AI Business Assistant

Thinking

Final Results

Execution Logs

\n \n \n \n `;\n}\n\nfunction clientQuestionnaireJs() {\n return `\n const messagesDiv = document.getElementById('messages');\n const inputArea = document.getElementById('input-area');\n const loadingIndicator = document.getElementById('loading-indicator');\n const errorDisplay = document.getElementById('error-display');\n const resultsDisplay = document.getElementById('results-display');\n const resultsContent = document.getElementById('results-content');\n const logsContent = document.getElementById('logs-content');\n\n let currentSessionId = null;\n let currentTask = null;\n let currentQuestion = null;\n let currentServerState = null;\n\n function showTaskSelection() {\n clearInputArea();\n if (!currentSessionId || resultsDisplay.style.display === 'block') {\n addBotMessage(\"Hello! What can I help you with today?\");\n } else {\n addBotMessage(\"What else can I help you with?\");\n }\n enableInteraction(false);\n resultsDisplay.style.display = 'none';\n\n const select = document.createElement('select');\n select.id = 'task-select';\n select.classList.add('form-control');\n select.innerHTML = \\`\n \n \n \n \n \n \n \n \\`;\n\n const button = document.createElement('button');\n button.id = 'task-submit';\n button.textContent = 'Start';\n button.onclick = handleTaskSelection;\n\n inputArea.appendChild(select);\n inputArea.appendChild(button);\n enableInteraction(true);\n select.focus();\n }\n\n async function handleTaskSelection() {\n const select = document.getElementById('task-select');\n const task = select.value;\n if (!task) {\n showError(\"Please select a task first.\");\n return;\n }\n clearError();\n currentTask = task;\n addUserMessage(\\`I need help with: \\${select.options[select.selectedIndex].text}\\`);\n clearInputArea();\n enableInteraction(false);\n setLoading(true);\n\n await sendToServer({\n task: currentTask,\n sessionId: currentSessionId,\n lastAnswer: null,\n state: currentServerState\n });\n }\n\n function addMessage(text, sender) {\n const msgDiv = document.createElement('div');\n msgDiv.classList.add('message', sender);\n text = text.replace(/\\\\[(.*?)\\\\]\\\$(.*?)\\\$/g, '$1');\n msgDiv.innerHTML = text;\n messagesDiv.appendChild(msgDiv);\n setTimeout(() => { messagesDiv.scrollTop = messagesDiv.scrollHeight; }, 50);\n }\n function addBotMessage(text) { addMessage(escapeHtml(text), 'bot'); }\n function addUserMessage(text) { addMessage(escapeHtml(text), 'user'); }\n\n function clearInputArea() {\n inputArea.innerHTML = '';\n }\n\n function enableInteraction(enabled) {\n const inputEls = inputArea.querySelectorAll('input, select, textarea');\n const buttonEls = inputArea.querySelectorAll('button');\n inputEls.forEach(el => el.disabled = !enabled);\n buttonEls.forEach(el => el.disabled = !enabled);\n }\n\n function setupInputForQuestion(question) {\n clearInputArea();\n currentQuestion = question;\n let inputElement;\n if (question.type === 'select') {\n inputElement = document.createElement('select');\n inputElement.id = 'user-input';\n inputElement.classList.add('form-control');\n (question.options || []).forEach(opt => {\n const option = document.createElement('option');\n option.value = opt.value;\n option.textContent = opt.label;\n inputElement.appendChild(option);\n });\n } else if (question.type === 'textarea') {\n inputElement = document.createElement('textarea');\n inputElement.id = 'user-input';\n inputElement.rows = 3;\n inputElement.placeholder = \"Your answer...\";\n inputElement.classList.add('form-control');\n inputElement.addEventListener('keypress', function(event) {\n if (event.key === 'Enter' && !event.shiftKey) {\n event.preventDefault();\n handleSend();\n }\n });\n } else {\n inputElement = document.createElement('input');\n inputElement.type = 'text';\n inputElement.id = 'user-input';\n inputElement.placeholder = \"Your answer...\";\n inputElement.classList.add('form-control');\n inputElement.addEventListener('keypress', function(event) {\n if (event.key === 'Enter') {\n event.preventDefault();\n handleSend();\n }\n });\n }\n\n const sendButton = document.createElement('button');\n sendButton.id = 'send-button';\n sendButton.textContent = 'Send';\n sendButton.onclick = handleSend;\n\n inputArea.appendChild(inputElement);\n inputArea.appendChild(sendButton);\n enableInteraction(true);\n inputElement.focus();\n }\n\n async function handleSend() {\n const inputElement = document.getElementById('user-input');\n if (!inputElement || !currentQuestion) return;\n const answerValue = inputElement.value;\n clearError();\n\n let displayValue = answerValue;\n if (currentQuestion.type === 'select') {\n const selectedOption = Array.from(inputElement.options).find(opt => opt.selected || opt.value === answerValue);\n displayValue = selectedOption ? selectedOption.text : answerValue;\n if (answerValue === \"\") {\n displayValue = \"(Skipped optional question)\";\n }\n } else if (!answerValue.trim()) {\n displayValue = \"(Provided no input)\";\n }\n addUserMessage(displayValue);\n\n const lastAnswer = {\n field: currentQuestion.field,\n value: answerValue\n };\n\n clearInputArea();\n enableInteraction(false);\n setLoading(true);\n\n await sendToServer({\n task: currentTask,\n sessionId: currentSessionId,\n lastAnswer: lastAnswer,\n state: currentServerState\n });\n }\n\n async function sendToServer(payload) {\n try {\n console.log(\"Sending to /questionnaire:\", payload);\n const response = await fetch('/questionnaire', {\n method: 'POST',\n headers: { 'Content-Type': 'application/json' },\n body: JSON.stringify(payload),\n });\n\n setLoading(false);\n clearError();\n\n if (!response.ok) {\n let errorMsg = \\`Server error: \\${response.status} \\${response.statusText}\\`;\n try {\n const errorData = await response.json();\n errorMsg += \\` - \\${errorData.error || errorData.message || JSON.stringify(errorData)}\\`;\n } catch(e) { /* Ignore */ }\n throw new Error(errorMsg);\n }\n\n const data = await response.json();\n console.log(\"Received from /questionnaire:\", data);\n\n currentSessionId = data.sessionId;\n currentServerState = data.state;\n\n if (data.error) {\n showError(data.error);\n if (currentQuestion) {\n addBotMessage(\\`Sorry, there was an issue with that answer: \\${data.error} Let's try that again.\\`);\n setupInputForQuestion(currentQuestion);\n } else {\n addBotMessage(\\`Sorry, an error occurred: \\${data.error}. Please try selecting a task again.\\`);\n showTaskSelection();\n }\n return;\n }\n\n if (data.isComplete) {\n addBotMessage(\"Okay, I have all the information I need! Processing your request...\");\n enableInteraction(false);\n clearInputArea();\n setLoading(true);\n await executeWorkflow(data.gatheredData);\n } else if (data.nextQuestion) {\n addBotMessage(data.nextQuestion.text);\n setupInputForQuestion(data.nextQuestion);\n } else {\n showError(\"Received an unexpected response from the server.\");\n enableInteraction(false);\n addBotMessage(\"Something unexpected happened. Please refresh to start over.\");\n }\n\n } catch (error) {\n console.error(\"Communication error:\", error);\n setLoading(false);\n showError(\\`Communication error: \\${error.message}\\`);\n enableInteraction(false);\n addBotMessage(\"Sorry, a connection problem occurred. You might need to refresh the page to start over.\");\n }\n }\n\n async function executeWorkflow(finalPayload) {\n enableInteraction(false);\n try {\n console.log(\"Sending to /execute:\", finalPayload);\n if (!finalPayload || !finalPayload.task || !finalPayload.data) {\n throw new Error(\"Invalid payload structure received from questionnaire for execution.\");\n }\n\n const response = await fetch('/execute', {\n method: 'POST',\n headers: { 'Content-Type': 'application/json' },\n body: JSON.stringify(finalPayload),\n });\n\n setLoading(false);\n const resultData = await response.json();\n console.log(\"Execution Result:\", resultData);\n\n resultsDisplay.style.display = 'block';\n resultsContent.innerHTML = '';\n logsContent.innerHTML = '';\n\n if (!response.ok || resultData.error) {\n const errorMsg = resultData.error || \\`Execution failed with status \\${response.status}\\`;\n resultsContent.innerHTML = \\`

\\${escapeHtml(errorMsg)}

\\`;\n if (resultData.details) {\n resultsContent.innerHTML += \\`

Details: \\${escapeHtml(JSON.stringify(resultData.details))}

\\`;\n }\n addBotMessage(\\`Sorry, there was an error processing your request: \\${escapeHtml(errorMsg)}\\`);\n } else {\n resultsContent.innerHTML = formatResults(resultData.result, finalPayload.task);\n addBotMessage(\"I've completed your request! You can see the results below.\");\n }\n logsContent.innerHTML = formatLogs(resultData.debugLog);\n showTaskSelection();\n\n } catch (error) {\n console.error(\"Execution or result processing error:\", error);\n setLoading(false);\n resultsDisplay.style.display = 'block';\n resultsContent.innerHTML = \\`

Execution process failed: \\${escapeHtml(error.message)}

\\`;\n logsContent.innerHTML = formatLogs(null);\n addBotMessage(\"Sorry, there was a critical error trying to complete your request.\");\n showTaskSelection();\n }\n }\n\n function setLoading(isLoading) { loadingIndicator.style.display = isLoading ? 'block' : 'none'; }\n function showError(message) { errorDisplay.textContent = message; errorDisplay.style.display = 'block'; }\n function clearError() { errorDisplay.textContent = ''; errorDisplay.style.display = 'none'; }\n\n function escapeHtml(unsafe) {\n if (unsafe === null || unsafe === undefined) return '';\n let str = unsafe.toString();\n const map = { '&': '&', '<': '<', '>': '>', '\"': '"', \"'\": ''' };\n return str.replace(/[&<>\\\"']/g, function(m) { return map[m]; });\n }\n\n function formatResults(result, task) {\n let html = '';\n if (!result) {\n html = '

No specific results generated, but the process completed.

';\n } else if (typeof result === 'string') {\n html = \\`

\\${escapeHtml(result)}

\\`;\n } else if (task === 'create_content' && result.content) {\n html = \\`

Generated Content:

\\${escapeHtml(result.content)}

\\`;\n if (result.suggestions) html += \\`

Suggestions:

\\${escapeHtml(result.suggestions)}

\\`;\n } else if (task === 'develop_strategy' && result.strategy) {\n html = \\`

Marketing Strategy:

\\${escapeHtml(result.strategy)}

\\`;\n if (result.adIdeas && result.adIdeas.length > 0) html += \\`

Ad Ideas:

\\${escapeHtml(idea)}

\\`;\n if (result.audienceSuggestions && result.audienceSuggestions.length > 0) html += \\`

Audience Suggestions:

\\${escapeHtml(s)}

\\`;\n } else if (task === 'generate_logo_concept') {\n html = '

Logo Concepts & Descriptions:

';\n if (result.concepts && result.concepts.length > 0) {\n result.concepts.forEach((concept, index) => { html += \\`

Concept \\${index + 1}

\\${escapeHtml(concept)}

\\`; });\n } else { html += '

No concepts generated.

'; }\n if (result.promptsForImageModel && result.promptsForImageModel.length > 0) html += \\`

Prompts for Image Models (e.g., DALL-E, Midjourney):

```
\\${escapeHtml(p)}
```

\\`;\n if (result.svgAttempt) html += \\`

Basic SVG Attempt (Experimental):

\\${escapeHtml(result.svgAttempt)}

(Note: SVG quality from LLMs varies greatly. Copy and save as a .svg file to view)

\\`;\n html += '

Note: These are concepts. Use them as inspiration or with an image generation tool/designer.

';\n } else if (task === 'build_brand_identity') {\n html = \\`

Brand Voice:

\\${escapeHtml(result.brandVoice || 'Not specified')}

\\`;\n if (result.keyMessages && result.keyMessages.length > 0) html += \\`

Key Messages:

\\${escapeHtml(msg)}

\\`;\n html += \\`

Color Palette Suggestions:

\\${escapeHtml(result.colorPaletteSuggestions || 'Not specified')}

\\`;\n html += \\`

Typography Suggestions:

\\${escapeHtml(result.typographySuggestions || 'Not specified')}

\\`;\n } else if (task === 'run_simulation' && result.simulationResult) {\n html = \\`

Simulation Result:

\\${escapeHtml(result.simulationResult)}

\\`;\n if (result.analysis) html += \\`

Analysis:

\\${escapeHtml(result.analysis)}

\\`;\n } else if (task === 'meta_prompt' && result.system_prompt) {\n html = \\`

Generated System Prompt:

\\${escapeHtml(result.system_prompt)}

\\`;\n } else {\n html = \\`

Raw Results:

\\${escapeHtml(JSON.stringify(result, null, 2))}

\\`;\n }\n return html;\n }\n\n function formatLogs(logs) {\n if (!logs || !Array.isArray(logs) || logs.length === 0) {\n return '

No execution logs generated or available.

';\n }\n return logs.map(log => {\n let detailsString = '';\n if (log.details) {\n try { detailsString = escapeHtml(JSON.stringify(log.details)); }\n catch (e) { detailsString = escapeHtml(\\`[Error stringifying details: \\${e.message}]\\`); }\n }\n return \\`

\n [\\${log.timestamp || '?'}]\n [\\${log.level || '?'}]\n [\\${log.component || 'System'}]\n \\${escapeHtml(log.message || '')}\n \\${detailsString ? \\` | Details: \\${detailsString}\\` : ''}\n

\\`;\n }).join('');\n }\n showTaskSelection();\n `;\n}\n\nclass MockLoggingAgent {\n logs: LogEntry[];\n mandateIdContext?: string;\n\n constructor(mandateId?: string) {\n this.logs = [];\n this.mandateIdContext = mandateId;\n }\n\n log(level: LogLevel, component: string, message: string, details?: any, mandateId?: string, taskId?: string) {\n const effectiveMandateId = mandateId || this.mandateIdContext || \"N/A\";\n let safeDetails = undefined;\n if (details !== undefined) {\n try {\n safeDetails = JSON.parse(JSON.stringify(details));\n }\n catch (e) {\n safeDetails = `[Error logging details: ${e.message}]`;\n }\n }\n const entry: LogEntry = {\n timestamp: new Date().toISOString(),\n mandateId: effectiveMandateId,\n taskId: taskId,\n level,\n component,\n message,\n details: safeDetails,\n };\n this.logs.push(entry);\n const detailsString = safeDetails ? `| Details: ${JSON.stringify(safeDetails)}` : \"\";\n console.log(\n `[${entry.timestamp}] [${entry.mandateId}] [${\n entry.taskId || \"---\"\n }] [${entry.level}] [${entry.component}] ${entry.message} ${detailsString}`,\n );\n }\n\n getLogs(): LogEntry[] {\n try {\n return JSON.parse(JSON.stringify(this.logs));\n }\n catch (e) {\n console.error(\"Failed to deep copy logs:\", e);\n return [...this.logs];\n }\n }\n\n clearLogs() {\n this.logs = [];\n }\n setMandateIdContext(mandateId: string) {\n this.mandateIdContext = mandateId;\n }\n}\n\nasync function callOpenAI(\n systemPrompt: string,\n userPrompt: string,\n mandateId: string,\n taskId: string,\n log: LogFunction,\n): Promise {\n log(\n \"DEBUG\",\n \"callOpenAI\",\n `Calling OpenAI for task ${taskId}`,\n { sysLen: systemPrompt.length, usrLen: userPrompt.length },\n mandateId,\n taskId,\n );\n try {\n const openai = new OpenAI();\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o-mini\",\n messages: [{ role: \"system\", content: systemPrompt }, { role: \"user\", content: userPrompt }],\n temperature: 0.7,\n max_tokens: 2048,\n });\n const resultText = completion.choices[0]?.message?.content;\n const usage = completion.usage;\n if (!resultText) {\n log(\"WARN\", \"callOpenAI\", `OpenAI returned no content for task ${taskId}.`, { usage }, mandateId, taskId);\n return null;\n }\n log(\n \"DEBUG\",\n \"callOpenAI\",\n `OpenAI call successful for task ${taskId}`,\n { resLen: resultText.length, usage },\n mandateId,\n taskId,\n );\n return resultText.trim();\n } catch (error) {\n const errorDetails = { msg: error.message, name: error.name, status: error.status, type: error.type };\n log(\"ERROR\", \"callOpenAI\", `OpenAI API call failed for task ${taskId}`, { error: errorDetails }, mandateId, taskId);\n throw new Error(`OpenAI API error: ${error.message}`);\n }\n}\n\nclass QuestionnaireAgent {\n log: LogFunction;\n\n constructor(log: LogFunction) {\n this.log = log;\n }\n\n private getRequiredFieldsForTask(task: UserInput[\"task\"]): string[] {\n const baseReq = [\"businessType\", \"serviceDescription\", \"targetAudience\"];\n const optBase = [\"brandTone\", \"goals\"];\n const taskReq: Partial> = {\n \"create_content\": [\"contentType\", \"platform\"],\n \"develop_strategy\": [\"platforms\"],\n \"generate_logo_concept\": [\"businessName\"],\n \"build_brand_identity\": [\"businessName\"],\n \"run_simulation\": [\"simulationType\", \"specificParams\"],\n \"meta_prompt\": [\"prompt_input\"],\n };\n const taskOpt: Partial> = {\n \"create_content\": [\"specificInstructions\"],\n \"develop_strategy\": [\"budgetLevel\", \"duration\"],\n \"generate_logo_concept\": [\"stylePreferences\", \"colorPreferences\"],\n \"build_brand_identity\": [\"coreValues\"],\n };\n if (task === \"meta_prompt\") { return taskReq[task] || []; }\n const required = [...baseReq, ...(taskReq[task] || [])];\n const optional = [...optBase, ...(taskOpt[task] || [])];\n return [...new Set([...required, ...optional])];\n }\n\n private selectNextField(fields: string[], gathered: Record): string | null {\n for (const field of fields) {\n if (\n !(field in gathered) || gathered[field] === null || gathered[field] === undefined || gathered[field] === \"\"\n ) { return field; }\n }\n return null;\n }\n\n private formulateQuestion(field: string, currentState: QuestionnaireState): Question {\n const qId = `${field}-${Date.now()}`;\n let text = `What is your ${field}?`;\n let type: Question[\"type\"] = \"text\";\n let options: Question[\"options\"] | undefined = undefined;\n const bizName = currentState.gatheredAnswers.businessName || \"your business\";\n\n switch (field) {\n case \"businessType\":\n text = \"First, what type of business do you run? (e.g., Cafe, Plumber, Online Coach, Graphic Designer)\";\n break;\n case \"serviceDescription\":\n text = `Great! Could you briefly describe the main products or services ${bizName} offers?`;\n type = \"textarea\";\n break;\n case \"targetAudience\":\n text =\n `Who is your ideal customer or target audience for these offerings? (e.g., Local families, Small business owners, Tech startups)`;\n type = \"textarea\";\n break;\n case \"brandTone\":\n text = \"How would you describe your desired brand tone or personality? (Optional - choose one or leave blank)\";\n type = \"select\";\n options = [\n { label: \"Select Tone (Optional)\", value: \"\" },\n { label: \"Professional & Authoritative\", value: \"professional\" },\n { label: \"Friendly & Approachable\", value: \"friendly\" },\n { label: \"Casual & Conversational\", value: \"casual\" },\n { label: \"Playful & Humorous\", value: \"humorous\" },\n { label: \"Informative & Educational\", value: \"informative\" },\n { label: \"Formal & Reserved\", value: \"formal\" },\n { label: \"Enthusiastic & Energetic\", value: \"enthusiastic\" },\n ];\n break;\n case \"goals\":\n text =\n \"What are the main goals for this specific task or marketing effort? (Optional, e.g., Increase website traffic, Generate leads, Build brand awareness, Announce a new product)\";\n type = \"textarea\";\n break;\n case \"businessName\":\n text = \"What is the exact name of your business?\";\n break;\n case \"contentType\":\n text = \"What type of content do you need help creating?\";\n type = \"select\";\n options = [\n { label: \"-- Select Content Type --\", value: \"\" },\n { label: \"Promotional Social Media Post\", value: \"promo_post\" },\n { label: \"Blog Post Idea\", value: \"blog_idea\" },\n { label: \"Blog Post Draft/Outline\", value: \"blog_draft\" },\n { label: \"Advertising Copy (e.g., Google/Facebook Ad)\", value: \"ad_copy\" },\n ];\n break;\n case \"platform\":\n text = \"Which platform is this content primarily for?\";\n type = \"select\";\n options = [\n { label: \"-- Select Platform --\", value: \"\" },\n { label: \"Generic / Any Platform\", value: \"generic\" },\n { label: \"Facebook\", value: \"facebook\" },\n { label: \"Instagram\", value: \"instagram\" },\n { label: \"TikTok\", value: \"tiktok\" },\n { label: \"LinkedIn\", value: \"linkedin\" },\n { label: \"Google Ads\", value: \"google_ads\" },\n { label: \"Blog / Website\", value: \"blog\" },\n ];\n break;\n case \"specificInstructions\":\n text = \"Any specific instructions, keywords, topics, calls to action, or offers to include? (Optional)\";\n type = \"textarea\";\n break;\n case \"platforms\":\n text =\n \"Which specific online platforms are you most interested in focusing on for this strategy? (e.g., facebook, instagram, google_ads, linkedin, tiktok, email - separate multiple with commas)\";\n break;\n case \"budgetLevel\":\n text = \"What's your approximate budget level for this marketing campaign or effort? (Optional)\";\n type = \"select\";\n options = [\n { label: \"Select Budget (Optional)\", value: \"\" },\n { label: \"Low (e.g., organic focus, minimal ad spend)\", value: \"low\" },\n { label: \"Medium (e.g., moderate ad spend, some tools)\", value: \"medium\" },\n { label: \"High (e.g., significant ad spend, agency possible)\", value: \"high\" },\n ];\n break;\n case \"duration\":\n text =\n \"What is the planned timeframe or duration for this strategy or campaign? (Optional, e.g., Next 3 months, Ongoing, Specific event date)\";\n break;\n case \"stylePreferences\":\n text =\n \"Do you have any preferred styles for the logo? (Optional, e.g., minimalist, modern, vintage, bold, playful, corporate)\";\n break;\n case \"colorPreferences\":\n text =\n \"Are there any preferred colors or color palettes for the logo? (Optional, e.g., blue and white, warm tones, earthy colors, avoid pink)\";\n break;\n case \"coreValues\":\n text =\n `What are the core values, mission, or unique selling points of ${bizName}? (Optional, e.g., Reliability, Innovation, Community Focus, Sustainability, Affordability)`;\n type = \"textarea\";\n break;\n case \"simulationType\":\n text = \"What type of simulation would you like to run?\";\n type = \"select\";\n options = [\n { label: \"-- Select Simulation --\", value: \"\" },\n { label: \"Estimate Ad Response/CTR\", value: \"ad_response_estimate\" },\n { label: \"Simulate a Customer Service Interaction\", value: \"customer_interaction\" },\n { label: \"Other (Describe Below)\", value: \"other\" },\n ];\n break;\n case \"specificParams\":\n text =\n `Please provide specific details or parameters for the simulation.\\n- For 'Ad Response', mention the ad platform, target audience details, and maybe the ad copy or offer.\\n- For 'Customer Interaction', describe the customer's scenario or question.\\n- For 'Other', clearly explain what scenario or process you want to simulate.`;\n type = \"textarea\";\n break;\n case \"prompt_input\":\n text =\n \"Please describe the task you want the AI system prompt to achieve, or provide the existing prompt you want to refine or analyze.\";\n type = \"textarea\";\n break;\n default:\n text = `Please provide the '${field}'.`;\n if (\n field.toLowerCase().includes(\"description\") || field.toLowerCase().includes(\"details\")\n || field.toLowerCase().includes(\"instructions\") || field.toLowerCase().includes(\"values\")\n || field.toLowerCase().includes(\"params\") || field.toLowerCase().includes(\"prompt\")\n || field.toLowerCase().includes(\"audience\") || field.toLowerCase().includes(\"goals\")\n ) { type = \"textarea\"; }\n break;\n }\n return { id: qId, field, text, type, options };\n }\n\n async getNextStep(\n currentSessionId: string | null,\n task: UserInput[\"task\"],\n lastAnswer: { field: string; value: any } | null,\n currentState?: QuestionnaireState | null,\n ): Promise {\n let state: QuestionnaireState;\n if (!currentState || !currentSessionId) {\n const sessionId = `session-${Date.now()}-${Math.random().toString(36).substring(2, 8)}`;\n this.log(\n \"INFO\",\n \"QuestionnaireAgent\",\n `Starting new session: ${sessionId} for task: ${task}`,\n undefined,\n sessionId,\n );\n state = {\n sessionId: sessionId,\n task: task,\n requiredFields: this.getRequiredFieldsForTask(task),\n gatheredAnswers: {},\n questionHistory: [],\n isComplete: false,\n };\n } else {\n state = {\n ...currentState,\n gatheredAnswers: { ...currentState.gatheredAnswers },\n questionHistory: [...currentState.questionHistory],\n };\n if (lastAnswer?.field) {\n this.log(\"DEBUG\", \"QuestionnaireAgent\", `Received answer for field '${lastAnswer.field}'`, {\n value: lastAnswer.value,\n }, state.sessionId);\n const value = typeof lastAnswer.value === \"string\" ? lastAnswer.value.trim() : lastAnswer.value;\n state.gatheredAnswers[lastAnswer.field] = value;\n }\n if (state.task !== task) {\n this.log(\n \"INFO\",\n \"QuestionnaireAgent\",\n `Task changed from '${state.task}' to '${task}', retaining shared answers.`,\n undefined,\n state.sessionId,\n );\n const newRequiredFields = this.getRequiredFieldsForTask(task);\n const commonFields = Object.keys(state.gatheredAnswers).filter(field =>\n newRequiredFields.includes(field) && state.gatheredAnswers[field] !== null\n && state.gatheredAnswers[field] !== undefined\n && (typeof state.gatheredAnswers[field] !== \"string\" || state.gatheredAnswers[field].trim() !== \"\")\n );\n const retainedAnswers: Record = {};\n commonFields.forEach(field => {\n retainedAnswers[field] = state.gatheredAnswers[field];\n });\n this.log(\"DEBUG\", \"QuestionnaireAgent\", `Retaining answers for fields: [${commonFields.join(\", \")}]`, {\n retained: retainedAnswers,\n }, state.sessionId);\n state.task = task;\n state.requiredFields = newRequiredFields;\n state.gatheredAnswers = retainedAnswers;\n state.questionHistory = [];\n state.isComplete = false;\n }\n }\n\n const nextField = this.selectNextField(state.requiredFields, state.gatheredAnswers);\n\n if (!nextField) {\n state.isComplete = true;\n this.log(\n \"INFO\",\n \"QuestionnaireAgent\",\n `Questionnaire complete for task '${state.task}'. Finalizing data.`,\n undefined,\n state.sessionId,\n );\n const finalData = this.finalizeData(state.gatheredAnswers, state.task, state.sessionId);\n try {\n UserInputSchema.parse(finalData);\n this.log(\"INFO\", \"QuestionnaireAgent\", \"Final data validation successful.\", { finalData }, state.sessionId);\n return {\n sessionId: state.sessionId,\n isComplete: true,\n gatheredData: finalData,\n state: state,\n nextQuestion: undefined,\n error: undefined,\n };\n } catch (validationError) {\n const zodError = validationError as z.ZodError;\n const errorMessages = zodError.issues.map(i => `Field '${i.path.join(\".\")}': ${i.message}`).join(\", \");\n this.log(\"ERROR\", \"QuestionnaireAgent\", \"Final data validation failed\", {\n errors: zodError.issues,\n data: finalData,\n }, state.sessionId);\n state.isComplete = false;\n return {\n sessionId: state.sessionId,\n isComplete: false,\n error: `Data validation failed: ${errorMessages}. Required information might be missing or invalid.`,\n state: state,\n nextQuestion: undefined,\n gatheredData: undefined,\n };\n }\n } else {\n const nextQuestion = this.formulateQuestion(nextField, state);\n if (!state.questionHistory.find(q => q.questionId === nextQuestion.id)) {\n state.questionHistory.push({ questionId: nextQuestion.id, field: nextQuestion.field, text: nextQuestion.text });\n }\n this.log(\n \"INFO\",\n \"QuestionnaireAgent\",\n `Asking next question for field: ${nextField}`,\n undefined,\n state.sessionId,\n );\n return {\n sessionId: state.sessionId,\n isComplete: false,\n nextQuestion: nextQuestion,\n state: state,\n gatheredData: undefined,\n error: undefined,\n };\n }\n }\n\n private finalizeData(gatheredData: Record, task: UserInput[\"task\"], sessionId?: string): UserInput {\n let finalPayloadData: Record = { ...gatheredData };\n const baseReq = [\"businessType\", \"serviceDescription\", \"targetAudience\"];\n const taskSpecificReqMap: Partial> = {\n \"create_content\": [\"contentType\", \"platform\"],\n \"develop_strategy\": [\"platforms\"],\n \"generate_logo_concept\": [\"businessName\"],\n \"build_brand_identity\": [\"businessName\"],\n \"run_simulation\": [\"simulationType\", \"specificParams\"],\n \"meta_prompt\": [\"prompt_input\"],\n };\n const strictlyRequiredFields = new Set([...baseReq, ...(taskSpecificReqMap[task] || [])]);\n\n for (const key in finalPayloadData) {\n if (\n (finalPayloadData[key] === \"\" && !strictlyRequiredFields.has(key)) || finalPayloadData[key] === null\n || finalPayloadData[key] === undefined\n ) {\n delete finalPayloadData[key];\n this.log(\"DEBUG\", \"finalizeData\", `Removed empty/null optional field: ${key}`, undefined, sessionId);\n }\n }\n\n if (task === \"develop_strategy\" && typeof finalPayloadData.platforms === \"string\") {\n finalPayloadData.platforms = finalPayloadData.platforms.split(\",\").map((p: string) => p.trim()).filter((\n p: string,\n ) => p.length > 0);\n if (!Array.isArray(finalPayloadData.platforms) || finalPayloadData.platforms.length === 0) {\n this.log(\"WARN\", \"finalizeData\", \"Platforms string parsing resulted in empty array.\", {\n raw: gatheredData.platforms,\n }, sessionId);\n }\n }\n\n if (task === \"run_simulation\" && typeof finalPayloadData.specificParams === \"string\") {\n try {\n const parsedParams = JSON.parse(finalPayloadData.specificParams);\n if (typeof parsedParams === \"object\" && parsedParams !== null) {\n finalPayloadData.specificParams = parsedParams;\n } else {\n finalPayloadData.specificParams = { description: finalPayloadData.specificParams };\n this.log(\"WARN\", \"finalizeData\", \"Simulation specificParams JSON not an object, using as description.\", {\n raw: gatheredData.specificParams,\n }, sessionId);\n }\n } catch (e) {\n finalPayloadData.specificParams = { description: finalPayloadData.specificParams };\n this.log(\"WARN\", \"finalizeData\", \"Simulation specificParams not valid JSON, using as description.\", {\n raw: gatheredData.specificParams,\n }, sessionId);\n }\n }\n return { task: task, data: finalPayloadData };\n }\n}\n\nasync function runContentCreationAgent(\n task: TaskInput,\n log: LogFunction,\n): Promise> {\n const { mandateId, taskId, payload } = task;\n log(\n \"INFO\",\n \"ContentCreationAgent\",\n `Starting task ${taskId}`,\n { type: payload.contentType, platform: payload.platform },\n mandateId,\n taskId,\n );\n const systemPrompt =\n `You are an expert marketing copywriter specializing in creating engaging content for small businesses. Your goal is to generate content that matches the user's business, audience, tone, and specific requirements. Business Type: ${\n payload.businessType || \"Not specified\"\n } Service/Product: ${payload.serviceDescription || \"Not specified\"} Target Audience: ${\n payload.targetAudience || \"General audience\"\n } Brand Tone: ${payload.brandTone || \"Friendly and professional\"} Task Goal: ${\n payload.goals || \"General promotion\"\n } Content Type Requested: ${payload.contentType} Platform: ${payload.platform} Specific Instructions/Keywords: ${\n payload.specificInstructions || \"None\"\n } Instructions: Generate content suitable for the specified PLATFORM and CONTENT TYPE. Adapt the TONE and language to the TARGET AUDIENCE. Incorporate any SPECIFIC INSTRUCTIONS or keywords naturally. For social media posts, include relevant hashtags if appropriate for the platform. For ad copy, focus on clear calls to action and benefits. For blog ideas, provide 3-5 distinct and relevant titles/concepts. For blog drafts, provide a structured outline or initial draft (approx 300-500 words). Output *only* the requested content. Do not add extra explanations unless asked for suggestions.`;\n const userPrompt =\n `Please generate the ${payload.contentType} for ${payload.platform} based on the details provided in the system prompt.`;\n try {\n const generatedContent = await callOpenAI(systemPrompt, userPrompt, mandateId, taskId, log);\n if (!generatedContent) {\n log(\"WARN\", \"ContentCreationAgent\", `LLM returned no content for task ${taskId}`, undefined, mandateId, taskId);\n return {\n mandateId,\n correlationId: taskId,\n payload: { content: \"\" },\n error: \"Content generation failed: No output from AI.\",\n };\n }\n const output: ContentCreationOutput = { content: generatedContent };\n log(\n \"INFO\",\n \"ContentCreationAgent\",\n `Task ${taskId} completed successfully.`,\n { outLen: output.content.length },\n mandateId,\n taskId,\n );\n return { mandateId, correlationId: taskId, payload: output };\n } catch (error) {\n log(\"ERROR\", \"ContentCreationAgent\", `Task ${taskId} failed`, { error: error.message }, mandateId, taskId);\n return {\n mandateId,\n correlationId: taskId,\n payload: { content: \"\" },\n error: `Content generation failed: ${error.message}`,\n };\n }\n}\n\nasync function runMarketingStrategyAgent(\n task: TaskInput,\n log: LogFunction,\n): Promise> {\n const { mandateId, taskId, payload } = task;\n log(\n \"INFO\",\n \"MarketingStrategyAgent\",\n `Starting task ${taskId}`,\n { platforms: payload.platforms, budget: payload.budgetLevel },\n mandateId,\n taskId,\n );\n const systemPrompt =\n `You are a digital marketing strategist providing actionable advice for small businesses. Your goal is to outline a clear, concise online marketing strategy based on the user's input. Business Type: ${\n payload.businessType || \"Not specified\"\n } Service/Product: ${payload.serviceDescription || \"Not specified\"} Target Audience: ${\n payload.targetAudience || \"General audience\"\n } Goals: ${payload.goals || \"Increase online presence and leads\"} Platforms of Interest: ${\n Array.isArray(payload.platforms) ? payload.platforms.join(\", \") : \"Not specified\"\n } Budget Level: ${payload.budgetLevel || \"Not specified\"} Campaign Duration: ${\n payload.duration || \"Ongoing / Not specified\"\n } Brand Tone: ${\n payload.brandTone || \"Professional\"\n } Instructions: Structure the output using clear Markdown headings (e.g., ## Overall Approach, ## Recommended Platforms, ## Content Themes, ## Ad Copy Ideas, ## Audience Segment Suggestions, ## Next Steps). Provide specific, actionable recommendations for the RECOMMENDED PLATFORMS (chosen from the user's interest list or suggest alternatives if appropriate). Suggest 2-3 relevant CONTENT THEMES. Provide 2-3 examples under ## Ad Copy Ideas suitable for the business and goals. Suggest 1-2 specific ## Audience Segment Suggestions for targeting (if applicable). Keep recommendations practical for the indicated BUDGET LEVEL. Conclude with brief ## Next Steps. Output only the strategy document.`;\n const userPrompt = `Please develop the marketing strategy based on the details provided in the system prompt.`;\n try {\n const strategyContent = await callOpenAI(systemPrompt, userPrompt, mandateId, taskId, log);\n if (!strategyContent) {\n log(\"WARN\", \"MarketingStrategyAgent\", `LLM returned no content for task ${taskId}`, undefined, mandateId, taskId);\n return {\n mandateId,\n correlationId: taskId,\n payload: { strategy: \"\" },\n error: \"Strategy generation failed: No output from AI.\",\n };\n }\n const adIdeas =\n strategyContent.match(/## Ad Copy Ideas\\n([\\s\\S]*?)(?=\\n##|$)/)?.[1]?.split(/\\n\\s*[-\\*]\\s*/).map(s => s.trim())\n .filter(s => s && s.length > 5) || [];\n const audienceSuggestions =\n strategyContent.match(/## Audience Segment Suggestions\\n([\\s\\S]*?)(?=\\n##|$)/)?.[1]?.split(/\\n\\s*[-\\*]\\s*/).map(\n s => s.trim()\n ).filter(s => s && s.length > 5) || [];\n const output: MarketingStrategyOutput = {\n strategy: strategyContent,\n adIdeas: adIdeas,\n audienceSuggestions: audienceSuggestions,\n };\n log(\n \"INFO\",\n \"MarketingStrategyAgent\",\n `Task ${taskId} completed successfully.`,\n { stratLen: output.strategy.length },\n mandateId,\n taskId,\n );\n return { mandateId, correlationId: taskId, payload: output };\n } catch (error) {\n log(\"ERROR\", \"MarketingStrategyAgent\", `Task ${taskId} failed`, { error: error.message }, mandateId, taskId);\n return {\n mandateId,\n correlationId: taskId,\n payload: { strategy: \"\" },\n error: `Strategy generation failed: ${error.message}`,\n };\n }\n}\n\nasync function runLogoConceptAgent(\n task: TaskInput,\n log: LogFunction,\n): Promise> {\n const { mandateId, taskId, payload } = task;\n log(\n \"INFO\",\n \"LogoConceptAgent\",\n `Starting task ${taskId}`,\n { name: payload.businessName, style: payload.stylePreferences },\n mandateId,\n taskId,\n );\n const systemPrompt =\n `You are a creative assistant specializing in generating logo concepts and descriptions. Your goal is to provide 2-3 distinct logo ideas based on the user's business details. Business Name: ${payload.businessName} Business Type: ${payload.businessType} Service/Product: ${payload.serviceDescription} Target Audience: ${payload.targetAudience} Style Preferences: ${\n payload.stylePreferences || \"Not specified\"\n } Color Preferences: ${payload.colorPreferences || \"Not specified\"} Brand Tone: ${\n payload.brandTone || \"Professional\"\n } Instructions: Generate 2-3 distinct logo concepts. For each concept, provide: - A clear \"Concept Title:\" (e.g., Concept Title: Minimalist Icon) - A detailed \"Description:\" explaining the visual elements, symbolism, and font suggestions. - A potential \"Image Prompt:\" suitable for AI image generation tools (like DALL-E or Midjourney) based on the description. Start prompts with \"/imagine prompt:\" or similar format if helpful. Structure the output clearly, separating each concept. Optionally, if feasible, attempt to generate a very basic conceptual SVG representation for one concept, enclosed in \\`\\`\\`svg ... \\`\\`\\` markers. This is experimental. Focus on the description and prompts; the SVG is secondary. Output only the concepts. Do not add introductory or concluding remarks outside the concept descriptions.`;\n const userPrompt = `Generate logo concepts for \"${payload.businessName}\" based on the details in the system prompt.`;\n try {\n const conceptContent = await callOpenAI(systemPrompt, userPrompt, mandateId, taskId, log);\n if (!conceptContent) {\n log(\"WARN\", \"LogoConceptAgent\", `LLM returned no content for task ${taskId}`, undefined, mandateId, taskId);\n return {\n mandateId,\n correlationId: taskId,\n payload: { concepts: [] },\n error: \"Logo concept generation failed: No output from AI.\",\n };\n }\n const conceptSections = conceptContent.split(/Concept Title:/i).slice(1);\n const concepts: string[] = [];\n const prompts: string[] = [];\n conceptSections.forEach(section => {\n const descMatch = section.match(/Description:([\\s\\S]*?)(Image Prompt:|```svg|$)/i);\n const promptMatch = section.match(/Image Prompt:([\\s\\S]*?)(```svg|$)/i);\n const title = section.split(/Description:/i)[0]?.trim();\n const description = descMatch ? descMatch[1].trim() : section.trim();\n concepts.push(`Title: ${title}\\nDescription: ${description}`);\n if (promptMatch) { prompts.push(promptMatch[1].trim()); }\n });\n const svgAttemptMatch = conceptContent.match(/```svg\\n([\\s\\S]*?)\\n```/);\n const svgAttempt = svgAttemptMatch ? svgAttemptMatch[1].trim() : undefined;\n const output: LogoConceptOutput = { concepts: concepts, promptsForImageModel: prompts, svgAttempt: svgAttempt };\n log(\n \"INFO\",\n \"LogoConceptAgent\",\n `Task ${taskId} completed. Found ${output.concepts.length} concepts.`,\n { prompts: output.promptsForImageModel?.length, svg: !!output.svgAttempt },\n mandateId,\n taskId,\n );\n return { mandateId, correlationId: taskId, payload: output };\n } catch (error) {\n log(\"ERROR\", \"LogoConceptAgent\", `Task ${taskId} failed`, { error: error.message }, mandateId, taskId);\n return {\n mandateId,\n correlationId: taskId,\n payload: { concepts: [] },\n error: `Logo concept generation failed: ${error.message}`,\n };\n }\n}\n\nasync function runBrandIdentityAgent(\n task: TaskInput,\n log: LogFunction,\n): Promise> {\n const { mandateId, taskId, payload } = task;\n log(\n \"INFO\",\n \"BrandIdentityAgent\",\n `Starting task ${taskId}`,\n { name: payload.businessName, values: payload.coreValues },\n mandateId,\n taskId,\n );\n const systemPrompt =\n `You are a brand strategist assistant helping define foundational brand identity elements. Your goal is to create a concise guide based on the user's input. Business Name: ${payload.businessName} Business Type: ${payload.businessType} Service/Product: ${payload.serviceDescription} Target Audience: ${payload.targetAudience} Core Values/Mission: ${\n payload.coreValues || \"Not specified\"\n } Desired Brand Tone: ${\n payload.brandTone || \"Professional and trustworthy\"\n } Instructions: Develop the following sections using Markdown headings: - ## Brand Voice: Describe the tone, language style, and personality. - ## Key Messages: List 3-5 core messages or taglines that communicate the brand's value proposition. Use bullet points. - ## Color Palette Suggestions: Suggest a primary and secondary color palette (e.g., \"Primary: Deep Blue (#003366), Secondary: Bright Orange (#FFA500), Accent: Light Grey (#CCCCCC)\") with brief justification. - ## Typography Suggestions: Recommend font pairings for headings and body text (e.g., \"Heading: Montserrat Bold, Body: Lato Regular\") with brief justification. Keep the descriptions concise and actionable. Ensure suggestions align with the business type, audience, and desired tone. Output only the brand identity guide.`;\n const userPrompt =\n `Build the brand identity guide for \"${payload.businessName}\" based on the details in the system prompt.`;\n try {\n const identityContent = await callOpenAI(systemPrompt, userPrompt, mandateId, taskId, log);\n if (!identityContent) {\n log(\"WARN\", \"BrandIdentityAgent\", `LLM returned no content for task ${taskId}`, undefined, mandateId, taskId);\n return {\n mandateId,\n correlationId: taskId,\n payload: { brandVoice: \"\", keyMessages: [], colorPaletteSuggestions: \"\", typographySuggestions: \"\" },\n error: \"Brand identity generation failed: No output from AI.\",\n };\n }\n const brandVoice = identityContent.match(/## Brand Voice\\n([\\s\\S]*?)(?=\\n##|$)/)?.[1]?.trim() || \"N/A\";\n const keyMessagesRaw = identityContent.match(/## Key Messages\\n([\\s\\S]*?)(?=\\n##|$)/)?.[1] || \"\";\n const keyMessages = keyMessagesRaw.split(/\\n\\s*[-\\*]\\s*/).map(s => s.trim()).filter(s => s.length > 5);\n const colorPaletteSuggestions =\n identityContent.match(/## Color Palette Suggestions\\n([\\s\\S]*?)(?=\\n##|$)/)?.[1]?.trim() || \"N/A\";\n const typographySuggestions = identityContent.match(/## Typography Suggestions\\n([\\s\\S]*?)(?=\\n##|$)/)?.[1]?.trim()\n || \"N/A\";\n const output: BrandIdentityOutput = { brandVoice, keyMessages, colorPaletteSuggestions, typographySuggestions };\n log(\"INFO\", \"BrandIdentityAgent\", `Task ${taskId} completed successfully.`, undefined, mandateId, taskId);\n return { mandateId, correlationId: taskId, payload: output };\n } catch (error) {\n log(\"ERROR\", \"BrandIdentityAgent\", `Task ${taskId} failed`, { error: error.message }, mandateId, taskId);\n return {\n mandateId,\n correlationId: taskId,\n payload: { brandVoice: \"\", keyMessages: [], colorPaletteSuggestions: \"\", typographySuggestions: \"\" },\n error: `Brand identity generation failed: ${error.message}`,\n };\n }\n}\n\nasync function runSimulationAgent(\n task: TaskInput,\n log: LogFunction,\n): Promise> {\n const { mandateId, taskId, payload } = task;\n log(\n \"INFO\",\n \"SimulationAgent\",\n `Starting task ${taskId}`,\n { type: payload.simulationType, params: payload.specificParams },\n mandateId,\n taskId,\n );\n const systemPrompt =\n `You are an advanced simulation model capable of predicting outcomes or role-playing scenarios based on provided business context. Business Type: ${payload.businessType} Service/Product: ${payload.serviceDescription} Target Audience: ${payload.targetAudience} Simulation Type Requested: ${payload.simulationType} Specific Parameters/Scenario: ${\n JSON.stringify(payload.specificParams)\n } Instructions: Based on the SIMULATION TYPE and SPECIFIC PARAMETERS, perform the requested simulation. For 'ad_response_estimate': Predict potential click-through rates (CTR), conversion rates, or provide qualitative feedback on ad effectiveness based on the parameters. State assumptions clearly. For 'customer_interaction': Role-play the interaction from the perspective of the customer or the business representative, as appropriate to the scenario. Make the dialogue realistic. For 'other': Execute the simulation as described in the parameters. Structure the output clearly, starting with the core simulation result. If applicable, add a brief \"## Analysis\" section providing insights or takeaways from the simulation. Output only the simulation result and analysis.`;\n const userPrompt = `Run the ${payload.simulationType} simulation using the parameters provided in the system prompt.`;\n try {\n const simContent = await callOpenAI(systemPrompt, userPrompt, mandateId, taskId, log);\n if (!simContent) {\n log(\"WARN\", \"SimulationAgent\", `LLM returned no content for task ${taskId}`, undefined, mandateId, taskId);\n return {\n mandateId,\n correlationId: taskId,\n payload: { simulationResult: \"\" },\n error: \"Simulation failed: No output from AI.\",\n };\n }\n const simResult = simContent.match(/([\\s\\S]*?)(?=\\n## Analysis|$)/)?.[1]?.trim() || simContent;\n const analysis = simContent.match(/## Analysis\\n([\\s\\S]*)/)?.[1]?.trim();\n const output: SimulationOutput = { simulationResult: simResult, analysis };\n log(\n \"INFO\",\n \"SimulationAgent\",\n `Task ${taskId} completed successfully.`,\n { hasAnalysis: !!output.analysis },\n mandateId,\n taskId,\n );\n return { mandateId, correlationId: taskId, payload: output };\n } catch (error) {\n log(\"ERROR\", \"SimulationAgent\", `Task ${taskId} failed`, { error: error.message }, mandateId, taskId);\n return {\n mandateId,\n correlationId: taskId,\n payload: { simulationResult: \"\" },\n error: `Simulation failed: ${error.message}`,\n };\n }\n}\n\nasync function runMetaPromptAgent(\n task: TaskInput,\n log: LogFunction,\n): Promise> {\n const { mandateId, taskId, payload } = task;\n log(\n \"INFO\",\n \"MetaPromptAgent\",\n `Starting task ${taskId}`,\n { inputLen: payload.prompt_input.length },\n mandateId,\n taskId,\n );\n const systemPromptForMeta =\n `You are an expert AI Prompt Engineer. Your task is to refine or generate a high-quality system prompt for another AI model (like GPT-4) based on the user's request. Guidelines for creating the system prompt: 1. **Role Definition:** Clearly define the AI's role, expertise, and persona (e.g., \"You are a helpful marketing assistant specializing in social media content for small businesses.\"). 2. **Task Definition:** State the primary goal or task the AI should accomplish (e.g., \"Your goal is to generate engaging Instagram post captions.\"). 3. **Context Inclusion:** Instruct the AI to consider specific context provided by the user later (e.g., \"Use the following user-provided details: Business Description, Target Audience, Specific Offer...\"). List the expected context keys. 4. **Process/Steps (If Applicable):** Outline key steps the AI should follow (e.g., \"1. Analyze the user input. 2. Draft three variations. 3. Select the best one.\"). 5. **Output Format:** Specify the desired output format clearly (e.g., \"Output only the final caption text.\", \"Structure the output using Markdown headings: ## Strategy, ## Content Ideas.\"). Use examples if necessary. 6. **Constraints & Tone:** Define any constraints (e.g., \"Keep responses under 100 words.\", \"Avoid technical jargon.\") and the desired tone (e.g., \"Maintain a friendly and encouraging tone.\"). 7. **Refinement:** If the user provided an existing prompt, analyze it and incorporate its strengths while addressing weaknesses based on these guidelines. User Request: The user wants a system prompt for an AI that will perform the following task OR wants to refine the following existing prompt: \\`\\`\\` ${payload.prompt_input} \\`\\`\\` Instructions: Generate the complete system prompt based on the user's request and the guidelines above. Enclose the final system prompt in \\`\\`\\`markdown ... \\`\\`\\` or just output the raw text. Do not add any explanations outside the generated prompt itself.`;\n const userPromptForMeta =\n `Generate or refine the system prompt based on my request provided in the system prompt context.`;\n try {\n const genSysPrompt = await callOpenAI(systemPromptForMeta, userPromptForMeta, mandateId, taskId, log);\n if (!genSysPrompt) {\n log(\"WARN\", \"MetaPromptAgent\", `LLM returned no content for task ${taskId}`, undefined, mandateId, taskId);\n return {\n mandateId,\n correlationId: taskId,\n payload: { system_prompt: \"\" },\n error: \"Meta-prompt generation failed: No output from AI.\",\n };\n }\n const finalPrompt = genSysPrompt.replace(/^```(markdown)?\\n?/, \"\").replace(/\\n?```$/, \"\").trim();\n const output: MetaPromptOutput = { system_prompt: finalPrompt };\n log(\n \"INFO\",\n \"MetaPromptAgent\",\n `Task ${taskId} completed successfully.`,\n { outputLen: output.system_prompt.length },\n mandateId,\n taskId,\n );\n return { mandateId, correlationId: taskId, payload: output };\n } catch (error) {\n log(\"ERROR\", \"MetaPromptAgent\", `Task ${taskId} failed`, { error: error.message }, mandateId, taskId);\n return {\n mandateId,\n correlationId: taskId,\n payload: { system_prompt: \"\" },\n error: `Meta-prompt generation failed: ${error.message}`,\n };\n }\n}\n\nclass Orchestrator {\n baseLog: LogFunction;\n mandateId: string;\n log: LogFunction;\n\n constructor(baseLoggerInstance: MockLoggingAgent) {\n this.baseLog = baseLoggerInstance.log.bind(baseLoggerInstance);\n this.mandateId = `mandate-${Date.now()}-${Math.random().toString(36).substring(2, 8)}`;\n baseLoggerInstance.setMandateIdContext(this.mandateId);\n this.log = (level, component, message, details?, _mandateIdIgnored?, taskId?) => {\n this.baseLog(level, component, message, details, this.mandateId, taskId);\n };\n this.log(\"INFO\", \"Orchestrator\", `Initialized execution mandate: ${this.mandateId}`);\n }\n\n async runWorkflow(input: UserInput): Promise {\n this.log(\"INFO\", \"Orchestrator\", `Starting workflow for task: ${input.task}`, {\n inputKeys: Object.keys(input.data),\n });\n const taskId = `${input.task}-exec-${Date.now()}-${Math.random().toString(36).substring(2, 5)}`;\n const taskInput: TaskInput = { mandateId: this.mandateId, taskId, payload: input.data };\n let taskOutput: TaskOutput;\n\n try {\n const agentLogger: LogFunction = (level, component, message, details?, mandateIdOverride?, taskIdFromAgent?) => {\n this.log(level, component, message, details, this.mandateId, taskIdFromAgent || taskId);\n };\n\n switch (input.task) {\n case \"create_content\":\n taskOutput = await runContentCreationAgent(taskInput as TaskInput, agentLogger);\n break;\n case \"develop_strategy\":\n taskOutput = await runMarketingStrategyAgent(taskInput as TaskInput, agentLogger);\n break;\n case \"generate_logo_concept\":\n taskOutput = await runLogoConceptAgent(taskInput as TaskInput, agentLogger);\n break;\n case \"build_brand_identity\":\n taskOutput = await runBrandIdentityAgent(taskInput as TaskInput, agentLogger);\n break;\n case \"run_simulation\":\n taskOutput = await runSimulationAgent(taskInput as TaskInput, agentLogger);\n break;\n case \"meta_prompt\":\n taskOutput = await runMetaPromptAgent(taskInput as TaskInput, agentLogger);\n break;\n default:\n this.log(\"ERROR\", \"Orchestrator\", `Unknown task type: ${input.task}`, undefined, this.mandateId, taskId);\n throw new Error(`Unsupported task type: ${(input as any).task}`);\n }\n\n if (taskOutput.error) {\n this.log(\"WARN\", \"Orchestrator\", `Agent for task ${taskId} reported error`, {\n error: taskOutput.error,\n agentPayload: taskOutput.payload,\n });\n return {\n mandateId: this.mandateId,\n error: `Agent execution failed: ${taskOutput.error}`,\n result: taskOutput.payload,\n debugLog: [],\n details: \"Agent reported an error during its execution.\",\n };\n }\n\n this.log(\"INFO\", \"Orchestrator\", `Workflow for task ${taskId} completed successfully.`);\n return { mandateId: this.mandateId, result: taskOutput.payload, error: undefined, debugLog: [] };\n } catch (error) {\n this.log(\"ERROR\", \"Orchestrator\", `Workflow failed critically for task ${taskId}`, {\n error: error.message,\n stack: error.stack?.substring(0, 200),\n }); // Limit stack trace length\n return { mandateId: this.mandateId, error: `Workflow execution failed: ${error.message}`, debugLog: [] };\n }\n }\n}\n\nexport default async function(req: Request): Promise {\n const requestLogger = new MockLoggingAgent(\"http-request\");\n const log = requestLogger.log.bind(requestLogger);\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(\"INFO\", \"Handler\", \"Handling OPTIONS preflight request\");\n return new Response(null, { status: 204, headers });\n }\n\n try {\n if (req.method === \"GET\" && (url.pathname === \"/\" || url.pathname === \"\")) {\n log(\"INFO\", \"Handler\", \"GET /: Serving Questionnaire UI shell.\");\n const html = generateQuestionnaireHtmlShell();\n return new Response(html, { status: 200, headers: { ...headers, \"Content-Type\": \"text/html\" } });\n }\n else if (req.method === \"POST\" && url.pathname === \"/questionnaire\") {\n log(\"INFO\", \"Handler\", \"POST /questionnaire step\");\n let requestBody;\n try {\n requestBody = await req.json();\n if (\n !requestBody || typeof requestBody !== \"object\" || !requestBody.task || typeof requestBody.task !== \"string\"\n ) {\n throw new Error(\"Invalid request body: missing or invalid 'task' (string).\");\n }\n } catch (e) {\n log(\"WARN\", \"Handler\", \"Invalid JSON or structure in /questionnaire request\", {\n error: e.message,\n body: await req.text().catch(() => \"Failed to read body\"),\n });\n return new Response(JSON.stringify({ error: `Bad Request: ${e.message}`, state: null }), {\n status: 400,\n headers,\n });\n }\n\n const questionnaireAgent = new QuestionnaireAgent(log);\n const { sessionId, task, lastAnswer, state: prevState } = requestBody;\n const responsePayload = await questionnaireAgent.getNextStep(sessionId, task, lastAnswer, prevState);\n return new Response(JSON.stringify(responsePayload), { status: responsePayload.error ? 400 : 200, headers });\n }\n else if (req.method === \"POST\" && url.pathname === \"/execute\") {\n log(\"INFO\", \"Handler\", \"POST /execute workflow\");\n let validatedInput: UserInput;\n let rawPayload;\n try {\n rawPayload = await req.json();\n log(\"DEBUG\", \"Handler\", \"Received payload for /execute\", { payload: rawPayload });\n validatedInput = UserInputSchema.parse(rawPayload);\n log(\"INFO\", \"Handler\", \"Payload validation successful for /execute\");\n } catch (error) {\n let errorDetails = \"Invalid input format.\";\n let issues = undefined;\n if (error instanceof z.ZodError) {\n issues = error.issues;\n errorDetails = issues.map(i => `Field '${i.path.join(\".\")}': ${i.message}`).join(\"; \");\n log(\"WARN\", \"Handler\", \"Invalid final payload for /execute (Zod)\", {\n error: errorDetails,\n issues: issues,\n received: rawPayload,\n });\n } else {\n errorDetails = error.message;\n log(\"WARN\", \"Handler\", \"Error parsing/validating /execute payload\", {\n error: errorDetails,\n received: rawPayload,\n });\n }\n const errorLogs = requestLogger.getLogs();\n return new Response(\n JSON.stringify({ error: \"Invalid input for execution.\", details: errorDetails, debugLog: errorLogs }),\n { status: 400, headers },\n );\n }\n\n const executionLogger = new MockLoggingAgent();\n const orchestrator = new Orchestrator(executionLogger);\n const result: FinalResult = await orchestrator.runWorkflow(validatedInput);\n result.debugLog = executionLogger.getLogs();\n const status = result.error ? 500 : 200;\n log(\"INFO\", \"Handler\", `/execute finished with status ${status}`, {\n mandateId: result.mandateId,\n error: result.error,\n });\n return new Response(JSON.stringify(result), { status, headers });\n }\n else {\n log(\"WARN\", \"Handler\", `Unsupported request: ${req.method} ${url.pathname}`);\n return new Response(JSON.stringify({ error: \"Not Found or Method Not Allowed\" }), { status: 404, headers });\n }\n } catch (error) {\n log(\"ERROR\", \"Handler\", \"Unhandled exception in HTTP handler\", {\n error: error.message,\n stack: error.stack?.substring(0, 200),\n });\n const crashLogs = requestLogger.getLogs();\n return new Response(\n JSON.stringify({ error: \"Internal Server Error\", details: error.message, debugLog: crashLogs }),\n { status: 500, headers },\n );\n }\n}","deleted","label","children",[],[154],"files",[152],"counts",{"_175":31,"_176":177,"_178":31,"_179":31},"pulls","branches",1,"forks","envVars","contents",[182],{"_107":153,"_100":154,"_155":120,"_156":-5,"_157":140,"_158":139,"_159":113,"_109":183,"_104":161,"_162":163,"_164":165,"_166":27,"_167":-5,"_103":184},["D",1745712428316],[154],"file","upstreamProject",{"_107":188,"_100":101,"_116":-5,"_117":-5,"_114":115,"_21":189},"a8cef910-2797-11f0-9d52-569c3dd06744",{"_98":190,"_135":191},"salon","https://img.clerk.com/eyJ0eXBlIjoiZGVmYXVsdCIsImlpZCI6Imluc18yTm10cGhQUklwU0tqZHNiOHNnNGd2ZjJURFUiLCJyaWQiOiJ1c2VyXzJ3NnI5NzZ4ajBmUUZoTUxYdExrMmZYWnAzdyJ9","upstreamBranch",{"_107":143,"_109":194,"_195":196,"_100":42,"_141":188,"_104":197,"_142":-5,"_144":-5,"_198":199,"_111":-5,"_133":27,"_114":115,"_149":190,"_148":101},["D",1746219277414],"updatedAt",["D",1745689548780],8,"lockfileKey","1746268811108-esm.town_v_salon_svvc@8-main","emptyEnvVars","isNew","routes/_app.x.$handle.$name._index",{"_98":99,"_100":101,"_102":-7,"_104":-7,"_105":204,"_180":208,"_211":-7,"_212":213,"_216":-5,"_171":217,"_221":222},{"_107":108,"_100":101,"_109":205,"_111":-5,"_112":113,"_114":115,"_116":-5,"_117":-5,"_12":118,"_119":120,"_121":113,"_122":-5,"_123":31,"_124":125,"_126":125,"_127":128,"_129":128,"_130":128,"_131":125,"_132":125,"_133":27,"_21":206,"_137":207},["D",1746228430817],{"_98":99,"_135":136},{"_104":139,"_100":42,"_107":140,"_141":108,"_142":143,"_144":31,"_145":115,"_146":147,"_148":101,"_149":99},[209],{"_107":153,"_100":154,"_155":120,"_156":-5,"_157":140,"_158":139,"_159":113,"_109":183,"_104":161,"_162":163,"_164":165,"_166":27,"_167":-5,"_103":210},[154],"readme","upstream",{"_107":143,"_109":214,"_195":215,"_100":42,"_141":188,"_104":197,"_142":-5,"_144":-5,"_198":199,"_111":-5,"_133":27,"_114":115,"_149":190,"_148":101},["D",1746219277414],["D",1745689548780],"pull",[218],{"_107":153,"_100":154,"_155":120,"_156":-5,"_157":140,"_158":139,"_159":113,"_109":160,"_104":161,"_162":163,"_164":165,"_166":27,"_167":-5,"_168":219,"_103":220},[],[154],"previews",[218],"routes/_app","actionData","errors"]