEM prompt for an AI model like GPT-4o-mini.\nPrinciples to consider for the generated system prompt:\n1. Role: Clearly define the AI's role.\n2. Task/Goal: Specify the main task and desired outcome.\n3. Context/Input: Mention what context or inputs the AI will receive.\n4. Process: If the task is complex, outline steps.\n5. Output Format/Constraints: Define the desired output format (e.g., Markdown, JSON structure), length, tone, things to avoid. Provide examples if helpful.\n6. Ambiguity: Try to resolve potential ambiguities.\n7. Refinement: If refining an existing prompt, identify weaknesses and suggest improvements.\n\nOutput ONLY a valid JSON object with the following structure:\n{\n \"systemPrompt\": \"The complete generated or refined system prompt text as a string. Do not include markdown code fences (like \\`\\`\\`json) inside this string unless they are part of the prompt itself meant for the target AI.\"\n}\nDo not include any explanatory text before or after the JSON object.`,\n user: `Generate or refine a system prompt based on my request.`,\n }),\n ps: (rawJson: string): MetaOut => {\n const parsed = JSON.parse(rawJson);\n if (typeof parsed.systemPrompt !== \"string\" || parsed.systemPrompt.length < 10) { // Basic sanity check\n throw new Error(\n \"Parsed JSON for MetaOut is missing, has invalid 'systemPrompt' field, or prompt is too short.\",\n );\n }\n return {\n systemPrompt: parsed.systemPrompt,\n };\n },\n },\n};\n\nasync function runAgent(\n task: TskIn,\n log: LogFn,\n cfg: AgentConfig,\n): Promise> {\n const { mid, tid, p } = task;\n log(\"IN\", cfg.n, `Start tid=${tid}`, {}, mid, tid);\n try {\n const { sys, user } = cfg.p(p);\n const rawC = await callOpenAI(sys, user, mid, tid, log);\n if (!rawC) {\n log(\"WN\", cfg.n, `No LLM content tid=${tid}.`, {}, mid, tid);\n return { mid, cid: tid, p: {} as T_Out, e: `AI failed to gen ${cfg.n}.` };\n }\n let out: T_Out;\n try {\n out = cfg.ps(rawC); // rawC should be a JSON string\n } catch (e: any) {\n log(\"ER\", cfg.n, `Parse FAIL tid=${tid}. Message: ${e.message}`, { raw: rawC.slice(0, 500) }, mid, tid);\n return { mid, cid: tid, p: {} as T_Out, e: `AI output was unparseable or invalid structure: ${e.message}` };\n }\n if (cfg.pp) {\n try {\n out = cfg.pp(out, p);\n } catch (e: any) {\n log(\"ER\", cfg.n, `PP FAIL tid=${tid}.`, { e: e.message }, mid, tid);\n // Non-fatal, proceed with parsed output\n }\n }\n log(\"IN\", cfg.n, `OK tid=${tid}.`, {}, mid, tid);\n return { mid, cid: tid, p: out };\n } catch (err: any) {\n log(\"ER\", cfg.n, `Task FAIL tid=${tid}.`, { e: err.message }, mid, tid);\n return { mid, cid: tid, p: {} as T_Out, e: `${cfg.n} fail: ${err.message}` };\n }\n}\n\n// --- Orchestrator Class (Minor Abbreviation) ---\nclass Orchestrator {\n baseL: LogFn;\n mId: string;\n lg: LogFn;\n constructor(baseLgr: LogAgent) {\n this.mId = `m-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`;\n baseLgr.setMidCtx(this.mId);\n this.baseL = baseLgr.log.bind(baseLgr);\n this.lg = (lvl, comp, msg, dt?, _midIgn?, tid?) => {\n this.baseL(lvl, comp, msg, dt, this.mId, tid);\n };\n this.lg(\"IN\", \"Orch\", `Init mandate: ${this.mId}`);\n }\n async runWorkflow(input: UserIn): Promise {\n this.lg(\"IN\", \"Orch\", `Wf start: ${input.task}`, {}, this.mId);\n const tid = `${input.task.slice(0, 3)}-${Date.now()}-${Math.random().toString(36).slice(2, 7)}`;\n const tInput: TskIn = { mid: this.mId, tid: tid, p: input.data };\n let tOutput: TskOut;\n const aCfg = AGENT_CFGS[input.task];\n if (!aCfg) {\n this.lg(\"ER\", \"Orch\", `Unsup task: ${input.task}`, {}, this.mId, tid);\n return { mId: this.mId, err: `Unsup task: ${input.task}`, dbgLog: [] };\n }\n try {\n tOutput = await runAgent(tInput, this.lg, aCfg);\n if (tOutput.e) {\n this.lg(\"WN\", \"Orch\", `Agent ${tid} err.`, { e: tOutput.e }, this.mId);\n return { mid: this.mId, err: `Agent fail: ${tOutput.e}`, res: tOutput.p, dbgLog: [] };\n }\n this.lg(\"IN\", \"Orch\", `Wf task ${tid} OK.`, {}, this.mId);\n return { mid: this.mId, res: tOutput.p, dbgLog: [] };\n } catch (err: any) {\n this.lg(\"ER\", \"Orch\", `Wf critical fail ${tid}.`, { e: err.message }, this.mId);\n return { mid: this.mId, err: `Wf critical fail: ${err.message}`, dbgLog: [] };\n }\n }\n}\n\n// --- HTML Generation (Minor Abbreviation & CSS/JS reduction) ---\n// No changes to HTML/JS needed for this refactor, so it's kept as is.\n// It's long, so I'll use a comment to represent it to keep the response shorter.\n// Original HTML generation function generateOrthogonalUI() and its embedded JS remains here.\nfunction generateOrthogonalUI(): string {\n return `AI Biz Asst\n

\n \n \n \n \n

\n \n `;\n}\n\n// --- Main Request Handler (Concise) ---\nexport default async function(req: Request): Promise {\n const reqId = `r-${Date.now()}-${Math.random().toString(36).slice(2, 7)}`;\n const reqLgr = new LogAgent(reqId);\n const log = reqLgr.log.bind(reqLgr);\n const url = new URL(req.url);\n const headers = {\n \"Content-Type\": \"application/json\",\n \"Access-Control-Allow-Origin\": \"*\",\n \"Access-Control-Allow-Methods\": \"GET, POST, OPTIONS\",\n \"Access-Control-Allow-Headers\": \"Content-Type, Authorization\",\n };\n\n if (req.method === \"OPTIONS\") {\n log(\"IN\", \"H\", \"OPT\");\n return new Response(null, { status: 204, headers });\n }\n\n try {\n if (req.method === \"GET\" && (url.pathname === \"/\" || url.pathname === \"\")) {\n log(\"IN\", \"H\", \"GET /\");\n const html = generateOrthogonalUI();\n return new Response(html, { status: 200, headers: { ...headers, \"Content-Type\": \"text/html; charset=utf-8\" } });\n }\n\n if (req.method === \"POST\" && url.pathname === \"/questionnaire\") {\n log(\"IN\", \"H\", \"POST /q\");\n let b;\n let rawB = \"[No body]\";\n try {\n rawB = await req.text();\n b = JSON.parse(rawB);\n if (\n !b?.task || typeof b.task !== \"string\" || !b.action\n || ![\"start\", \"answer\", \"back\", \"restart\"].includes(b.action) || (b.action === \"answer\" && !b.lastAnswer)\n ) throw new Error(\"Bad req.\");\n } catch (e: any) {\n log(\"WN\", \"H\", \"Bad /q body\", { e: e.message, b: rawB.slice(0, 300) });\n return new Response(JSON.stringify({ err: `Bad Req: ${e.message}`, st: null, dbgL: reqLgr.getLogs() }), {\n status: 400,\n headers,\n });\n }\n\n const qA = new QAgent(log);\n let qRes: QResp;\n try {\n log(\"DB\", \"H\", `Proc: ${b.action}`, { t: b.task, s: b.sessionId, hasS: !!b.state, hasA: !!b.lastAnswer });\n switch (b.action) {\n case \"start\":\n qRes = await qA.nextStep(b.sessionId, b.task, b.lastAnswer, null);\n break;\n case \"answer\":\n qRes = await qA.nextStep(b.sessionId, b.task, b.lastAnswer, b.state);\n break;\n case \"back\":\n qRes = !b.state\n ? { sid: b.sessionId || `e-${Date.now()}`, isDone: false, err: \"No state to go back.\", state: null }\n : await qA.goBack(b.sessionId, b.task, b.state);\n break;\n case \"restart\":\n qRes = await qA.restart(b.sessionId, b.task);\n break;\n default: // Should not be reached due to initial validation\n qRes = {\n sid: b.sessionId || `e-${Date.now()}`,\n isDone: false,\n err: \"Invalid action.\",\n state: b.state || null,\n };\n }\n const respSt = qRes.state ? { ...qRes.state } : null;\n log(\"DB\", \"H\", \"Q step proc.\", {\n sid: qRes.sid,\n done: qRes.isDone,\n hasE: !!qRes.err,\n nextF: qRes.nextQ?.field,\n });\n return new Response(JSON.stringify({ ...qRes, state: respSt }), { status: qRes.err ? 400 : 200, headers });\n } catch (e: any) {\n log(\"ER\", \"H\", `QAgent proc err: ${b.action}.`, { e: e.message, s: e.stack?.slice(0, 300) });\n return new Response(\n JSON.stringify({ err: `Internal Q err: ${e.message}`, st: null, dbgL: reqLgr.getLogs(), mid: reqId }),\n { status: 500, headers },\n );\n }\n }\n\n if (req.method === \"POST\" && url.pathname === \"/execute\") {\n log(\"IN\", \"H\", \"POST /exec\");\n let vInput: UserIn;\n let rawP = \"[No payload]\";\n try {\n const p = await req.json();\n rawP = JSON.stringify(p).slice(0, 300);\n log(\"DB\", \"H\", \"Payload /exec\", { p: rawP });\n vInput = UserInSchema.parse(p);\n log(\"IN\", \"H\", \"Payload valid OK /exec.\", { t: vInput.task });\n } catch (e: any) {\n let d = \"Bad input.\";\n if (e instanceof z.ZodError) d = e.issues.map(i => `${i.path.join(\".\")}: ${i.message}`).join(\"; \"); else d =\n e.message;\n log(\"WN\", \"H\", \"Valid fail /exec.\", { e: d, p: rawP });\n return new Response(\n JSON.stringify({ err: \"Bad input data.\", details: d, dbgL: reqLgr.getLogs(), mid: `m-v-${reqId}` }),\n { status: 400, headers },\n );\n }\n const execLgr = new LogAgent();\n const orch = new Orchestrator(execLgr);\n const finalRes: FinalRes = await orch.runWorkflow(vInput);\n finalRes.dbgLog = execLgr.getLogs();\n const respS = finalRes.err ? (finalRes.err.startsWith(\"AI output was unparseable\") ? 422 : 500) : 200; // 422 for unprocessable AI entity\n log(\"IN\", \"H\", `/exec wf end status ${respS}`, { mid: finalRes.mid, hasE: !!finalRes.err });\n return new Response(JSON.stringify(finalRes), { status: respS, headers });\n }\n\n log(\"WN\", \"H\", `Unsup req: ${req.method} ${url.pathname}`);\n const errL = reqLgr.getLogs();\n return new Response(JSON.stringify({ err: \"Not Found\", dbgL: errL, mid: reqId }), { status: 404, headers });\n } catch (err: any) {\n log(\"ER\", \"H\", \"Unhandled exc.\", { e: err.message, s: err.stack?.slice(0, 300) });\n const crashL = reqLgr.getLogs();\n return new Response(\n JSON.stringify({\n err: \"Internal Error\",\n details: \"Unexpected server error.\",\n dbgL: crashL,\n mid: `m-c-${reqId}`,\n st: null, // Ensure 'st' (state) is defined as null for error consistency if frontend expects it\n }),\n { status: 500, headers },\n );\n }\n}","import { fetch } from \"https://esm.town/v/std/fetch\";\nimport { OpenAI } from \"https://esm.town/v/std/openai\";\nimport { z } from \"npm:zod\";\n\n// --- Types (Minimal & Core) ---\ntype TskIn = { mid: string; tid: string; p: T };\ntype TskOut = { mid: string; cid: string; p: T; e?: string };\ntype Lvl = \"DB\" | \"IN\" | \"WN\" | \"ER\"; // Abbreviated log levels\ntype LogFn = (lvl: Lvl, comp: string, msg: string, dt?: any, mid?: string, tid?: string) => void;\ntype LogEntry = { ts: string; mid: string; tid?: string; lvl: Lvl; comp: string; msg: string; dt?: any };\ntype BaseIn = { bType?: string; sDesc?: string; tAud?: string; bTone?: string; goals?: string };\n\n// Task-specific inputs (reusing types from before, minimal change needed)\ntype CntIn = BaseIn & { cType: string; plat: string; specInstr?: string };\ntype StratIn = BaseIn & { plats: string[]; budgetLvl?: string; duration?: string };\ntype LogoIn = BaseIn & { bName: string; stylePrefs?: string; colorPrefs?: string };\ntype BrandIn = BaseIn & { bName: string; coreVals?: string };\ntype SimIn = BaseIn & { simType: string; specParams: Record | string };\ntype MetaIn = { pInput: string };\n\n// Consolidated User Input Schema (reusing schema, essential for validation)\nconst UserInSchema = z.object({\n task: z.enum([\n \"create_content\",\n \"develop_strategy\",\n \"generate_logo_concept\",\n \"build_brand_identity\",\n \"run_simulation\",\n \"meta_prompt\",\n ]),\n data: z.record(z.any()),\n}).refine((input) => {\n const d = input.data;\n const has = (key: string) => d && d[key] != null && d[key] !== \"\";\n const hasArr = (key: string) => Array.isArray(d?.[key]) && d[key].length > 0;\n switch (input.task) {\n case \"create_content\":\n return has(\"businessType\") && has(\"serviceDescription\") && has(\"targetAudience\") && has(\"contentType\")\n && has(\"platform\");\n case \"develop_strategy\":\n return has(\"businessType\") && has(\"serviceDescription\") && has(\"targetAudience\") && hasArr(\"platforms\");\n case \"generate_logo_concept\":\n return has(\"businessType\") && has(\"serviceDescription\") && has(\"targetAudience\") && has(\"businessName\");\n case \"build_brand_identity\":\n return has(\"businessType\") && has(\"serviceDescription\") && has(\"targetAudience\") && has(\"businessName\");\n case \"run_simulation\":\n return has(\"businessType\") && has(\"serviceDescription\") && has(\"targetAudience\") && has(\"simulationType\")\n && has(\"specificParams\") && (typeof d?.specificParams === \"object\" || typeof d?.specificParams === \"string\");\n case \"meta_prompt\":\n return has(\"prompt_input\");\n default:\n return false;\n }\n}, { message: \"Missing required fields for the selected task or invalid data format.\" });\ntype UserIn = z.infer;\n\n// Task-specific outputs (MODIFIED LogoOut)\ntype CntOut = { content: string; suggestions?: string };\ntype StratOut = { strategyMarkdown: string; advertisingIdeas?: string[]; audienceInsights?: string[] }; // Renamed fields for clarity\n// MODIFIED LogoOut for better JSON structure\ntype LogoConceptDetail = { title: string; description: string; aiPrompt: string };\ntype LogoOut = { concepts: LogoConceptDetail[]; svgExample?: string };\n\ntype BrandOut = {\n brandVoiceAndPersonality: string;\n keyMessages: string[];\n colorPaletteSuggestions: string; // Detailed string, potentially with hex and justification\n typographySuggestions: string; // Detailed string\n};\ntype SimOut = { simulationResult: string; analysis?: string };\ntype MetaOut = { systemPrompt: string };\n\ntype FinalRes = { res?: any; err?: string; details?: string; dbgLog: LogEntry[]; mid: string };\n\n// Questionnaire Types (reusing types)\ntype QState = {\n sid: string;\n task: UserIn[\"task\"];\n reqFlds: string[];\n answers: Partial>;\n qHist: { qid: string; field: string; text: string }[];\n isDone: boolean;\n};\ntype Quest = {\n id: string;\n field: string;\n text: string;\n type: \"text\" | \"textarea\" | \"select\";\n opts?: { lbl: string; val: string }[];\n};\ntype QResp = { sid: string; nextQ?: Quest; isDone: boolean; data?: UserIn; err?: string; state: QState | null };\n\n// --- Data Structure for Questionnaire Fields (Simplified) ---\nconst Q_FIELDS: {\n [key: string]: {\n txt: string;\n typ?: Quest[\"type\"];\n opts?: Quest[\"opts\"];\n opt?: true;\n tasks?: Array;\n ph?: string;\n map?: string;\n };\n} = {\n bType: { txt: \"Type of business?\", tasks: [\"base\"], ph: \"e.g., SaaS, Bakery\", map: \"businessType\" },\n sDesc: {\n txt: \"Describe products/services.\",\n typ: \"textarea\",\n tasks: [\"base\"],\n ph: \"Briefly explain what you offer\",\n map: \"serviceDescription\",\n },\n tAud: {\n txt: \"Ideal customer? (Specific)\",\n typ: \"textarea\",\n tasks: [\"base\"],\n ph: \"e.g., Small business owners in marketing\",\n map: \"targetAudience\",\n },\n bTone: {\n txt: \"Desired tone?\",\n typ: \"select\",\n opt: true,\n tasks: [\"base\"],\n opts: [\n { lbl: \"-- Tone --\", val: \"\" },\n { lbl: \"Professional\", val: \"professional\" },\n { lbl: \"Friendly\", val: \"friendly\" },\n { lbl: \"Casual\", val: \"casual\" },\n { lbl: \"Humorous\", val: \"humorous\" },\n { lbl: \"Informative\", val: \"informative\" },\n ],\n map: \"brandTone\",\n },\n goals: { txt: \"Main goals for this task?\", typ: \"textarea\", opt: true, tasks: [\"base\"], map: \"goals\" },\n bName: { txt: \"Exact business name?\", tasks: [\"generate_logo_concept\", \"build_brand_identity\"], map: \"businessName\" },\n cType: {\n txt: \"Content type needed?\",\n typ: \"select\",\n tasks: [\"create_content\"],\n opts: [\n { lbl: \"-- Type --\", val: \"\" },\n { lbl: \"Promo Post\", val: \"promo_post\" },\n { lbl: \"Blog Ideas\", val: \"blog_idea\" },\n { lbl: \"Blog Draft\", val: \"blog_draft\" },\n { lbl: \"Ad Copy\", val: \"ad_copy\" },\n ],\n map: \"contentType\",\n },\n plat: {\n txt: \"Platform for content?\",\n typ: \"select\",\n tasks: [\"create_content\"],\n opts: [\n { lbl: \"-- Platform --\", val: \"\" },\n { lbl: \"Generic\", val: \"generic\" },\n { lbl: \"Facebook\", val: \"facebook\" },\n { lbl: \"Instagram\", val: \"instagram\" },\n { lbl: \"LinkedIn\", val: \"linkedin\" },\n { lbl: \"Blog\", val: \"blog\" },\n ],\n map: \"platform\",\n },\n specInstr: {\n txt: \"Specific instructions/keywords?\",\n typ: \"textarea\",\n opt: true,\n tasks: [\"create_content\"],\n map: \"specificInstructions\",\n },\n plats: {\n txt: \"Strategy platforms? (comma separated)\",\n tasks: [\"develop_strategy\"],\n ph: \"e.g., LinkedIn, SEO\",\n map: \"platforms\",\n },\n budgetLvl: {\n txt: \"Approx budget level?\",\n typ: \"select\",\n opt: true,\n tasks: [\"develop_strategy\"],\n opts: [{ lbl: \"-- Budget --\", val: \"\" }, { lbl: \"Low (Organic)\", val: \"low\" }, {\n lbl: \"Medium (Mod Ads)\",\n val: \"medium\",\n }, { lbl: \"High (Sig Ads)\", val: \"high\" }],\n map: \"budgetLevel\",\n },\n duration: { txt: \"Strategy duration?\", opt: true, tasks: [\"develop_strategy\"], map: \"duration\" },\n stylePrefs: { txt: \"Logo style preferences?\", opt: true, tasks: [\"generate_logo_concept\"], map: \"stylePreferences\" },\n colorPrefs: { txt: \"Logo color preferences?\", opt: true, tasks: [\"generate_logo_concept\"], map: \"colorPreferences\" },\n coreVals: {\n txt: \"Core values/mission?\",\n typ: \"textarea\",\n opt: true,\n tasks: [\"build_brand_identity\"],\n map: \"coreValues\",\n },\n simType: {\n txt: \"Simulation type?\",\n typ: \"select\",\n tasks: [\"run_simulation\"],\n opts: [{ lbl: \"-- Sim --\", val: \"\" }, { lbl: \"Ad Response Est.\", val: \"ad_response_estimate\" }, {\n lbl: \"Customer Interaction\",\n val: \"customer_interaction\",\n }, { lbl: \"Other\", val: \"other\" }],\n map: \"simulationType\",\n },\n specParams: {\n txt: \"Sim details/params?\",\n typ: \"textarea\",\n tasks: [\"run_simulation\"],\n ph: \"Ad: audience, platform, offer. Cust: scenario, issue. Other: describe.\",\n map: \"specificParams\",\n },\n pInput: { txt: \"AI task desc or prompt to refine?\", typ: \"textarea\", tasks: [\"meta_prompt\"], map: \"prompt_input\" },\n};\n\n// --- LogAgent Class (Minor Abbreviation) ---\nclass LogAgent {\n logs: LogEntry[];\n midCtx?: string;\n constructor(mid?: string) {\n this.logs = [];\n this.midCtx = mid;\n }\n log(lvl: Lvl, comp: string, msg: string, dt?: any, mid?: string, tid?: string): void {\n const effMid = mid || this.midCtx || \"N/A\";\n let safeDt: any;\n if (dt != null) {\n try {\n safeDt = typeof structuredClone === \"function\"\n ? structuredClone(dt)\n : JSON.parse(JSON.stringify(dt, this.getCircReplacer()));\n } catch (e) {\n safeDt = `[Log Err: ${(e as Error).message}]`;\n }\n } else safeDt = dt;\n const entry: LogEntry = { ts: new Date().toISOString(), mid: effMid, tid, lvl, comp, msg, dt: safeDt };\n this.logs.push(entry);\n const dtStr = safeDt != null ? `| D: ${JSON.stringify(safeDt)}` : \"\";\n console.log(\n `[${entry.ts}] [${entry.mid}] [${entry.tid || \"---\"}] [${entry.lvl}] [${entry.comp}] ${entry.msg} ${dtStr}`,\n );\n }\n private getCircReplacer = () => {\n const s = new WeakSet();\n return (_k: string, v: any) => {\n if (typeof v === \"object\" && v !== null) { // Corrected 'o' to 'object'\n if (s.has(v)) return \"[C]\";\n s.add(v);\n }\n return v;\n };\n };\n getLogs(): LogEntry[] {\n try {\n return typeof structuredClone === \"function\"\n ? structuredClone(this.logs)\n : JSON.parse(JSON.stringify(this.logs, this.getCircReplacer()));\n } catch (e) {\n console.error(\"[LogAgent] Fail copy:\", e);\n return [...this.logs];\n }\n }\n clearLogs(): void {\n this.logs = [];\n }\n setMidCtx(mid: string): void {\n this.midCtx = mid;\n }\n}\n\n// --- OpenAI Call Wrapper (MODIFIED to enable JSON mode) ---\nasync function callOpenAI(\n sysP: string,\n userP: string,\n mid: string,\n tid: string,\n log: LogFn,\n): Promise {\n log(\"DB\", \"oai\", `Call tid=${tid}`, { spL: sysP.length, upL: userP.length }, mid, tid);\n try {\n const oai = new OpenAI();\n const comp = await oai.chat.completions.create({\n model: \"gpt-4o-mini\",\n messages: [{ role: \"system\", content: sysP }, {\n role: \"user\",\n content: userP + \"\\n\\nRespond with JSON only or you will break the server.\",\n }],\n temperature: 0.7,\n response_format: { type: \"json_object\" }, // Enable JSON mode\n });\n const resT = comp.choices[0]?.message?.content;\n const usage = comp.usage;\n if (!resT) {\n log(\"WN\", \"oai\", `No text tid=${tid}.`, { usage, fin: comp.choices[0]?.finish_reason }, mid, tid);\n return null;\n }\n log(\"IN\", \"oai\", `OK tid=${tid}`, { resL: resT.length, usage, fin: comp.choices[0]?.finish_reason }, mid, tid);\n return resT.trim(); // The result should be a JSON string\n } catch (err: any) {\n const d = { m: err.message, n: err.name, s: err.stack, t: err.type, c: err.code }; // Abbreviated error details\n log(\"ER\", \"oai\", `Fail tid=${tid}`, { e: d }, mid, tid);\n throw new Error(`API fail: ${err.message}` + (err.code ? ` (Code: ${err.code})` : \"\"));\n }\n}\n\n// --- QAgent Class (Refactored & Concise) ---\nclass QAgent {\n lg: LogFn;\n fDefs = Q_FIELDS;\n constructor(lgFn: LogFn) {\n this.lg = lgFn;\n }\n\n private getTaskFields(task: UserIn[\"task\"]) {\n const fields = Object.keys(this.fDefs).filter(k =>\n this.fDefs[k].tasks?.includes(task) || (this.fDefs[k].tasks?.includes(\"base\") && task !== \"meta_prompt\")\n );\n const req = fields.filter(k => !this.fDefs[k].opt);\n const opt = fields.filter(k => this.fDefs[k].opt);\n return { req, opt };\n }\n\n private isOptional(f: string): boolean {\n return this.fDefs[f]?.opt === true;\n }\n\n private selectNextField(req: string[], opt: string[], ans: Record): string | null {\n for (const f of req)\n if (!(f in ans) || (ans[f] === \"\" && !this.isOptional(f))) {\n this.lg(\"DB\", \"QA\", `Sel REQ: ${f}`, { ak: Object.keys(ans) });\n return f;\n }\n for (const f of opt)\n if (!(f in ans)) {\n this.lg(\"DB\", \"QA\", `Sel OPT: ${f}`, { ak: Object.keys(ans) });\n return f;\n }\n this.lg(\"DB\", \"QA\", `All fields done.`, { ak: Object.keys(ans) });\n return null;\n }\n\n private formatQuest(f: string, st: QState): Quest {\n const id = `${f}-${Date.now()}`;\n const fC = this.fDefs[f];\n let txt = fC?.txt || `Provide '${f}'.`;\n const typ = fC?.typ || \"text\";\n const opts = fC?.opts;\n const isOpt = fC?.opt || false;\n const bN = st?.answers?.bName || \"your business\";\n if (f === \"bType\") txt = `Type of business for ${bN}?`;\n if (f === \"sDesc\") txt = `Describe products/services for ${bN}.`;\n if (f === \"tAud\") txt = `Ideal customer for ${bN}? (Specific)`;\n if (f === \"coreVals\") txt = `Core values/mission of ${bN}?${isOpt ? \" (Optional)\" : \"\"}`;\n txt += isOpt && !txt.includes(\"(Optional)\") ? \" (Optional)\" : \"\";\n return { id, field: f, text: txt, type: typ, opts };\n }\n\n async nextStep(\n sid: string | null,\n task: UserIn[\"task\"],\n lastAns: { field: string; value: any } | null,\n curSt?: QState | null,\n ): Promise {\n let st: QState;\n const sessId = sid || `s-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`;\n const { req: reqFlds, opt: optFlds } = this.getTaskFields(task);\n const allFlds = [...reqFlds, ...optFlds];\n\n if (!curSt || !sid || curSt.sid !== sessId || curSt.task !== task) {\n this.lg(\"IN\", \"QA\", `Init/Reset state: ${task}`, { os: curSt?.sid, ot: curSt?.task, ns: sessId }, sessId);\n st = { sid: sessId, task, reqFlds: allFlds, answers: {}, qHist: [], isDone: false };\n if (lastAns?.field && allFlds.includes(lastAns.field)) {\n const v = typeof lastAns.value === \"string\" ? lastAns.value.trim() : lastAns.value;\n if (v !== \"\" || this.isOptional(lastAns.field)) st.answers[lastAns.field] = v;\n else this.lg(\"DB\", \"QA\", `Ign empty req init: ${lastAns.field}`, { v }, sessId);\n }\n } else {\n st = { ...curSt, answers: { ...curSt.answers }, qHist: [...curSt.qHist] };\n if (lastAns?.field && allFlds.includes(lastAns.field)) {\n this.lg(\"DB\", \"QA\", `Proc ans: ${lastAns.field}`, { v: lastAns.value }, st.sid);\n const v = typeof lastAns.value === \"string\" ? lastAns.value.trim() : lastAns.value;\n if (v !== \"\" || !reqFlds.includes(lastAns.field)) st.answers[lastAns.field] = v;\n else {\n this.lg(\"WN\", \"QA\", `Empty req ans: ${lastAns.field}. Re-ask.`, undefined, st.sid);\n delete st.answers[lastAns.field];\n }\n const lastQ = st.qHist[st.qHist.length - 1];\n if (lastQ && lastQ.field !== lastAns.field)\n this.lg(\"WN\", \"QA\", `Ans for ${lastAns.field}, last Q was ${lastQ?.field}. Sync issue?`, undefined, st.sid);\n }\n }\n\n const nextF = this.selectNextField(reqFlds, optFlds, st.answers);\n\n if (!nextF) {\n this.lg(\"IN\", \"QA\", `All fields for ${st.task}. Validate.`, { g: Object.keys(st.answers) }, st.sid);\n const finalD = this.finalizeData(st.answers, st.task, st.sid);\n try {\n UserInSchema.parse(finalD);\n this.lg(\"IN\", \"QA\", \"Valid OK.\", undefined, st.sid);\n st.isDone = true;\n return { sid: st.sid, isDone: true, data: finalD, state: st };\n } catch (validErr) {\n if (validErr instanceof z.ZodError) {\n const errMsgs = validErr.issues.map(i => `${i.path.join(\".\")}: ${i.message}`).join(\"; \");\n this.lg(\"ER\", \"QA\", \"Valid FAIL.\", { e: validErr.issues }, st.sid);\n st.isDone = false;\n const firstFailPath = validErr.issues[0]?.path;\n let fToReask: string | null = null;\n if (firstFailPath && firstFailPath.length > 1) {\n const schemaF = firstFailPath[1].toString();\n fToReask = Object.keys(this.fDefs).find(k => this.fDefs[k].map === schemaF) || null;\n if (fToReask && !allFlds.includes(fToReask)) fToReask = null;\n }\n if (fToReask) {\n delete st.answers[fToReask];\n const lastQIdx = st.qHist.findIndex(q => q.field === fToReask);\n if (lastQIdx !== -1) st.qHist.splice(lastQIdx, 1);\n const nextQ = this.formatQuest(fToReask, st);\n this.lg(\"WN\", \"QA\", `Valid fail on ${fToReask}. Re-ask.`, { e: errMsgs }, st.sid);\n st.qHist.push({ qid: nextQ.id, field: nextQ.field, text: nextQ.text });\n return {\n sid: st.sid,\n isDone: false,\n nextQ,\n state: st,\n err: `Check input for '${nextQ.text.replace(/\\s+\$Optional\$\\s*$/, \"\")}': ${errMsgs}`,\n };\n } else {\n this.lg(\"ER\", \"QA\", \"Valid fail, no field to re-ask.\", { e: errMsgs }, st.sid);\n return { sid: st.sid, isDone: false, err: `Valid fail: ${errMsgs}. Cannot proceed.`, state: st };\n }\n } else {\n this.lg(\"ER\", \"QA\", \"Unexpected valid error.\", { e: (validErr as Error).message }, st.sid);\n return {\n sid: st.sid,\n isDone: false,\n err: `Unexpected valid error: ${(validErr as Error).message}`,\n state: st,\n };\n }\n }\n } else {\n const nextQ = this.formatQuest(nextF, st);\n st.qHist.push({ qid: nextQ.id, field: nextQ.field, text: nextQ.text });\n this.lg(\"IN\", \"QA\", `Ask: ${nextF}`, { qT: nextQ.text }, st.sid);\n return { sid: st.sid, isDone: false, nextQ, state: st };\n }\n }\n\n async goBack(sid: string, task: UserIn[\"task\"], curSt: QState): Promise {\n this.lg(\"IN\", \"QA\", `Back action: ${task}`, { sessId: sid });\n const st = { ...curSt, answers: { ...curSt.answers }, qHist: [...curSt.qHist] };\n if (st.qHist.length < 1) {\n this.lg(\"WN\", \"QA\", \"Back fail: no history.\", undefined, st.sid);\n return { sid, isDone: false, err: \"Cannot go back further.\", state: st };\n }\n const lastQ = st.qHist.pop();\n if (lastQ && st.answers) {\n delete st.answers[lastQ.field];\n this.lg(\"DB\", \"QA\", `Back: removed ans for '${lastQ.field}'.`, { qid: lastQ.qid }, st.sid);\n }\n const QToReask = st.qHist[st.qHist.length - 1];\n if (QToReask) {\n const nextQ = this.formatQuest(QToReask.field, st);\n st.qHist[st.qHist.length - 1] = { qid: nextQ.id, field: nextQ.field, text: nextQ.text };\n this.lg(\"IN\", \"QA\", `Back to Q: ${QToReask.field}`, { qT: nextQ.text }, st.sid);\n return { sid, isDone: false, nextQ, state: st };\n } else {\n this.lg(\"WN\", \"QA\", \"Back fail: history empty after pop. Restarting.\", undefined, st.sid);\n return this.nextStep(sid, task, null, null);\n }\n }\n\n async restart(sid: string, task: UserIn[\"task\"]): Promise {\n this.lg(\"IN\", \"QA\", `Restart action: ${task}`, { sessId: sid });\n return this.nextStep(sid, task, null, null);\n }\n\n private finalizeData(gatheredAnswers: Record, task: UserIn[\"task\"], sid?: string): UserIn {\n let processedAnswers: Record = { ...gatheredAnswers };\n const { req: reqFlds, opt: optFlds } = this.getTaskFields(task);\n\n optFlds.forEach(internalKey => {\n if (\n processedAnswers[internalKey] == null\n || (typeof processedAnswers[internalKey] === \"string\" && processedAnswers[internalKey].trim() === \"\")\n ) {\n delete processedAnswers[internalKey];\n }\n });\n\n reqFlds.forEach(internalKey => {\n const fieldDef = this.fDefs[internalKey];\n const isFieldOptionalInDef = fieldDef?.opt === true;\n if (\n !isFieldOptionalInDef\n && (processedAnswers[internalKey] == null\n || (typeof processedAnswers[internalKey] === \"string\" && processedAnswers[internalKey].trim() === \"\"))\n ) {\n this.lg(\n \"WN\",\n \"finalize\",\n `Required internal field '${internalKey}' is empty/null before mapping. UserInSchema will validate mapped fields.`,\n { v: processedAnswers[internalKey] },\n sid,\n );\n }\n });\n\n const platsInternalKey = \"plats\";\n if (task === \"develop_strategy\" && processedAnswers[platsInternalKey] !== undefined) {\n if (typeof processedAnswers[platsInternalKey] === \"string\") {\n processedAnswers[platsInternalKey] = processedAnswers[platsInternalKey].split(\",\").map((p: string) => p.trim())\n .filter((p: string) => p.length > 0);\n } else if (!Array.isArray(processedAnswers[platsInternalKey])) {\n processedAnswers[platsInternalKey] = [];\n }\n processedAnswers[platsInternalKey] = (processedAnswers[platsInternalKey] as any[]).filter((p: any) =>\n typeof p === \"string\"\n );\n }\n\n const specParamsInternalKey = \"specParams\";\n if (task === \"run_simulation\" && processedAnswers[specParamsInternalKey] !== undefined) {\n const val = processedAnswers[specParamsInternalKey];\n if (typeof val === \"string\") {\n const pStr = val.trim();\n if (pStr.startsWith(\"{\") && pStr.endsWith(\"}\")) {\n try {\n const p = JSON.parse(pStr);\n processedAnswers[specParamsInternalKey] = typeof p === \"object\" && p !== null && !Array.isArray(p)\n ? p\n : { description: pStr };\n } catch (e) {\n processedAnswers[specParamsInternalKey] = { description: pStr };\n }\n } else {\n processedAnswers[specParamsInternalKey] = { description: pStr };\n }\n } else if (typeof val !== \"object\" || val === null || Array.isArray(val)) {\n try {\n processedAnswers[specParamsInternalKey] = { description: JSON.stringify(val) };\n } catch (e) {\n processedAnswers[specParamsInternalKey] = { description: \"[Error serializing specificParams]\" };\n }\n }\n }\n\n const schemaData: Record = {};\n for (const internalKey in processedAnswers) {\n if (Object.prototype.hasOwnProperty.call(processedAnswers, internalKey)) {\n const fieldDef = this.fDefs[internalKey];\n if (fieldDef && fieldDef.map) {\n schemaData[fieldDef.map] = processedAnswers[internalKey];\n } else if (fieldDef) {\n this.lg(\n \"WN\",\n \"finalize\",\n `Internal field '${internalKey}' defined in Q_FIELDS but has no 'map' property. It will NOT be added to schema data unless '${internalKey}' is a direct schema key.`,\n { val: processedAnswers[internalKey] },\n sid,\n );\n } else {\n this.lg(\n \"WN\",\n \"finalize\",\n `Field '${internalKey}' from answers is not defined in Q_FIELDS. It will be excluded from schema data.`,\n { val: processedAnswers[internalKey] },\n sid,\n );\n }\n }\n }\n this.lg(\"DB\", \"finalize\", \"Data mapped to schema keys for validation\", {\n fromAnswers: processedAnswers,\n toSchemaData: schemaData,\n }, sid);\n return { task, data: schemaData };\n }\n}\n\n// --- Generic Agent Execution Function & Configurations (MODIFIED) ---\ntype AgentConfig = {\n n: string; // Name\n p: (p: T_In) => { sys: string; user: string }; // Prompts\n ps: (rawJson: string) => T_Out; // Parse JSON output\n pp?: (out: T_Out, p: T_In) => T_Out; // Post-process\n};\n\nconst AGENT_CFGS: { [key in UserIn[\"task\"]]: AgentConfig } = {\n create_content: {\n n: \"Cnt\",\n p: (p: CntIn) => ({\n sys: `You are a helpful AI copywriter.\nContext: Business Type: ${p.bType || \"N/A\"}, Services: ${p.sDesc || \"N/A\"}, Target Audience: ${\n p.tAud || \"General Public\"\n }, Tone: ${p.bTone || \"Professional\"}, Goal: ${p.goals || \"Engage audience\"}.\nTask: Create content of type \"${p.cType}\" for platform \"${p.plat}\". Specific Instructions: ${p.specInstr || \"None\"}.\nPlatform Guidelines:\n- Social media posts (Facebook, Instagram, LinkedIn): Include relevant hashtags, calls to action, and aim for brevity and engagement.\n- Blog Ideas: Generate 3-5 distinct blog post titles or brief concepts.\n- Blog Draft: Write a 300-500 word draft in Markdown format.\n- Ad Copy: Create compelling ad copy with a clear headline, description, and call to action.\n- Generic: Create adaptable content suitable for general use.\n\nOutput ONLY a valid JSON object with the following structure:\n{\n \"content\": \"The generated content as a string. For 'Blog Ideas', this should be a list of titles/concepts separated by newlines.\",\n \"suggestions\": \"Optional brief suggestions for improving or using the content (e.g., 'Consider adding an image/video.', 'Research relevant keywords for blog posts.') as a string.\"\n}\nDo not include any explanatory text before or after the JSON object.`,\n user: `Generate the ${p.cType} based on the provided context.`,\n }),\n ps: (rawJson: string): CntOut => {\n const parsed = JSON.parse(rawJson);\n if (typeof parsed.content !== \"string\") {\n throw new Error(\"Parsed JSON for CntOut is missing or has invalid 'content' field.\");\n }\n // suggestions is optional\n return {\n content: parsed.content,\n suggestions: typeof parsed.suggestions === \"string\" ? parsed.suggestions : undefined,\n };\n },\n pp: (out, p) => { // Post-processing suggestions can still be useful\n let suggestions = out.suggestions || \"\";\n if (p.cType === \"promo_post\" || p.cType === \"ad_copy\") suggestions += \" Consider A/B testing different versions.\";\n else if (p.cType === \"blog_idea\" && (!suggestions || !suggestions.toLowerCase().includes(\"keyword\")))\n suggestions += \" Research keywords for each idea.\";\n else if (p.cType === \"blog_draft\") suggestions += \" Add internal/external links and a concluding paragraph.\";\n return { ...out, suggestions: suggestions.trim() || undefined };\n },\n },\n develop_strategy: {\n n: \"Strat\",\n p: (p: StratIn) => {\n const pS = p.plats?.join(\", \") || \"N/A\";\n return {\n sys: `You are an AI Marketing Strategist.\nContext: Business Type: ${p.bType || \"N/A\"}, Services: ${p.sDesc || \"N/A\"}, Target Audience: ${\n p.tAud || \"General Public\"\n }, Primary Goals: ${\n p.goals || \"Generate leads and increase brand awareness\"\n }, Key Platforms: ${pS}, Budget Level: ${p.budgetLvl || \"Low\"}, Strategy Duration: ${\n p.duration || \"Ongoing\"\n }, Desired Tone: ${p.bTone || \"Professional\"}.\n\nTask: Develop a concise marketing strategy.\n\nOutput ONLY a valid JSON object with the following structure:\n{\n \"strategyMarkdown\": \"A comprehensive strategy document in Markdown format. It should include sections like: Executive Summary, Target Audience Deep Dive (insights, behaviors, needs), Recommended Platforms (evaluation of ${pS} and any other suggestions, with justification), Content Themes & Key Messages (2-4 themes with example messages), Platform-Specific Tactics (2-3 actionable tactics per chosen platform, prioritized by budget), KPIs (3-5 key performance indicators like Traffic, Conversion Rate, Engagement Rate), and Next Steps.\",\n \"advertisingIdeas\": [\"A list of 2-3 distinct advertising ideas or campaign concepts as strings, if applicable based on budget and goals.\"],\n \"audienceInsights\": [\"A list of 2-3 key bullet-point insights about the target audience behavior or needs as strings.\"]\n}\nEnsure the 'strategyMarkdown' is well-structured. Do not include any explanatory text before or after the JSON object.`,\n user: `Develop the marketing strategy.`,\n };\n },\n ps: (rawJson: string): StratOut => {\n const parsed = JSON.parse(rawJson);\n if (typeof parsed.strategyMarkdown !== \"string\") {\n throw new Error(\"Parsed JSON for StratOut is missing or has invalid 'strategyMarkdown' field.\");\n }\n return {\n strategyMarkdown: parsed.strategyMarkdown,\n advertisingIdeas: Array.isArray(parsed.advertisingIdeas)\n ? parsed.advertisingIdeas.filter(s => typeof s === \"string\")\n : undefined,\n audienceInsights: Array.isArray(parsed.audienceInsights)\n ? parsed.audienceInsights.filter(s => typeof s === \"string\")\n : undefined,\n };\n },\n },\n generate_logo_concept: {\n n: \"Logo\",\n p: (p: LogoIn) => ({\n sys: `You are a Creative Design Assistant.\nContext: Business Name: \"${p.bName}\", Business Type: ${p.bType || \"N/A\"}, Services: ${\n p.sDesc || \"N/A\"\n }, Target Audience: ${p.tAud || \"General Public\"}, Style Preferences: ${\n p.stylePrefs || \"Versatile, modern\"\n }, Color Preferences: ${p.colorPrefs || \"Suggest based on business type and audience\"}, Desired Tone: ${\n p.bTone || \"Professional\"\n }.\n\nTask: Generate 2-3 distinct logo concepts. For each concept, provide a title, a detailed description (visuals, symbolism, font ideas, color palette considerations), and a detailed AI art generation prompt suitable for models like DALL-E or Midjourney. Optionally, provide a very basic SVG example if it's simple (e.g., an abstract shape or initial).\n\nOutput ONLY a valid JSON object with the following structure:\n{\n \"concepts\": [\n {\n \"title\": \"Concept Title 1 (e.g., Modern Initial Mark)\",\n \"description\": \"Detailed description of the concept, including visual elements, symbolism, potential font styles, and color palette ideas...\",\n \"aiPrompt\": \"A detailed AI art generation prompt for this concept. Example: 'Minimalist vector logo for [Business Name], letter [X], [style preferences e.g., sleek lines, geometric shapes], [color preferences e.g., blue and silver], white background, high resolution.'\"\n }\n // ... (include 1 or 2 more concept objects)\n ],\n \"svgExample\": \"Optional: A string containing very basic SVG code for one of the concepts, or leave empty if not applicable/too complex. Example: ''\"\n}\nDo not include any explanatory text before or after the JSON object.`,\n user: `Generate logo concepts for \"${p.bName}\".`,\n }),\n ps: (rawJson: string): LogoOut => {\n const parsed = JSON.parse(rawJson);\n if (\n !Array.isArray(parsed.concepts)\n || parsed.concepts.some(c =>\n typeof c.title !== \"string\" || typeof c.description !== \"string\" || typeof c.aiPrompt !== \"string\"\n )\n ) {\n throw new Error(\"Parsed JSON for LogoOut has invalid 'concepts' structure.\");\n }\n const concepts: LogoConceptDetail[] = parsed.concepts.map(c => ({\n title: c.title,\n description: c.description,\n aiPrompt: c.aiPrompt,\n }));\n return {\n concepts: concepts,\n svgExample: typeof parsed.svgExample === \"string\" && parsed.svgExample.trim() !== \"\"\n ? parsed.svgExample\n : undefined,\n };\n },\n },\n build_brand_identity: {\n n: \"Brand\",\n p: (p: BrandIn) => ({\n sys: `You are an AI Brand Strategist.\nContext: Business Name: \"${p.bName}\", Business Type: ${p.bType || \"N/A\"}, Services: ${\n p.sDesc || \"N/A\"\n }, Target Audience: ${p.tAud || \"General Public\"}, Core Values/Mission: ${p.coreVals || \"N/A\"}, Desired Tone: ${\n p.bTone || \"Professional\"\n }.\n\nTask: Define key elements of the brand identity.\n\nOutput ONLY a valid JSON object with the following structure:\n{\n \"brandVoiceAndPersonality\": \"A detailed description of the brand's voice, personality, and communication style. Should align with the desired tone: '${\n p.bTone || \"Professional\"\n }'.\",\n \"keyMessages\": [\n \"A list of 3-5 concise key messages (as strings) that communicate the brand's value proposition, targeted at the audience, and reflecting its core values.\"\n ],\n \"colorPaletteSuggestions\": \"A description of a suggested color palette, including primary, secondary, and accent colors. Mention example HEX codes and justify choices (e.g., 'Primary: #0A2A4E (Deep Blue) - conveys trust and professionalism. Secondary: #F1C40F (Gold) - adds a touch of quality and optimism.').\",\n \"typographySuggestions\": \"Suggestions for typography, including a headline font and a body text font (preferably from Google Fonts for web accessibility). Justify choices (e.g., 'Headline: Montserrat (Bold) - modern and strong. Body: Open Sans - highly readable and versatile.').\"\n}\nDo not include any explanatory text before or after the JSON object.`,\n user: `Develop brand identity guide for \"${p.bName}\".`,\n }),\n ps: (rawJson: string): BrandOut => {\n const parsed = JSON.parse(rawJson);\n if (\n typeof parsed.brandVoiceAndPersonality !== \"string\"\n || !Array.isArray(parsed.keyMessages) || parsed.keyMessages.some(m => typeof m !== \"string\")\n || typeof parsed.colorPaletteSuggestions !== \"string\"\n || typeof parsed.typographySuggestions !== \"string\"\n ) {\n throw new Error(\"Parsed JSON for BrandOut has invalid structure or missing fields.\");\n }\n return {\n brandVoiceAndPersonality: parsed.brandVoiceAndPersonality,\n keyMessages: parsed.keyMessages,\n colorPaletteSuggestions: parsed.colorPaletteSuggestions,\n typographySuggestions: parsed.typographySuggestions,\n };\n },\n },\n run_simulation: {\n n: \"Sim\",\n p: (p: SimIn) => {\n let pD = \"N/A\";\n try {\n pD = typeof p.specParams === \"string\" ? p.specParams : JSON.stringify(p.specParams);\n } catch (e) {\n pD = \"[Serialization failed for specificParams]\";\n }\n return {\n sys: `You are an AI Simulation Model.\nContext: Business Type: ${p.bType || \"N/A\"}, Services: ${p.sDesc || \"N/A\"}, Target Audience: ${\n p.tAud || \"General Public\"\n }, Desired Tone: ${p.bTone || \"Professional\"}.\nSimulation Task Type: \"${p.simType}\".\nSpecific Parameters/Scenario: ${pD}.\n\nTask: Execute the simulation based on the type and parameters.\n- 'ad_response_estimate': Qualitatively estimate potential ad effectiveness. Consider likelihoods, target audience alignment. Do not give financial guarantees.\n- 'customer_interaction': Role-play a customer interaction. Provide dialogue for both 'Customer:' and 'Agent:'. Aim for 3-5 exchanges.\n- 'other': Execute the simulation as described in the 'Specific Parameters'. Interpret the request and state any assumptions made.\n\nOutput ONLY a valid JSON object with the following structure:\n{\n \"simulationResult\": \"The direct output of the simulation. For 'customer_interaction', this should be the dialogue. For 'ad_response_estimate', a summary of the estimated response. For 'other', the result of the custom simulation.\",\n \"analysis\": \"Optional: A brief (2-4 sentences) analysis or key takeaways from the simulation outcome as a string.\"\n}\nDo not include any explanatory text before or after the JSON object.`,\n user: `Run simulation '${p.simType}'.`,\n };\n },\n ps: (rawJson: string): SimOut => {\n const parsed = JSON.parse(rawJson);\n if (typeof parsed.simulationResult !== \"string\") {\n throw new Error(\"Parsed JSON for SimOut is missing or has invalid 'simulationResult' field.\");\n }\n return {\n simulationResult: parsed.simulationResult,\n analysis: typeof parsed.analysis === \"string\" ? parsed.analysis : undefined,\n };\n },\n },\n meta_prompt: {\n n: \"Meta\",\n p: (p: MetaIn) => ({\n sys: `You are an AI Prompt Engineering Assistant.\nUser's Request for a System Prompt (this is the input you need to process):\n\\`\\`\\`\n${p.pInput}\n\\`\\`\\`\nTask: Based ONLY on the user's request above and general prompt engineering principles, generate or refine a SYSTEM prompt for an AI model like GPT-4o-mini.\nPrinciples to consider for the generated system prompt:\n1. Role: Clearly define the AI's role.\n2. Task/Goal: Specify the main task and desired outcome.\n3. Context/Input: Mention what context or inputs the AI will receive.\n4. Process: If the task is complex, outline steps.\n5. Output Format/Constraints: Define the desired output format (e.g., Markdown, JSON structure), length, tone, things to avoid. Provide examples if helpful.\n6. Ambiguity: Try to resolve potential ambiguities.\n7. Refinement: If refining an existing prompt, identify weaknesses and suggest improvements.\n\nOutput ONLY a valid JSON object with the following structure:\n{\n \"systemPrompt\": \"The complete generated or refined system prompt text as a string. Do not include markdown code fences (like \\`\\`\\`json) inside this string unless they are part of the prompt itself meant for the target AI.\"\n}\nDo not include any explanatory text before or after the JSON object.`,\n user: `Generate or refine a system prompt based on my request.`,\n }),\n ps: (rawJson: string): MetaOut => {\n const parsed = JSON.parse(rawJson);\n if (typeof parsed.systemPrompt !== \"string\" || parsed.systemPrompt.length < 10) { // Basic sanity check\n throw new Error(\n \"Parsed JSON for MetaOut is missing, has invalid 'systemPrompt' field, or prompt is too short.\",\n );\n }\n return {\n systemPrompt: parsed.systemPrompt,\n };\n },\n },\n};\n\nasync function runAgent(\n task: TskIn,\n log: LogFn,\n cfg: AgentConfig,\n): Promise> {\n const { mid, tid, p } = task;\n log(\"IN\", cfg.n, `Start tid=${tid}`, {}, mid, tid);\n try {\n const { sys, user } = cfg.p(p);\n const rawC = await callOpenAI(sys, user, mid, tid, log);\n if (!rawC) {\n log(\"WN\", cfg.n, `No LLM content tid=${tid}.`, {}, mid, tid);\n return { mid, cid: tid, p: {} as T_Out, e: `AI failed to gen ${cfg.n}.` };\n }\n let out: T_Out;\n try {\n out = cfg.ps(rawC); // rawC should be a JSON string\n } catch (e: any) {\n log(\"ER\", cfg.n, `Parse FAIL tid=${tid}. Message: ${e.message}`, { raw: rawC.slice(0, 500) }, mid, tid);\n return { mid, cid: tid, p: {} as T_Out, e: `AI output was unparseable or invalid structure: ${e.message}` };\n }\n if (cfg.pp) {\n try {\n out = cfg.pp(out, p);\n } catch (e: any) {\n log(\"ER\", cfg.n, `PP FAIL tid=${tid}.`, { e: e.message }, mid, tid);\n // Non-fatal, proceed with parsed output\n }\n }\n log(\"IN\", cfg.n, `OK tid=${tid}.`, {}, mid, tid);\n return { mid, cid: tid, p: out };\n } catch (err: any) {\n log(\"ER\", cfg.n, `Task FAIL tid=${tid}.`, { e: err.message }, mid, tid);\n return { mid, cid: tid, p: {} as T_Out, e: `${cfg.n} fail: ${err.message}` };\n }\n}\n\n// --- Orchestrator Class (Minor Abbreviation) ---\nclass Orchestrator {\n baseL: LogFn;\n mId: string;\n lg: LogFn;\n constructor(baseLgr: LogAgent) {\n this.mId = `m-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`;\n baseLgr.setMidCtx(this.mId);\n this.baseL = baseLgr.log.bind(baseLgr);\n this.lg = (lvl, comp, msg, dt?, _midIgn?, tid?) => {\n this.baseL(lvl, comp, msg, dt, this.mId, tid);\n };\n this.lg(\"IN\", \"Orch\", `Init mandate: ${this.mId}`);\n }\n async runWorkflow(input: UserIn): Promise {\n this.lg(\"IN\", \"Orch\", `Wf start: ${input.task}`, {}, this.mId);\n const tid = `${input.task.slice(0, 3)}-${Date.now()}-${Math.random().toString(36).slice(2, 7)}`;\n const tInput: TskIn = { mid: this.mId, tid: tid, p: input.data };\n let tOutput: TskOut;\n const aCfg = AGENT_CFGS[input.task];\n if (!aCfg) {\n this.lg(\"ER\", \"Orch\", `Unsup task: ${input.task}`, {}, this.mId, tid);\n return { mId: this.mId, err: `Unsup task: ${input.task}`, dbgLog: [] };\n }\n try {\n tOutput = await runAgent(tInput, this.lg, aCfg);\n if (tOutput.e) {\n this.lg(\"WN\", \"Orch\", `Agent ${tid} err.`, { e: tOutput.e }, this.mId);\n return { mid: this.mId, err: `Agent fail: ${tOutput.e}`, res: tOutput.p, dbgLog: [] };\n }\n this.lg(\"IN\", \"Orch\", `Wf task ${tid} OK.`, {}, this.mId);\n return { mid: this.mId, res: tOutput.p, dbgLog: [] };\n } catch (err: any) {\n this.lg(\"ER\", \"Orch\", `Wf critical fail ${tid}.`, { e: err.message }, this.mId);\n return { mid: this.mId, err: `Wf critical fail: ${err.message}`, dbgLog: [] };\n }\n }\n}\n\n// --- HTML Generation (Minor Abbreviation & CSS/JS reduction) ---\n// No changes to HTML/JS needed for this refactor, so it's kept as is.\n// It's long, so I'll use a comment to represent it to keep the response shorter.\n// Original HTML generation function generateOrthogonalUI() and its embedded JS remains here.\nfunction generateOrthogonalUI(): string {\n return `AI Biz Asst\n

\n \n \n \n \n

Diff · v8