[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_153":154,"_155":156},"routes/_app.x.$handle.$name",{"_95":94,"_32":45,"_103":85,"_27":104,"_13":14,"_105":-5,"_11":106,"_107":108,"_109":110,"_111":112,"_113":114,"_20":-5,"_115":-5,"_116":-5,"_117":25,"_118":119,"_120":57,"_121":-5,"_122":-5,"_123":124},"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",11,"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.ts","latestPath","fileId","3524a2c8-ad5f-11f0-9cb2-0224a6c84d84","name","fileType","http","content","// @ts-nocheck\n// Taskflow AI - Professional Autonomous Agent Factory\n// A modern, user-friendly platform for AI-powered task execution\n// ✅ [v2 UI OVERHAUL COMPLETE]\n// ✅ [v3 DYNAMIC APPROACHES COMPLETE]\n\nimport { Hono } from \"npm:hono@4.4.12\";\nimport { OpenAI } from \"https://esm.town/v/std/openai\";\nimport type { Context } from \"npm:hono@4.4.12\";\nimport { streamText } from \"npm:hono@4.4.12/streaming\";\n\n// --- TYPE DEFINITIONS ---\ninterface CompanyContext {\n company_name: string;\n company_description: string;\n use_case: string;\n language: string;\n industry?: string;\n}\ninterface ConversationHistory {\n role: \"assistant\" | \"user\";\n content: string;\n}\ninterface GatherBody {\n task: string;\n domain: string;\n history: ConversationHistory[];\n}\ninterface DynamicPromptBody {\n occupation_title: string;\n task: string;\n company_context: CompanyContext;\n}\ninterface ExecuteBody {\n refined_prompt: string;\n user_inputs: Record;\n company_context: CompanyContext;\n category_object: any; // Changed from 'category'\n}\n\n// --- CENTRALIZED CONFIGURATION ---\nconst config = {\n models: {\n major: \"gpt-4o\",\n minor: \"gpt-4o-mini\",\n },\n prompts: {\n INFORMATION_GATHERER:\n `You are a native JSON speaking expert information gatherer designed to acquire all necessary details for a given task and domain. Your goal is to ask the fewest, most concise, and natural questions possible, one at a time. If information is missing or refused, make a reasonable assumption or broaden your scope. Iteratively ask questions and integrate responses until all required information for the {{task}} within the {{domain}} is gathered.\n\n# RULES\n1. **Primary Goal**: Ask single, concise questions to gather the minimum required information for the task, aiming for completion efficiently.\n2. **Decision Point**: After analyzing the conversation history, decide if you have enough information.\n - **If YES**: Your immediate response MUST be the token \"__COMPLETE__\" followed by a single, minified JSON object of all gathered data. Do not ask for confirmation. Example: \\`__COMPLETE__\\`{\"party_date\":\"next Saturday\",\"guests\":50,\"location\":\"rented hall\"}\n - **If NO**: You are missing critical information. Proceed to the next rule.\n3. **Question Formulation**: If more info is needed, formulate the single most impactful and concise question.\n4. **Output Format (Questions)**: The question must be a short, direct interrogative. DO NOT add conversational filler, commentary, apologies, or instructions.\n\n# CONTEXT\n- Task: {{task}}\n- Domain: {{domain}}\n- Conversation History:\n{{history}}`,\n PROMPT_REFINER:\n `You are a native JSON speaking prompt engineer. Transmute raw occupational data into a concise, forceful directive for a target AI. The AI must perform the task end-to-end and return only the finished deliverable(s)—never a plan. You will receive an \"Occupation,\" \"Task,\" and \"Company Context.\" Your prompt must be a production-grade instruction that directly yields the final artifact. CORE RULE: You MUST return a prompt with {{variables}}. Your goal is to identify important, undefined entities in the Task and convert them to descriptive, {{snake_case_placeholders}}. You MUST include at least 2 placeholders. INSTRUCTIONS: 1. Persona: Start with \"You are a top 0.1% [Occupation]...\" for an expert tone. 2. Objective: Translate the Task into a formal, actionable objective implying completion. 3. Context: Use the 'Company Context' to make the objective specific (e.g., \"Your mandate for [Company Name] is to...\"). 4. Placeholders: Convert only USER-SUPPLIED functional inputs into {{snake_case_variables}}. 5. Delivery Rules: Require the AI to: - Return only the finished deliverable(s) now. - Exclude plans, methods, or explanations. - Avoid meta-commentary, caveats, or apologies. - If details are missing, make minimal, reasonable assumptions and proceed, optionally adding a short \"Assumptions\" list (<=5 bullets). - Use production-ready formatting (e.g., ready-to-send copy; runnable code in fenced blocks). - Never include sections like \"Plan\", \"Approach\", \"Methodology\". 6. Output Format: Specify the exact output structure. 7. Mandate: Conclude with a clear call to action like \"Produce the deliverable now.\" Your entire response MUST be a single, minified JSON object with one key, \"refined_prompt\". No other text or markdown.`,\n MASTER_PROMPT_ENGINEER: // This is now used by the \"prompt_building\" dynamic approach\n `You are a native JSON speaking master prompt architect. Your objective is to construct a comprehensive, self-contained blueprint for another AI to execute a task flawlessly, not to do it yourself.\n\n# INSTRUCTIONS\n1. **Deconstruct Task:** Analyze the task to identify sub-tasks, knowledge domains, constraints, and desired output formats.\n2. **Establish Persona:** Define an expert persona for the target AI (e.g., \"You are a tenured professor of economics...\").\n3. **Provide Context:** Synthesize company context into a rich narrative explaining the task's strategic importance.\n4. **Formulate Logic:** Create a clear, logical sequence of operations using headings and lists.\n5. **Set Negative Constraints:** Explicitly state what the AI should *not* do (e.g., \"Do not use informal language,\" \"Do not output a plan\").\n6. **Define Output Schema:** Meticulously define the final output's structure, format, length, and tone.\n7. **Include Examples:** Where helpful, provide concise examples of good and bad outputs.\n8. **Final Mandate:** Conclude with a clear command to initiate execution.\n\nYour output must be the prompt itself, as a clean markdown document. Do not wrap it in JSON or add commentary.`,\n// --- NEW PROMPT ---\n APPROACH_GENERATOR:\n `You are a native JSON speaking expert AI system architect. Your task is to generate 3-5 distinct, professional \"Approaches\" for completing a given task, tailored to a specific company and occupation.\n\n# RULES\n1. **Analyze Input**: You will receive a Task, Occupation, and Company Context.\n2. **Generate Approaches**: Create a list of 3-5 unique, actionable approaches an expert would take.\n3. **Define Output Object**: Each approach in the list MUST be a JSON object with this exact structure:\n - \"name\": (string) A short, \"button-friendly\" name (e.g., \"Formal Business Plan\", \"Quick Email Script\").\n - \"description\": (string) A 1-sentence description of this approach.\n - \"icon\": (string) A single, relevant emoji.\n - \"system_prompt_template\": (string) A complete, production-ready system prompt *template* for another AI to execute this approach. This template MUST include placeholders like {{company_name}}, {{company_description}}, {{language}}, {{use_case}}, {{refined_task}}, and {{user_inputs_json}} where appropriate.\n4. **Template is Key**: The \"system_prompt_template\" is the *most important part*. It is the \"DNA\" of the approach, defining the persona, rules, context, and final deliverable format for the AI that will do the work.\n5. **Output Format**: Your entire response MUST be a single, minified JSON object: {\"approaches\": [...]}\n\n# EXAMPLES OF system_prompt_template\n## Example 1: For a \"Formal Document\" approach\n\"You are a top 0.1% {{occupation}}. Your task is to produce a formal, production-ready document based on the user's request.\n# CONTEXT\n- Company: {{company_name}} ({{company_description}})\n- Goal: {{use_case}}\n- Language: {{language}}\n# TASK: {{refined_task}}\n# USER SPECIFICATIONS:\n{{user_inputs_json}}\n# RULES\n1. Return only the finished deliverable(s).\n2. Exclude plans, methods, or explanations.\n3. Use formal, professional language.\n4. Output MUST be in {{language}}.\nProduce the deliverable now.\"\n\n## Example 2: For a \"Code Generation\" approach\n\"You are an elite 10x software engineer. Your task is to write clean, efficient, and well-commented code.\n# TASK: {{refined_task}}\n# USER SPECIFICATIONS:\n{{user_inputs_json}}\n# RULES\n1. Return ONLY the runnable code in fenced markdown blocks.\n2. Do not write explanations before or after the code.\n3. Adhere to all modern best practices.\n4. All text output (comments, etc.) MUST be in {{language}}.\nProduce the code now.\"\n\n## Example 3: For a \"Prompt Engineering\" approach\n\"${config.prompts.MASTER_PROMPT_ENGINEER.replace(/\"/g, '\\\\\"')}\"\n\n# INPUT\n- Task: {{task}}\n- Occupation: {{occupation}}\n- Company Context: {{company_context}}`,\n// --- END NEW PROMPT ---\n INDUSTRY_GENERATOR:\n `You are a native JSON speaking expert business analyst. Generate a list of professional industries most relevant to the given company.\n\n# RULES\n1. **Analyze Context**: Deeply analyze the provided company context (name, description, use case) to understand their core business.\n2. **Generate Industries**: Produce a list of 10-12 diverse yet highly relevant industries.\n3. **Output Format**: Your entire response MUST be a single, minified JSON object with one key: \"industries\". The value must be an array of strings.\n\n# COMPANY CONTEXT\n{{company_context}}`,\n DYNAMIC_LIST_GENERATOR:\n `You are a native JSON speaking expert AI that generates context-aware lists for professional applications. You will receive a JSON object with company context and the list type to generate.\n\n# RULES\n1. **Analyze Context**: Deeply analyze the 'company_context' to understand their business.\n2. **Generate List**: Create a list of 10-15 relevant items based on 'generation_type'.\n - If 'generation_type' is \"occupations\", generate specific job titles within the 'industry'.\n - If 'generation_type' is \"tasks\", generate specific, actionable tasks for the 'occupation'.\n3. **Output Format**: Your response MUST be a single, minified JSON object with one key: \"generated_list\".\n4. **List Structure**: The value must be an array of objects, each with \"name\" (string) and \"description\" (concise string).\n5. **Relevance is Key**: Items must be sharply focused and directly applicable to the company's use case.\n6. **Input is JSON**: You must parse the JSON input to get the context.`,\n EVALUATOR_AGENT:\n `You are a native JSON speaking, meticulous, and impartial AI QA agent. Given a task briefing, success criteria, and AI output, you will:\n1. **Score:** Score the output 1-10 on each criterion.\n2. **Critique:** Provide a concise, professional critique of its strengths and weaknesses.\n# SCORING RULES\n1. **Gut Check:** Before all else, if the output has obvious placeholders ('example.com', '[Insert Name Here]') or is extremely generic, penalize heavily on \"Plausibility and Specificity.\"\n2. **Penalize Non-Deliverables:** If the output is a plan or commentary instead of the final deliverable, penalize heavily (scores <= 3) on the relevant criterion.\n# OUTPUT FORMAT\nYour entire response MUST be a single, minified JSON object: {\"scores\": [{\"criterion\": \"...\", \"score\": ...}], \"critique\": \"...\"}. You MUST write the critique in the requested language. Add no other text or markdown.`,\nSESSION:\n `You are a top-tier \"Refiner\" agent. You have your original task briefing, your first draft, and a critique from a QA agent. Your SOLE mission is to rewrite the first draft to fully address every point in the critique, elevating it to a 10/10 standard.\n\n# RULES\n1. **Analyze and Integrate:** Apply all suggested changes from the critique to the first draft.\n2. **No Commentary:** Do not apologize, explain changes, or include meta commentary.\n3. **Final Deliverable Only:** Your output must be only the new, improved, final version.\n4. **Language:** Write the final output in the requested language.`,\n SUCCESS_CRITERIA_GENERATOR:\n `You are a native JSON speaking expert QA analyst. Your task is to generate a concise list of success criteria for evaluating an AI-generated document.\n# RULES\n1. **Analyze Task:** Deeply analyze the task briefing to understand the desired output's purpose and requirements.\n2. **Generate Criteria:** Create 4-6 specific, measurable, relevant success criteria as short strings.\n3. **Output Format:** Your response MUST be a single, minified JSON object: {\"criteria\": [\"...\", \"...\", ...]}.\n\n# TASK BRIEFING\n{{task_briefing}}`,\n },\n};\n\n// --- STATIC DATA ---\n// NO LONGER NEEDED - This is now dynamic\n// const taskCategories = [ ... ];\n\n// --- BACKEND SERVICES ---\nconst services = {\n async callOpenAI(\n systemPrompt: string,\n userContent: string | object,\n options: {\n model?: string;\n isJson?: boolean;\n stream?: boolean;\n c?: Context;\n } = {},\n ) {\n const {\n model = config.models.major,\n isJson = false,\n stream = false,\n c = null,\n } = options;\n const openai = new OpenAI();\n const messages: any[] = [{ role: \"system\", content: systemPrompt }, {\n role: \"user\",\n content: typeof userContent === \"string\"\n ? userContent\n : JSON.stringify(userContent),\n }];\n const requestPayload: any = { model, messages, stream };\n if (isJson) requestPayload.response_format = { type: \"json_object\" };\n try {\n const completion = await openai.chat.completions.create(requestPayload);\n if (stream && c) {\n return streamText(c, async (s) => {\n for await (const part of completion) {\n await s.write(part.choices[0]?.delta?.content || \"\");\n }\n });\n }\n if (isJson && c) {\n return c.json(JSON.parse(completion.choices[0].message.content));\n }\n return completion;\n } catch (e) {\n console.error(`Error calling OpenAI: ${e.message}`);\n throw new Error(\"AI service failed.\");\n }\n },\n};\n\n// --- FRONTEND GENERATION ---\nfunction generateHtml(sourceUrl: string): string {\n const languages = {\n \"English\": \"en-US\",\n \"Spanish\": \"es-ES\",\n \"French\": \"fr-FR\",\n \"German\": \"de-DE\",\n \"Japanese\": \"ja-JP\",\n \"Chinese (Simplified)\": \"zh-CN\",\n };\n const languageOptions = Object.entries(languages).map(([name, code]) =>\n ``\n ).join(\"\");\n const taskCategoriesJson = \"[]\"; // No longer passed to frontend\n return `Taskflow AI | Intelligent Task Automation

Great job!

You're making excellent progress.

Taskflow AI

`;\n}\n\n// --- HONO APPLICATION SETUP ---\nconst app = new Hono();\n\n// --- API ROUTES ---\n\napp.get(\"/\", (c: Context) => c.html(generateHtml(c.req.url)));\n\napp.post(\"/api/industries\", async (c: Context) => {\n const { company_context } = await c.req.json();\n if (!company_context) {\n return c.json({ error: \"Company context is required\" }, 400);\n }\n const userContent = `Company Context: ${JSON.stringify(company_context)}`;\n return services.callOpenAI(config.prompts.INDUSTRY_GENERATOR, userContent, {\n c,\n isJson: true,\n model: config.models.minor,\n });\n});\n\napp.post(\"/api/gather\", async (c: Context) => {\n const { task, domain, history } = await c.req.json();\n const historyString = history.map((h) =>\n `${h.role === \"assistant\" ? \"AI\" : \"User\"}: ${h.content}`\n ).join(\"\\n\");\n const prompt = config.prompts.INFORMATION_GATHERER.replace(\"{{task}}\", task)\n .replace(\"{{domain}}\", domain).replace(\n \"{{history}}\",\n historyString ||\n \"No history yet. Start with your most important question.\",\n );\n try {\n const completion = await services.callOpenAI(\n \"You are a helpful assistant following strict output rules.\",\n prompt,\n { model: config.models.minor },\n );\n const responseText = completion.choices[0].message.content.trim();\n if (responseText.startsWith(\"__COMPLETE__\")) {\n const jsonPart = responseText.replace(\"__COMPLETE__\", \"\");\n return c.json({ status: \"complete\", data: JSON.parse(jsonPart) });\n } else {\n return c.json({ status: \"in-progress\", question: responseText });\n }\n } catch (e) {\n console.error(\"Gathering agent failed:\", e);\n return c.json({\n status: \"in-progress\",\n question:\n \"I'm having a little trouble with that. Could you please rephrase your last answer?\",\n }, 500);\n }\n});\n\napp.post(\"/api/generate/:type\", async (c: Context) => {\n const { type } = c.req.param();\n if (type !== \"occupations\" && type !== \"tasks\") {\n return c.json({ error: \"Invalid generation type\" }, 400);\n }\n const body = await c.req.json();\n const payload = {\n generation_type: type,\n company_context: body.company_context,\n ...(type === \"occupations\" && { industry: body.industry }),\nSESSION };\n return services.callOpenAI(config.prompts.DYNAMIC_LIST_GENERATOR, payload, {\n c,\n isJson: true,\n });\n});\n\n// --- NEW ENDPOINT ---\napp.post(\"/api/generate/approaches\", async (c: Context) => {\n const { task, occupation, company_context } = await c.req.json();\n if (!task || !occupation || !company_context) {\n return c.json({ error: \"Task, occupation, and context are required\" }, 400);\n }\n const userContent = config.prompts.APPROACH_GENERATOR\n .replace(\"{{task}}\", task)\n .replace(\"{{occupation}}\", occupation)\n .replace(\"{{company_context}}\", JSON.stringify(company_context));\n\n return services.callOpenAI(\n \"You are a helpful assistant following strict output rules.\",\n userContent,\n { c, isJson: true, model: config.models.minor }, // Use minor model for speed\n );\n});\n// --- END NEW ENDPOINT ---\n\napp.post(\"/api/prompt/dynamic\", async (c: Context) => {\n const { occupation_title, task, company_context } = await c.req.json<\n DynamicPromptBody\n >();\n const userContent = `Company Context: ${\n JSON.stringify(company_context)\n }\\n\\nOccupation: ${occupation_title}\\n\\nTask: ${task}`;\nsource:\n isJson: true,\n });\n\n\n// --- HEAVILY MODIFIED ENDPOINT ---\napp.post(\"/api/execute-stream\", async (c: Context) => {\n const {\n refined_prompt, // This is the \"Task\" (e.g., \"Write a Q4 sales report\")\n user_inputs, // This is the JSON from the questionnaire\n company_context,\n category_object, // This is our new dynamic object\n } = await c.req.json();\n\n if (!company_context || !category_object || !refined_prompt || !user_inputs) {\n return c.json({ error: \"Required fields missing\" }, 400);\n }\n\n // 1. Get the prompt template from the approach object\n let systemPrompt = category_object.system_prompt_template;\n\n // 2. Inject all our context into the template\n systemPrompt = systemPrompt\n .replace(/\\{\\{\\s*company_name\\s*\\}\\}/g, company_context.company_name)\n .replace(\n /\\{\\{\\s*company_description\\s*\\}\\}/g,\n company_context.company_description,\n )\n\n .replace(/\\{\\{\\s*use_case\\s*\\}\\}/g, company_context.use_case)\n .replace(/\\{\\{\\s*refined_task\\s*\\}\\}/g, refined_prompt); // Pass the task itself\n\n // 3. Inject the user's questionnaire answers\n systemPrompt = systemPrompt\n .replace(\n /\\{\\{\\s*user_inputs_json\\s*\\}\\}/g,\n JSON.stringify(user_inputs, null, 2),\n );\n\n // 4. The user prompt is now just a simple \"go\" command.\n const userPrompt = \"Proceed with the task defined in your system instructions.\";\n\n return services.callOpenAI(systemPrompt, userPrompt, { c, stream: true });\n});\n// --- END MODIFIED ENDPOINT ---\n\napp.post(\"/api/evaluate\", async (c: Context) => {\nSESSION:\n return c.json({\n error: \"Briefing, first_draft, and company_context are required\",\n }, 400);\n }\n try {\n const criteriaPrompt = config.prompts.SUCCESS_CRITERIA_GENERATOR.replace(\nIA:\n briefing,\n );\n const criteriaCompletion = await services.callOpenAI(\n \"You are a helpful assistant following strict output rules.\",\n criteriaPrompt,\n { model: config.models.minor, isJson: true },\n );\n const success_criteria =\n A JSON.parse(criteriaCompletion.choices[0].message.content).criteria;\n const userContent = {\n task_briefing: briefing,\n success_criteria: success_criteria,\n raw_output: first_draft,\n };\n return services.callOpenAI(config.prompts.EVALUATOR_AGENT, userContent, {\n c,\n isJson: true,\n model: config.models.minor,\n });\n } catch (e) {\n console.error(\"Evaluation failed:\", e);\nNB { error: `Failed to evaluate the document. ${e.message}` },\n 500,\n );\n }\n});\n\napp.post(\"/api/refine\", async (c: Context) => {\n const { briefing, first_draft, critique, language } = await c.req.json();\n const userContent = `# YOUR TASK\nRewrite your first draft to fully address every point in the critique, elevating the work to a 10/10 standard. Return only the improved final deliverable. Do not write an apology, explanation, or any meta-commentary. Produce the new, superior output in the requested language: ${language}.\n\n## Original Briefing\n${briefing}\n\n## Your First Draft\n${first_draft}\n\n## QA Critique\n${critique}`;\n return services.callOpenAI(config.prompts.REFINER_AGENT, userContent, {\n c,\n stream: true,\n });\n});\n\nexport default app.fetch;","// @ts-nocheck\n// Taskflow AI - Professional Autonomous Agent Factory\n// A modern, user-friendly platform for AI-powered task execution\n// ✅ [v2 UI OVERHAUL COMPLETE]\n// ✅ [v3 DYNAMIC APPROACHES COMPLETE]\n\nimport { Hono } from \"npm:hono@4.4.12\";\nimport { OpenAI } from \"https://esm.town/v/std/openai\";\nimport type { Context } from \"npm:hono@4.4.12\";\nimport { streamText } from \"npm:hono@4.4.12/streaming\";\n\n// --- TYPE DEFINITIONS ---\ninterface CompanyContext {\n company_name: string;\n company_description: string;\n use_case: string;\n language: string;\n industry?: string;\n}\ninterface ConversationHistory {\n role: \"assistant\" | \"user\";\n content: string;\n}\ninterface GatherBody {\n task: string;\n domain: string;\n history: ConversationHistory[];\n}\ninterface DynamicPromptBody {\n occupation_title: string;\n task: string;\n company_context: CompanyContext;\n}\ninterface ExecuteBody {\n refined_prompt: string;\n user_inputs: Record;\n company_context: CompanyContext;\n category_object: any; // Changed from 'category'\n}\n\n// --- CENTRALIZED CONFIGURATION ---\nconst config = {\n models: {\n major: \"gpt-4o\",\n minor: \"gpt-4o-mini\",\n },\n prompts: {\n INFORMATION_GATHERER:\n `You are a native JSON speaking expert information gatherer designed to acquire all necessary details for a given task and domain. Your goal is to ask the fewest, most concise, and natural questions possible, one at a time. If information is missing or refused, make a reasonable assumption or broaden your scope. Iteratively ask questions and integrate responses until all required information for the {{task}} within the {{domain}} is gathered.\n\n# RULES\n1. **Primary Goal**: Ask single, concise questions to gather the minimum required information for the task, aiming for completion efficiently.\n2. **Decision Point**: After analyzing the conversation history, decide if you have enough information.\n - **If YES**: Your immediate response MUST be the token \"__COMPLETE__\" followed by a single, minified JSON object of all gathered data. Do not ask for confirmation. Example: \\`__COMPLETE__\\`{\"party_date\":\"next Saturday\",\"guests\":50,\"location\":\"rented hall\"}\n - **If NO**: You are missing critical information. Proceed to the next rule.\n3. **Question Formulation**: If more info is needed, formulate the single most impactful and concise question.\n4. **Output Format (Questions)**: The question must be a short, direct interrogative. DO NOT add conversational filler, commentary, apologies, or instructions.\n\n# CONTEXT\n- Task: {{task}}\n- Domain: {{domain}}\n- Conversation History:\n{{history}}`,\n PROMPT_REFINER:\n `You are a native JSON speaking prompt engineer. Transmute raw occupational data into a concise, forceful directive for a target AI. The AI must perform the task end-to-end and return only the finished deliverable(s)—never a plan. You will receive an \"Occupation,\" \"Task,\" and \"Company Context.\" Your prompt must be a production-grade instruction that directly yields the final artifact. CORE RULE: You MUST return a prompt with {{variables}}. Your goal is to identify important, undefined entities in the Task and convert them to descriptive, {{snake_case_placeholders}}. You MUST include at least 2 placeholders. INSTRUCTIONS: 1. Persona: Start with \"You are a top 0.1% [Occupation]...\" for an expert tone. 2. Objective: Translate the Task into a formal, actionable objective implying completion. 3. Context: Use the 'Company Context' to make the objective specific (e.g., \"Your mandate for [Company Name] is to...\"). 4. Placeholders: Convert only USER-SUPPLIED functional inputs into {{snake_case_variables}}. 5. Delivery Rules: Require the AI to: - Return only the finished deliverable(s) now. - Exclude plans, methods, or explanations. - Avoid meta-commentary, caveats, or apologies. - If details are missing, make minimal, reasonable assumptions and proceed, optionally adding a short \"Assumptions\" list (<=5 bullets). - Use production-ready formatting (e.g., ready-to-send copy; runnable code in fenced blocks). - Never include sections like \"Plan\", \"Approach\", \"Methodology\". 6. Output Format: Specify the exact output structure. 7. Mandate: Conclude with a clear call to action like \"Produce the deliverable now.\" Your entire response MUST be a single, minified JSON object with one key, \"refined_prompt\". No other text or markdown.`,\n MASTER_PROMPT_ENGINEER: // This is now used by the \"prompt_building\" dynamic approach\n `You are a native JSON speaking master prompt architect. Your objective is to construct a comprehensive, self-contained blueprint for another AI to execute a task flawlessly, not to do it yourself.\n\n# INSTRUCTIONS\n1. **Deconstruct Task:** Analyze the task to identify sub-tasks, knowledge domains, constraints, and desired output formats.\n2. **Establish Persona:** Define an expert persona for the target AI (e.g., \"You are a tenured professor of economics...\").\n3. **Provide Context:** Synthesize company context into a rich narrative explaining the task's strategic importance.\n4. **Formulate Logic:** Create a clear, logical sequence of operations using headings and lists.\n5. **Set Negative Constraints:** Explicitly state what the AI should *not* do (e.g., \"Do not use informal language,\" \"Do not output a plan\").\n6. **Define Output Schema:** Meticulously define the final output's structure, format, length, and tone.\n7. **Include Examples:** Where helpful, provide concise examples of good and bad outputs.\n8. **Final Mandate:** Conclude with a clear command to initiate execution.\n\nYour output must be the prompt itself, as a clean markdown document. Do not wrap it in JSON or add commentary.`,\n// --- NEW PROMPT ---\n APPROACH_GENERATOR:\n `You are a native JSON speaking expert AI system architect. Your task is to generate 3-5 distinct, professional \"Approaches\" for completing a given task, tailored to a specific company and occupation.\n\n# RULES\n1. **Analyze Input**: You will receive a Task, Occupation, and Company Context.\n2. **Generate Approaches**: Create a list of 3-5 unique, actionable approaches an expert would take.\n3. **Define Output Object**: Each approach in the list MUST be a JSON object with this exact structure:\n - \"name\": (string) A short, \"button-friendly\" name (e.g., \"Formal Business Plan\", \"Quick Email Script\").\n - \"description\": (string) A 1-sentence description of this approach.\n - \"icon\": (string) A single, relevant emoji.\n - \"system_prompt_template\": (string) A complete, production-ready system prompt *template* for another AI to execute this approach. This template MUST include placeholders like {{company_name}}, {{company_description}}, {{language}}, {{use_case}}, {{refined_task}}, and {{user_inputs_json}} where appropriate.\n4. **Template is Key**: The \"system_prompt_template\" is the *most important part*. It is the \"DNA\" of the approach, defining the persona, rules, context, and final deliverable format for the AI that will do the work.\n5. **Output Format**: Your entire response MUST be a single, minified JSON object: {\"approaches\": [...]}\n\n# EXAMPLES OF system_prompt_template\n## Example 1: For a \"Formal Document\" approach\n\"You are a top 0.1% {{occupation}}. Your task is to produce a formal, production-ready document based on the user's request.\n# CONTEXT\n- Company: {{company_name}} ({{company_description}})\n- Goal: {{use_case}}\n- Language: {{language}}\n# TASK: {{refined_task}}\n# USER SPECIFICATIONS:\n{{user_inputs_json}}\n# RULES\n1. Return only the finished deliverable(s).\n2. Exclude plans, methods, or explanations.\n3. Use formal, professional language.\n4. Output MUST be in {{language}}.\nProduce the deliverable now.\"\n\n## Example 2: For a \"Code Generation\" approach\n\"You are an elite 10x software engineer. Your task is to write clean, efficient, and well-commented code.\n# TASK: {{refined_task}}\n# USER SPECIFICATIONS:\n{{user_inputs_json}}\n# RULES\n1. Return ONLY the runnable code in fenced markdown blocks.\n2. Do not write explanations before or after the code.\n3. Adhere to all modern best practices.\n4. All text output (comments, etc.) MUST be in {{language}}.\nProduce the code now.\"\n\n## Example 3: For a \"Prompt Engineering\" approach\n\"${config.prompts.MASTER_PROMPT_ENGINEER.replace(/\"/g, '\\\\\"')}\"\n\n# INPUT\n- Task: {{task}}\n- Occupation: {{occupation}}\n- Company Context: {{company_context}}`,\n// --- END NEW PROMPT ---\n INDUSTRY_GENERATOR:\n `You are a native JSON speaking expert business analyst. Generate a list of professional industries most relevant to the given company.\n\n# RULES\n1. **Analyze Context**: Deeply analyze the provided company context (name, description, use case) to understand their core business.\n2. **Generate Industries**: Produce a list of 10-12 diverse yet highly relevant industries.\n3. **Output Format**: Your entire response MUST be a single, minified JSON object with one key: \"industries\". The value must be an array of strings.\n\n# COMPANY CONTEXT\n{{company_context}}`,\n DYNAMIC_LIST_GENERATOR:\n `You are a native JSON speaking expert AI that generates context-aware lists for professional applications. You will receive a JSON object with company context and the list type to generate.\n\n# RULES\n1. **Analyze Context**: Deeply analyze the 'company_context' to understand their business.\n2. **Generate List**: Create a list of 10-15 relevant items based on 'generation_type'.\n - If 'generation_type' is \"occupations\", generate specific job titles within the 'industry'.\n - If 'generation_type' is \"tasks\", generate specific, actionable tasks for the 'occupation'.\n3. **Output Format**: Your response MUST be a single, minified JSON object with one key: \"generated_list\".\n4. **List Structure**: The value must be an array of objects, each with \"name\" (string) and \"description\" (concise string).\n5. **Relevance is Key**: Items must be sharply focused and directly applicable to the company's use case.\n6. **Input is JSON**: You must parse the JSON input to get the context.`,\n EVALUATOR_AGENT:\n `You are a native JSON speaking, meticulous, and impartial AI QA agent. Given a task briefing, success criteria, and AI output, you will:\n1. **Score:** Score the output 1-10 on each criterion.\n2. **Critique:** Provide a concise, professional critique of its strengths and weaknesses.\n# SCORING RULES\n1. **Gut Check:** Before all else, if the output has obvious placeholders ('example.com', '[Insert Name Here]') or is extremely generic, penalize heavily on \"Plausibility and Specificity.\"\n2. **Penalize Non-Deliverables:** If the output is a plan or commentary instead of the final deliverable, penalize heavily (scores <= 3) on the relevant criterion.\n# OUTPUT FORMAT\nYour entire response MUST be a single, minified JSON object: {\"scores\": [{\"criterion\": \"...\", \"score\": ...}], \"critique\": \"...\"}. You MUST write the critique in the requested language. Add no other text or markdown.`,\nSESSION:\n `You are a top-tier \"Refiner\" agent. You have your original task briefing, your first draft, and a critique from a QA agent. Your SOLE mission is to rewrite the first draft to fully address every point in the critique, elevating it to a 10/10 standard.\n\n# RULES\n1. **Analyze and Integrate:** Apply all suggested changes from the critique to the first draft.\n2. **No Commentary:** Do not apologize, explain changes, or include meta commentary.\n3. **Final Deliverable Only:** Your output must be only the new, improved, final version.\n4. **Language:** Write the final output in the requested language.`,\n SUCCESS_CRITERIA_GENERATOR:\n `You are a native JSON speaking expert QA analyst. Your task is to generate a concise list of success criteria for evaluating an AI-generated document.\n# RULES\n1. **Analyze Task:** Deeply analyze the task briefing to understand the desired output's purpose and requirements.\n2. **Generate Criteria:** Create 4-6 specific, measurable, relevant success criteria as short strings.\n3. **Output Format:** Your response MUST be a single, minified JSON object: {\"criteria\": [\"...\", \"...\", ...]}.\n\n# TASK BRIEFING\n{{task_briefing}}`,\n },\n};\n\n// --- STATIC DATA ---\n// NO LONGER NEEDED - This is now dynamic\n// const taskCategories = [ ... ];\n\n// --- BACKEND SERVICES ---\nconst services = {\n async callOpenAI(\n systemPrompt: string,\n userContent: string | object,\n options: {\n model?: string;\n isJson?: boolean;\n stream?: boolean;\n c?: Context;\n } = {},\n ) {\n const {\n model = config.models.major,\n isJson = false,\n stream = false,\n c = null,\n } = options;\n const openai = new OpenAI();\n const messages: any[] = [{ role: \"system\", content: systemPrompt }, {\n role: \"user\",\n content: typeof userContent === \"string\"\n ? userContent\n : JSON.stringify(userContent),\n }];\n const requestPayload: any = { model, messages, stream };\n if (isJson) requestPayload.response_format = { type: \"json_object\" };\n try {\n const completion = await openai.chat.completions.create(requestPayload);\n if (stream && c) {\n return streamText(c, async (s) => {\n for await (const part of completion) {\n await s.write(part.choices[0]?.delta?.content || \"\");\n }\n });\n }\n if (isJson && c) {\n return c.json(JSON.parse(completion.choices[0].message.content));\n }\n return completion;\n } catch (e) {\n console.error(`Error calling OpenAI: ${e.message}`);\n throw new Error(\"AI service failed.\");\n }\n },\n};\n\n// --- FRONTEND GENERATION ---\nfunction generateHtml(sourceUrl: string): string {\n const languages = {\n \"English\": \"en-US\",\n \"Spanish\": \"es-ES\",\n \"French\": \"fr-FR\",\n \"German\": \"de-DE\",\n \"Japanese\": \"ja-JP\",\n \"Chinese (Simplified)\": \"zh-CN\",\n };\n const languageOptions = Object.entries(languages).map(([name, code]) =>\n ``\n ).join(\"\");\n const taskCategoriesJson = \"[]\"; // No longer passed to frontend\n return `Taskflow AI | Intelligent Task Automation

Great job!

You're making excellent progress.

You're doing amazing! ✨

Welcome to Taskflow AI

What makes Taskflow AI special?

Quick Setup

Almost there!