Calendar Meeting Research Assistant

\n\n

\n Calendar Meeting Research Assistant\n

\n \n \n

📷 Calendar Screenshot (optional)

\n \n

Drag & drop or click to upload calendar screenshot

\n \n \n \n

\n\n

\n $\"Preview\"$ \n

\n \n

\n\n \n

\n and/or\n

\n\n \n

📝 Calendar Text (optional)

\n \n

\n\n \n

💡 Additional Context (optional)

\n \n

\n\n \n

\n \n \n

\n\n \n

\n \n \n

0\">\n \n Elapsed: seconds\n \n

\n\n \n

Meeting Research Results

\n\n \n

💬 Chat with Research Results

\n \n \n

\n \n

\n\n \n

\n \n \n

\n\n \n\n`;\n\n// Serve root with HTML content\napp.get('/', (c) => {\n return c.html(htmlContent);\n});\n\n// Configuration endpoint\napp.get('/api/config', (c) => {\n return c.json({\n researchTypes: Object.keys(researchTypes),\n config: researchTypes,\n adapters: Object.keys(researchAdapters)\n });\n});\n\n// Research adapters status endpoint\napp.get('/api/adapters', (c) => {\n const adapterStatus = {};\n \n for (const [name, adapter] of Object.entries(researchAdapters)) {\n let status = 'available';\n let apiKey = false;\n \n try {\n switch (name) {\n case 'compound':\n apiKey = !!Deno.env.get(\"GROQ_API_KEY\");\n break;\n }\n \n if (!apiKey) {\n status = 'missing_api_key';\n }\n } catch (error) {\n status = 'error';\n }\n \n adapterStatus[name] = {\n name: adapter.name,\n status,\n configured: apiKey\n };\n }\n \n return c.json({\n adapters: adapterStatus,\n environment_variables: {\n required: ['GROQ_API_KEY (always required)'],\n optional: []\n }\n });\n});\n\n// Chat API endpoint\napp.post('/api/chat', async (c) => {\n try {\n const body = await c.req.json();\n const { message, researchContext } = body;\n\n if (!message || !researchContext) {\n return c.json({ error: 'Message and research context are required' }, 400);\n }\n\n console.log('=== CHAT REQUEST ===');\n console.log('User Message:', message);\n console.log('Research Context Length:', researchContext.length);\n console.log('=== END CHAT REQUEST ===');\n\n const systemPrompt = `You are a helpful research assistant. You have access to detailed research results about meetings, people, and companies. Answer questions based on the provided research context. Be conversational but accurate.\n\nIf the user asks about something not covered in the research, politely say you don't have that information in the current research results.\n\nResearch Context:\n${researchContext}`;\n\n const chatResponse = await groq.chat.completions.create({\n model: 'moonshotai/kimi-k2-instruct',\n messages: [\n {\n role: 'system',\n content: systemPrompt\n },\n {\n role: 'user',\n content: message\n }\n ],\n temperature: 0.7,\n max_tokens: 1000\n });\n\n const response = chatResponse.choices[0].message.content;\n\n console.log('=== CHAT RESPONSE ===');\n console.log('Assistant Response:', response);\n console.log('=== END CHAT RESPONSE ===');\n\n return c.json({ response });\n\n } catch (error) {\n console.error('Chat error:', error);\n return c.json({ error: 'Chat service temporarily unavailable' }, 500);\n }\n});\n\n// API endpoint for calendar meeting research\napp.post('/api/research', async (c) => {\n try {\n // Check if GROQ_API_KEY is available\n if (!Deno.env.get(\"GROQ_API_KEY\")) {\n console.error('GROQ_API_KEY not found in environment variables');\n return c.json({ error: 'GROQ_API_KEY not configured. Please set your API key.' }, 500);\n }\n\n const formData = await c.req.formData();\n const imageFile = formData.get('image_input');\n const calendarText = formData.get('calendar_text');\n const additionalContext = (formData.get('additional_context') as string) || '';\n\n // Check that we have either image or text input\n if (!imageFile && !calendarText) {\n return c.json({ error: 'No input provided - need either image file or calendar text' }, 400);\n }\n\n console.log('Starting calendar analysis...');\n const startTime = Date.now();\n\n let extractionResponse;\n\n if (imageFile && imageFile instanceof Blob && calendarText) {\n // Process both image and text input\n console.log('Processing both image and text input...');\n \n // Convert blob to base64\n const arrayBuffer = await imageFile.arrayBuffer();\n const bytes = new Uint8Array(arrayBuffer);\n const base64Image = uint8ToBase64(bytes);\n\n // Prepare user message content for both inputs\n const userContent = [\n {\n type: 'text' as const,\n text: `Extract all 1:1 meetings from both the calendar screenshot and the provided text. Combine information from both sources and return as JSON array.\n\nCalendar Text:\n${calendarText}\n${additionalContext ? `\\\\n\\\\nAdditional context: ${additionalContext}` : ''}`\n },\n {\n type: 'image_url' as const,\n image_url: {\n url: `data:image/jpeg;base64,${base64Image}`\n }\n }\n ];\n\n extractionResponse = await groq.chat.completions.create({\n model: 'meta-llama/llama-4-maverick-17b-128e-instruct',\n messages: [\n {\n role: 'system',\n content: `You are an expert at analyzing calendar data from multiple sources. Extract all 1:1 scheduled meetings from both the calendar image and text provided, combining information from both sources. Return them as a JSON array. \n\nFor each meeting, include:\n- title: The meeting title/subject\n- time: The time (if visible)\n- date: The date (if visible) \n- attendees: Array of people mentioned (extract names from meeting titles, attendee lists, etc.)\n- companies: Array of companies mentioned (extract from meeting titles, attendee info, etc.)\n\nCross-reference information between image and text to get the most complete data. Use any additional context provided to better identify people and companies. Return ONLY valid JSON, no other text. If no 1:1 meetings are found, return an empty array []`\n },\n {\n role: 'user',\n content: userContent\n }\n ],\n temperature: 0.1\n });\n } else if (imageFile && imageFile instanceof Blob) {\n // Process image input only\n console.log('Processing image input...');\n \n // Convert blob to base64\n const arrayBuffer = await imageFile.arrayBuffer();\n const bytes = new Uint8Array(arrayBuffer);\n const base64Image = uint8ToBase64(bytes);\n\n // Prepare user message content for image\n const userContent = [\n {\n type: 'text' as const,\n text: `Extract all 1:1 meetings from this calendar screenshot and return as JSON array.${additionalContext ? `\\\\n\\\\nAdditional context: ${additionalContext}` : ''}`\n },\n {\n type: 'image_url' as const,\n image_url: {\n url: `data:image/jpeg;base64,${base64Image}`\n }\n }\n ];\n\n extractionResponse = await groq.chat.completions.create({\n model: 'meta-llama/llama-4-maverick-17b-128e-instruct',\n messages: [\n {\n role: 'system',\n content: `You are an expert at analyzing calendar screenshots. Extract all 1:1 scheduled meetings from the calendar image and return them as a JSON array. \n\nFor each meeting, include:\n- title: The meeting title/subject\n- time: The time (if visible)\n- date: The date (if visible) \n- attendees: Array of people mentioned (extract names from meeting titles, attendee lists, etc.)\n- companies: Array of companies mentioned (extract from meeting titles, attendee info, etc.)\n\nUse any additional context provided to better identify people and companies. Return ONLY valid JSON, no other text. If no 1:1 meetings are found, return an empty array []`\n },\n {\n role: 'user',\n content: userContent\n }\n ],\n temperature: 0.1\n });\n } else {\n // Process text input only\n console.log('Processing text input...');\n \n extractionResponse = await groq.chat.completions.create({\n model: 'moonshotai/kimi-k2-instruct',\n messages: [\n {\n role: 'system',\n content: `You are an expert at analyzing calendar text data. Extract all 1:1 scheduled meetings from the provided calendar text and return them as a JSON array. \n\nFor each meeting, include:\n- title: The meeting title/subject\n- time: The time (if visible)\n- date: The date (if visible) \n- attendees: Array of people mentioned (extract names from meeting titles, attendee lists, etc.)\n- companies: Array of companies mentioned (extract from meeting titles, attendee info, etc.)\n\nUse any additional context provided to better identify people and companies. Return ONLY valid JSON, no other text. If no 1:1 meetings are found, return an empty array []`\n },\n {\n role: 'user',\n content: `Extract all 1:1 meetings from this calendar text and return as JSON array:\n\n${calendarText}${additionalContext ? `\\\\n\\\\nAdditional context: ${additionalContext}` : ''}`\n }\n ],\n temperature: 0.1\n });\n }\n\n let meetings;\n try {\n const extractedContent = extractionResponse.choices[0].message.content.trim();\n console.log('Raw extraction response:', extractedContent);\n \n // Remove any markdown code blocks if present - fix the regex\n let jsonContent = extractedContent.replace(/```json\\s*\\n?|```\\s*$/g, '').trim();\n \n // Additional cleanup for any remaining backticks\n jsonContent = jsonContent.replace(/^```|```$/g, '').trim();\n \n console.log('Cleaned JSON content:', jsonContent);\n meetings = JSON.parse(jsonContent);\n } catch (parseError) {\n console.error('JSON parse error:', parseError);\n console.log('Raw response:', extractionResponse.choices[0].message.content);\n return c.json({ error: 'Failed to parse meeting data from calendar' }, 500);\n }\n\n console.log(`Found ${meetings.length} meetings:`, meetings);\n\n if (!Array.isArray(meetings) || meetings.length === 0) {\n return c.json({ \n text: 'No 1:1 meetings found in this calendar image.',\n html: '

No 1:1 meetings found in this calendar image.

'\n });\n }\n\n // Step 2: Research each meeting\n let markdownResult = '# Meeting Research Results\\n\\n';\n let allResearchData = ''; // Collect all research for final analysis\n let allRawAdapterData = []; // Collect all raw adapter responses\n \n for (let i = 0; i < meetings.length; i++) {\n const meeting = meetings[i];\n console.log(`Researching meeting ${i + 1}/${meetings.length}: ${meeting.title}`);\n \n markdownResult += `## Meeting ${i + 1}: ${meeting.title}\\n\\n`;\n \n if (meeting.time) markdownResult += `**Time:** ${meeting.time}\\n`;\n if (meeting.date) markdownResult += `**Date:** ${meeting.date}\\n`;\n \n markdownResult += '\\n';\n\n // Research companies with meeting context\n if (meeting.companies && meeting.companies.length > 0) {\n markdownResult += `

\\n

🏢 Company Research (${meeting.companies.length} companies)

\\n\\n`;\n \n for (const company of meeting.companies) {\n try {\n console.log(`Researching company: ${company}`);\n const researchResult = await performResearch(researchAdapters, groq, 'company', company, meeting, additionalContext);\n \n // Handle the new return format\n const synthesizedContent = typeof researchResult === 'string' ? researchResult : researchResult.synthesized;\n const rawData = typeof researchResult === 'object' ? researchResult.rawAdapterData : null;\n \n markdownResult += `### ${company}\\n\\n${synthesizedContent}\\n\\n`;\n allResearchData += `Company: ${company} - ${synthesizedContent}\\n\\n`;\n \n // Collect raw adapter data\n if (rawData) {\n allRawAdapterData.push(...rawData);\n }\n } catch (error) {\n console.error(`Error researching company ${company}:`, error);\n markdownResult += `### ${company}\\n\\n*Research unavailable for this company.*\\n\\n`;\n }\n }\n markdownResult += `

\\n\\n`;\n }\n\n // Research individuals with meeting context\n if (meeting.attendees && meeting.attendees.length > 0) {\n markdownResult += `

\\n

👤 People Research (${meeting.attendees.length} people)

\\n\\n`;\n \n for (const person of meeting.attendees) {\n try {\n console.log(`Researching person: ${person}`);\n const researchResult = await performResearch(researchAdapters, groq, 'individual', person, meeting, additionalContext);\n \n // Handle the new return format\n const synthesizedContent = typeof researchResult === 'string' ? researchResult : researchResult.synthesized;\n const rawData = typeof researchResult === 'object' ? researchResult.rawAdapterData : null;\n \n markdownResult += `### ${person}\\n\\n${synthesizedContent}\\n\\n`;\n allResearchData += `Person: ${person} - ${synthesizedContent}\\n\\n`;\n \n // Collect raw adapter data\n if (rawData) {\n allRawAdapterData.push(...rawData);\n }\n } catch (error) {\n console.error(`Error researching person ${person}:`, error);\n markdownResult += `### ${person}\\n\\n*Research unavailable for this person.*\\n\\n`;\n }\n }\n markdownResult += `

\\n\\n`;\n }\n\n markdownResult += '---\\n\\n';\n }\n\n // Step 3: Generate final strategic analysis\n console.log('Generating strategic analysis...');\n try {\n const finalAnalysis = await generateFinalAnalysis(groq, meetings, allResearchData);\n markdownResult += `

\\n

🎯 Strategic Briefing Summary

\\n\\n${finalAnalysis}\\n\\n

\\n\\n`;\n } catch (error) {\n console.error('Error generating final analysis:', error);\n markdownResult += `

\\n

🎯 Strategic Briefing Summary

\\n\\n*Analysis unavailable - see individual research above.*\\n\\n

\\n\\n`;\n }\n\n // Add summary\n const totalTime = (Date.now() - startTime) / 1000;\n markdownResult += `---\\n\\n*Research completed in ${totalTime.toFixed(2)} seconds*`;\n\n // Convert to HTML\n const html = marked(markdownResult);\n\n console.log('Research completed');\n console.log('=== FINAL RAW ADAPTER DATA ===');\n console.log(JSON.stringify(allRawAdapterData, null, 2));\n console.log('=== END FINAL RAW ADAPTER DATA ===');\n \n return c.json({ \n text: markdownResult,\n html,\n meetings,\n rawAdapterData: allRawAdapterData,\n totalTime\n });\n\n } catch (error) {\n console.error('Server error:', error);\n console.error('Error stack:', error.stack);\n return c.json({ \n error: 'Server error: ' + error.message,\n details: error.stack \n }, 500);\n }\n});\n\n// Export app.fetch for Val Town, otherwise export app — this is only for hono apps\nexport default (typeof Deno !== \"undefined\" && Deno.env.get(\"valtown\")) ? app.fetch : app;",[],[195],"01d17dce-6809-11f0-9fb3-0224a6c84d84","deno.json","5d64c9be-6806-11f0-8f64-0224a6c84d84",19,["D",1753304834252],"{\n \"$schema\": \"https://raw.githubusercontent.com/denoland/deno/348900b8b79f4a434cab4c74b3bc8d4d2fa8ee74/cli/schemas/config-file.v1.json\",\n \"lock\": false,\n \"tasks\": {\n \"dev\": \"deno serve --allow-net --allow-env --allow-read --allow-write main.js\",\n \"start\": \"deno serve --allow-net --allow-env --allow-read main.js\"\n },\n \"compilerOptions\": {\n \"noImplicitAny\": false,\n \"strict\": false,\n \"types\": [\n \"https://www.val.town/types/valtown.d.ts\"\n ],\n \"lib\": [\n \"dom\",\n \"dom.iterable\",\n \"dom.asynciterable\",\n \"deno.ns\",\n \"deno.unstable\"\n ]\n },\n \"lint\": {\n \"files\": {\n \"include\": [\n \"deno:/https/esm.town/**/*\"\n ]\n },\n \"rules\": {\n \"exclude\": [\n \"no-explicit-any\"\n ]\n }\n },\n \"node_modules_dir\": false,\n \"experimental\": {\n \"unstable-node-globals\": true,\n \"unstable-temporal\": true,\n \"unstable-worker-options\": true,\n \"unstable-sloppy-imports\": true\n }\n}",[],[202],"5d69c036-6806-11f0-96ff-0224a6c84d84",".vtignore","5d69c216-6806-11f0-96ff-0224a6c84d84",2,["D",1753303699438],".git\n.vscode\n.cursorrules\n.DS_Store\nnode_modules\nvendor",[],[210],["D",1753304048751],[],[18,64],["D",1753304048348],[180],[18],"2dbd327c-6807-11f0-8098-0224a6c84d84","research.js","2dbd3344-6807-11f0-8098-0224a6c84d84","2d83f228-6807-11f0-a782-0224a6c84d84",15,["D",1753304048963],"import { researchTypes } from '../config/research-types.js';\n\n// Clean and extract pure subject name using LLM\nasync function cleanSearchQuery(groq, rawSubject, type) {\n const cleaningPrompt = `Extract the clean, searchable name from this raw subject for ${type} research:\n\nRaw subject: \"${rawSubject}\"\n\nInstructions:\n- If this is a person (individual), return just their name (e.g., \"John Smith\")\n- If this is a company, return just the company name (e.g., \"Apple Inc\" or \"Microsoft\")\n- Remove any meeting context, parenthetical information, or extra details\n- Return ONLY the clean name, nothing else\n- If multiple names are mentioned, pick the main subject\n\nExamples:\n- \"Ari Brown (Meeting: Vibe Coding with Maverick Capi...\" → \"Ari Brown\"\n- \"Maverick Capital background current role recent...\" → \"Maverick Capital\"\n- \"John Smith at Tesla discussing AI\" → \"John Smith\"\n- \"Microsoft (Meeting: Azure demo)\" → \"Microsoft\"\n\nClean name:`;\n\n console.log(`=== QUERY CLEANING REQUEST ===`);\n console.log(`Type: ${type}`);\n console.log(`Raw Subject: ${rawSubject}`);\n console.log(`System Prompt: Extract clean searchable names`);\n console.log(`User Prompt: ${cleaningPrompt}`);\n console.log(`=== END QUERY CLEANING REQUEST ===`);\n\n try {\n const response = await groq.chat.completions.create({\n model: 'compound-beta',\n messages: [\n {\n role: 'system',\n content: 'Extract clean, searchable names from raw subjects. Return only the name, no explanations.'\n },\n {\n role: 'user',\n content: cleaningPrompt\n }\n ],\n temperature: 0.1,\n max_tokens: 50\n });\n\n const cleanedQuery = response.choices[0].message.content.trim().replace(/['\"]/g, '');\n console.log(`=== QUERY CLEANING RESULT ===`);\n console.log(`Original: ${rawSubject}`);\n console.log(`Cleaned: ${cleanedQuery}`);\n console.log(`=== END QUERY CLEANING RESULT ===`);\n \n return cleanedQuery;\n } catch (error) {\n console.error('Query cleaning failed, using original:', error);\n return rawSubject.trim();\n }\n}\n\n// Main research function - now runs ALL adapters in parallel\nexport async function performResearch(researchAdapters, groq, type, subject, meetingContext = null, additionalContext = '') {\n const config = researchTypes[type];\n if (!config) {\n throw new Error(`Unknown research type: ${type}`);\n }\n\n // Clean the search query using LLM to extract pure subject name\n const cleanedQuery = await cleanSearchQuery(groq, subject, type);\n \n // Keep context separate for synthesis only\n const contextInfo = {\n meetingContext,\n additionalContext,\n subject: subject, // Keep original subject for context\n cleanSubject: cleanedQuery // Clean subject for display\n };\n\n const adaptersToTry = config.adapters || ['compound'];\n let allResults = [];\n let allAdapterResponses = [];\n\n console.log(`=== RESEARCH SESSION START ===`);\n console.log(`Original Subject: ${subject}`);\n console.log(`Cleaned Search Query: ${cleanedQuery}`);\n console.log(`Research Type: ${type}`);\n console.log(`Additional Context: ${additionalContext || 'None'}`);\n console.log(`Meeting Context: ${meetingContext ? meetingContext.title : 'None'}`);\n console.log(`Running ${adaptersToTry.length} adapters in parallel: ${adaptersToTry.join(', ')}`);\n console.log(`=== END RESEARCH SESSION START ===`);\n\n // Create promises for all available adapters\n const adapterPromises = adaptersToTry.map(async (adapterName) => {\n const adapter = researchAdapters[adapterName];\n if (!adapter) {\n console.warn(`Adapter ${adapterName} not found, skipping`);\n return null;\n }\n\n try {\n console.log(`=== ${adapter.name.toUpperCase()} ADAPTER START ===`);\n console.log(`Adapter: ${adapter.name}`);\n console.log(`Type: ${type}`);\n console.log(`Clean Query: ${cleanedQuery}`);\n console.log(`Additional Context: ${additionalContext || 'None'}`);\n console.log(`=== END ${adapter.name.toUpperCase()} ADAPTER START ===`);\n \n // Pass additional context to the adapter\n const result = await adapter.search(cleanedQuery, type, additionalContext);\n \n if (result) {\n // Console log the raw result\n console.log(`=== RAW ${adapter.name.toUpperCase()} RESPONSE ===`);\n console.log(`Subject: ${cleanedQuery}`);\n console.log(`Type: ${type}`);\n console.log('Raw Data:', JSON.stringify(result, null, 2));\n console.log(`=== END ${adapter.name.toUpperCase()} RESPONSE ===`);\n \n const adapterData = {\n adapter: adapter.name,\n rawData: result,\n subject: cleanedQuery,\n type: type\n };\n\n return { adapter, result, adapterData };\n }\n return null;\n } catch (error) {\n console.error(`${adapter.name} adapter failed:`, error.message);\n return null;\n }\n });\n\n // Wait for all adapters to complete\n const adapterResults = await Promise.all(adapterPromises);\n\n // Filter out null results and collect successful responses\n const successfulResults = adapterResults.filter(result => result !== null);\n \n if (successfulResults.length === 0) {\n throw new Error(`All research adapters failed for ${cleanedQuery}`);\n }\n\n console.log(`=== RESEARCH RESULTS SUMMARY ===`);\n console.log(`${successfulResults.length} out of ${adaptersToTry.length} adapters succeeded for ${cleanedQuery}`);\n console.log(`Successful adapters: ${successfulResults.map(r => r.adapter.name).join(', ')}`);\n console.log(`=== END RESEARCH RESULTS SUMMARY ===`);\n\n // Collect all raw adapter data\n allAdapterResponses = successfulResults.map(r => r.adapterData);\n\n // Process results - if we have multiple sources, synthesize them together\n let finalSynthesized = '';\n let sources = [];\n\n for (const { adapter, result } of successfulResults) {\n sources.push(adapter.name);\n \n if (adapter.name === 'Compound') {\n // Compound adapter already provides synthesized content\n if (successfulResults.length === 1) {\n // If only compound succeeded, return its content directly\n return {\n synthesized: result.content,\n rawAdapterData: allAdapterResponses\n };\n } else {\n // If we have other sources too, include compound's content in synthesis\n finalSynthesized += `\\n\\n**${adapter.name} Analysis:**\\n${result.content}`;\n }\n } else {\n // For other adapters, synthesize their raw data WITH context\n try {\n const synthesized = await synthesizeResults(groq, result, type, config, contextInfo, adapter.name);\n finalSynthesized += `\\n\\n**${adapter.name} Research:**\\n${synthesized}`;\n } catch (error) {\n console.error(`Failed to synthesize ${adapter.name} results:`, error);\n // Include raw data as fallback\n const rawContent = extractRawContent(result);\n if (rawContent) {\n finalSynthesized += `\\n\\n**${adapter.name} Data (Raw):**\\n${rawContent}`;\n }\n }\n }\n }\n\n // If we have multiple sources, create a combined synthesis\n if (successfulResults.length > 1) {\n try {\n const combinedSynthesis = await createCombinedSynthesis(groq, finalSynthesized, type, config, contextInfo, sources);\n return {\n synthesized: `${combinedSynthesis}\\n\\n*Sources: ${sources.join(', ')}*`,\n rawAdapterData: allAdapterResponses\n };\n } catch (error) {\n console.error('Failed to create combined synthesis:', error);\n // Fall back to concatenated results\n return {\n synthesized: `${finalSynthesized.trim()}\\n\\n*Sources: ${sources.join(', ')}*`,\n rawAdapterData: allAdapterResponses\n };\n }\n } else {\n // Single source result\n return {\n synthesized: `${finalSynthesized.trim()}\\n\\n*Source: ${sources[0]}*`,\n rawAdapterData: allAdapterResponses\n };\n }\n}\n\n// Extract readable content from raw adapter results\nfunction extractRawContent(adapterResult) {\n if (adapterResult.content) {\n return adapterResult.content;\n } else if (adapterResult.results?.length > 0) {\n return adapterResult.results.slice(0, 3).map(r => \n `**${r.title}**\\n${r.summary || r.content || 'No content'}\\nSource: ${r.url}`\n ).join('\\n\\n');\n } else if (typeof adapterResult === 'string') {\n return adapterResult;\n }\n return 'Data available but could not extract readable content.';\n}\n\n// Create combined synthesis from multiple adapter results\nasync function createCombinedSynthesis(groq, allData, type, config, contextInfo, sources) {\n const { subject, meetingContext, additionalContext, cleanSubject } = contextInfo;\n \n const combinedPrompt = `Synthesize the following research data from multiple sources into a comprehensive, meeting-focused briefing:\n\nResearch Subject: ${cleanSubject}\nOriginal Context: ${subject}\n${meetingContext ? `Meeting Context: ${meetingContext.title}` : ''}\n${additionalContext ? `Additional Context: ${additionalContext}` : ''}\nData Sources: ${sources.join(', ')}\n\nResearch Guidelines:\n${config.prompt}\n\nCombined Research Data:\n${allData}\n\nInstructions:\n1. Combine insights from all sources to create a comprehensive view\n2. Prioritize the most recent and relevant information for the meeting context\n3. Remove duplicate information between sources\n4. Focus on actionable insights for the upcoming meeting\n5. Highlight any discrepancies or complementary information between sources\n6. Include key dates, numbers, or specific details when available\n\nProvide a clear, unified summary that leverages the best insights from all sources.`;\n\n console.log(`=== COMBINED SYNTHESIS REQUEST ===`);\n console.log(`System Prompt: Professional research analyst preparing executive briefings...`);\n console.log(`User Prompt: ${combinedPrompt}`);\n console.log(`Model: moonshotai/kimi-k2-instruct`);\n console.log(`Temperature: 0.2`);\n console.log(`=== END COMBINED SYNTHESIS REQUEST ===`);\n\n const response = await groq.chat.completions.create({\n model: 'moonshotai/kimi-k2-instruct',\n messages: [\n {\n role: 'system',\n content: `You are a professional research analyst preparing executive briefings. Synthesize research data from multiple sources into concise, unified insights for business meetings. Combine the best information from all sources while avoiding redundancy.${config.maxLength ? ` Keep responses under ${config.maxLength} words.` : ''}`\n },\n {\n role: 'user',\n content: combinedPrompt\n }\n ],\n temperature: 0.2\n });\n\n return response.choices[0].message.content;\n}\n\n// Synthesize raw data from adapters using LLM\nexport async function synthesizeResults(groq, adapterResult, type, config, contextInfo, adapterName = 'Unknown') {\n const { subject, meetingContext, additionalContext, cleanSubject } = contextInfo;\n \n // Format the raw data in a more readable way for synthesis\n let formattedData = '';\n \n if (adapterResult.source === 'Perplexity Sonar') {\n formattedData = `Source: ${adapterResult.source}\\n\\nContent:\\n${adapterResult.content}`;\n if (adapterResult.citations && adapterResult.citations.length > 0) {\n formattedData += `\\n\\nCitations:\\n${adapterResult.citations.map(c => `- ${c}`).join('\\n')}`;\n }\n } else if (adapterResult.source === 'Exa') {\n formattedData = `Source: ${adapterResult.source}\\n\\n`;\n if (adapterResult.results && adapterResult.results.length > 0) {\n formattedData += 'Search Results:\\n';\n adapterResult.results.forEach((result, i) => {\n formattedData += `\\n${i + 1}. ${result.title}\\n`;\n if (result.summary) formattedData += `Summary: ${result.summary}\\n`;\n if (result.highlights && result.highlights.length > 0) {\n formattedData += `Key Points: ${result.highlights.join('; ')}\\n`;\n }\n if (result.publishedDate) formattedData += `Published: ${result.publishedDate}\\n`;\n formattedData += `URL: ${result.url}\\n`;\n });\n }\n } else {\n // Default formatting for other adapters\n formattedData = `Source: ${adapterResult.source || adapterName}\\n\\n${JSON.stringify(adapterResult, null, 2)}`;\n }\n\n let synthesisPrompt = `Synthesize the following research data into a concise, meeting-focused briefing:\n\nResearch Subject: ${cleanSubject}\nOriginal Context: ${subject}\n${meetingContext ? `Meeting Context: ${meetingContext.title}` : ''}\n${additionalContext ? `Additional Context: ${additionalContext}` : ''}\n\nResearch Guidelines:\n${config.prompt}\n\n${formattedData}\n\nFocus on:\n1. Most recent and relevant information for the upcoming meeting\n2. Key insights that would help in conversation\n3. Notable achievements, news, or developments\n4. Specific details that show you're well-prepared\n\nProvide a clear, actionable summary. Include key dates, numbers, or specific details when available.`;\n\n console.log(`=== ${adapterName.toUpperCase()} SYNTHESIS REQUEST ===`);\n console.log(`System Prompt: Professional research analyst preparing executive briefings...`);\n console.log(`User Prompt: ${synthesisPrompt}`);\n console.log(`Model: moonshotai/kimi-k2-instruct`);\n console.log(`Temperature: 0.2`);\n console.log(`=== END ${adapterName.toUpperCase()} SYNTHESIS REQUEST ===`);\n\n try {\n const response = await groq.chat.completions.create({\n model: 'moonshotai/kimi-k2-instruct',\n messages: [\n {\n role: 'system',\n content: `You are a professional research analyst preparing executive briefings. Synthesize research data into concise, actionable insights for business meetings. Highlight the most important and recent information.${config.maxLength ? ` Keep responses under ${config.maxLength} words.` : ''}`\n },\n {\n role: 'user',\n content: synthesisPrompt\n }\n ],\n temperature: 0.2\n });\n\n return response.choices[0].message.content;\n } catch (error) {\n console.error('Synthesis failed, returning formatted data:', error);\n // Fallback to formatted data if synthesis fails\n return extractRawContent(adapterResult);\n }\n}\n\n// Generate final strategic analysis\nexport async function generateFinalAnalysis(groq, meetings, allResearchData) {\n const analysisPrompt = `Review all the meeting research data and provide a concise, actionable summary. Focus on:\n\n1. Key insights that connect across the meetings\n2. Important context for the upcoming conversations\n3. Notable patterns or themes\n4. Specific action items or discussion points to consider\n5. Any potential synergies or connections between the people/companies\n\nBe practical and meeting-focused. Remove generic information and highlight what's most relevant for actual conversations.`;\n\n const response = await groq.chat.completions.create({\n model: 'moonshotai/kimi-k2-instruct',\n messages: [\n {\n role: 'system',\n content: 'You are an executive assistant preparing a briefing summary. Be strategic, concise, and action-oriented.'\n },\n {\n role: 'user',\n content: `${analysisPrompt}\\n\\nMeetings Data:\\n${JSON.stringify(meetings, null, 2)}\\n\\nResearch Summary:\\n${allResearchData}`\n }\n ],\n temperature: 0.2\n });\n\n return response.choices[0].message.content;\n} ",[],[232,224],"services","2d83f156-6807-11f0-a782-0224a6c84d84",7,["D",1753304048609],[178],[232],"2d9d0150-6807-11f0-9573-0224a6c84d84","research-types.js","2d9d0204-6807-11f0-9573-0224a6c84d84","2d487d10-6807-11f0-a782-0224a6c84d84",9,["D",1753304048762],"export const researchTypes = {\n company: {\n prompt: `Research this company for meeting preparation. Focus on:\n - What they do (2-3 sentences max)\n - Recent significant news, funding, or strategic moves (last 6 months) \n - Current business priorities or challenges\n - Key decision makers or leadership\n - Anything notable that would be relevant conversation topics\n \n Skip generic industry information. Focus on specific, recent, actionable insights.`,\n temperature: 0.3,\n maxLength: 250,\n // Adapter preference order: try these in sequence until one works\n adapters: ['perplexity', 'exa', /* 'apollo', */ 'compound']\n },\n individual: {\n prompt: `Research this person for meeting preparation. Focus on:\n - Current role and what they're responsible for\n - Recent professional activities or announcements \n - Areas of expertise or notable achievements\n - Recent posts, articles, or public statements (if any)\n - Anything that would help facilitate a productive conversation\n \n If multiple people have this name, try to identify the most likely one based on meeting context. Skip generic biographical information.`,\n temperature: 0.3,\n maxLength: 200,\n // Adapter preference order: prioritize real-time search for people\n adapters: ['perplexity', 'exa', /* 'apollo', 'clay', */ 'compound']\n },\n // Future research types can be added here:\n // industry: { \n // adapters: ['exa', 'perplexity', 'compound'] \n // },\n // competitive: { \n // adapters: ['exa', 'perplexity', 'compound'] \n // },\n // financial: { \n // adapters: ['apollo', 'perplexity', 'compound'] \n // }\n}; ",[],[247,239],"config","2d487c3e-6807-11f0-a782-0224a6c84d84",4,["D",1753304048219],[176],[247],"2e00337e-6807-11f0-bead-0224a6c84d84","perplexity.js","2e003450-6807-11f0-bead-0224a6c84d84","2d710a6e-6807-11f0-a757-0224a6c84d84",17,["D",1753304048765],"import { ResearchAdapter } from './base.js';\n\n// Perplexity AI Adapter\nexport class PerplexityAdapter extends ResearchAdapter {\n constructor(config = {}) {\n super('Perplexity', config);\n this.apiKey = Deno.env.get(\"PERPLEXITY_API_KEY\");\n this.baseUrl = 'https://api.perplexity.ai';\n }\n\n async search(query, type = 'general', additionalContext = '') {\n if (!this.apiKey) {\n throw new Error('PERPLEXITY_API_KEY not configured');\n }\n\n // Create focused, clean search prompts based on research type\n let searchPrompt;\n if (type === 'company') {\n searchPrompt = `Find recent information about ${query} company:\n- What does the company do and their main business\n- Recent news, funding, or strategic developments \n- Current leadership and key decision makers\n- Financial performance or recent announcements\n- Market position and recent activities\n\n${additionalContext ? `Additional context to consider: ${additionalContext}` : ''}\n\nFocus on factual, recent information with sources.`;\n } else if (type === 'individual') {\n searchPrompt = `Find recent professional information about ${query}:\n- Current role and company\n- Professional background and expertise\n- Recent activities, announcements, or achievements\n- Notable projects or public statements\n- Career highlights and current responsibilities\n\n${additionalContext ? `Additional context to consider: ${additionalContext}` : ''}\n\nFocus on factual, recent professional information with sources.`;\n } else {\n searchPrompt = `Research ${query} for business context. ${additionalContext ? `Additional context: ${additionalContext}` : ''} Provide recent, factual information with sources.`;\n }\n\n const systemPrompt = 'You are a professional research assistant. Provide concise, recent, and factual information with sources. Focus on information that would be relevant for business contexts.';\n \n const requestPayload = {\n model: 'llama-3.1-sonar-large-128k-online',\n messages: [\n {\n role: 'system',\n content: systemPrompt\n },\n {\n role: 'user',\n content: searchPrompt\n }\n ],\n max_tokens: 1500,\n temperature: 0.1,\n return_citations: true,\n search_domain_filter: [\"linkedin.com\", \"crunchbase.com\", \"bloomberg.com\", \"reuters.com\", \"techcrunch.com\", \"forbes.com\"]\n };\n\n console.log(`=== PERPLEXITY API REQUEST ===`);\n console.log(`Query: ${query}`);\n console.log(`Type: ${type}`);\n console.log(`Additional Context: ${additionalContext || 'None'}`);\n console.log(`System Prompt: ${systemPrompt}`);\n console.log(`User Prompt: ${searchPrompt}`);\n console.log(`Full Payload:`, JSON.stringify(requestPayload, null, 2));\n console.log(`=== END PERPLEXITY API REQUEST ===`);\n\n try {\n const response = await fetch(`${this.baseUrl}/chat/completions`, {\n method: 'POST',\n headers: {\n 'Authorization': `Bearer ${this.apiKey}`,\n 'Content-Type': 'application/json'\n },\n body: JSON.stringify(requestPayload)\n });\n\n if (!response.ok) {\n const errorData = await response.text();\n console.log(`=== PERPLEXITY API ERROR ===`);\n console.log(`Status: ${response.status} ${response.statusText}`);\n console.log(`Error Data: ${errorData}`);\n console.log(`=== END PERPLEXITY API ERROR ===`);\n throw new Error(`Perplexity API error: ${response.status} ${response.statusText} - ${errorData}`);\n }\n\n const data = await response.json();\n \n console.log(`=== PERPLEXITY API RESPONSE ===`);\n console.log(`Response:`, JSON.stringify(data, null, 2));\n console.log(`=== END PERPLEXITY API RESPONSE ===`);\n \n return {\n source: 'Perplexity Sonar',\n content: data.choices?.[0]?.message?.content || '',\n citations: data.citations || [],\n usage: data.usage\n };\n } catch (error) {\n console.error('Perplexity search error:', error);\n throw error;\n }\n }\n} ",[],[262,254],"adapters","2da771ee-6807-11f0-a617-0224a6c84d84","index.js","2da77284-6807-11f0-a617-0224a6c84d84",11,["D",1753304048766],"import { PerplexityAdapter } from './perplexity.js';\nimport { ExaAdapter } from './exa.js';\nimport { CompoundAdapter } from './compound.js';\n\nexport { ResearchAdapter } from './base.js';\nexport { PerplexityAdapter } from './perplexity.js';\nexport { ExaAdapter } from './exa.js';\nexport { CompoundAdapter } from './compound.js';\n\n// Factory function to create all adapters\nexport function createAdapters(groq) {\n return {\n perplexity: new PerplexityAdapter(),\n exa: new ExaAdapter(),\n compound: new CompoundAdapter(groq)\n };\n} ",[],[262,264],"2da12bc2-6807-11f0-a757-0224a6c84d84","exa.js","2da12c80-6807-11f0-a757-0224a6c84d84",10,["D",1753304048763],"import { ResearchAdapter } from './base.js';\n\n// Exa AI Adapter\nexport class ExaAdapter extends ResearchAdapter {\n constructor(config = {}) {\n super('Exa', config);\n this.apiKey = Deno.env.get(\"EXA_API_KEY\");\n this.baseUrl = 'https://api.exa.ai';\n }\n\n async search(query, type = 'general', additionalContext = '') {\n if (!this.apiKey) {\n throw new Error('EXA_API_KEY not configured');\n }\n\n // Create focused search queries and target relevant domains\n let searchQuery;\n let includeDomains = [];\n \n if (type === 'company') {\n // Use clean company name/query with minimal enhancement\n searchQuery = `${query} company business`;\n // Add additional context if provided\n if (additionalContext) {\n searchQuery += ` ${additionalContext}`;\n }\n includeDomains = ['crunchbase.com', 'linkedin.com', 'bloomberg.com', 'techcrunch.com', 'forbes.com', 'reuters.com'];\n } else if (type === 'individual') {\n // Use clean person name with minimal professional context\n searchQuery = `${query} professional linkedin`;\n // Add additional context if provided\n if (additionalContext) {\n searchQuery += ` ${additionalContext}`;\n }\n includeDomains = ['linkedin.com', 'crunchbase.com', 'twitter.com', 'medium.com', 'forbes.com', 'techcrunch.com'];\n } else {\n searchQuery = query;\n if (additionalContext) {\n searchQuery += ` ${additionalContext}`;\n }\n }\n\n const searchPayload = {\n query: searchQuery,\n type: 'neural',\n useAutoprompt: true,\n numResults: 8,\n contents: {\n text: true,\n highlights: true,\n summary: true\n }\n };\n\n // Add domain filtering if we have specific domains for this research type\n if (includeDomains.length > 0) {\n searchPayload.includeDomains = includeDomains;\n }\n\n console.log(`=== EXA API REQUEST ===`);\n console.log(`Original Query: ${query}`);\n console.log(`Type: ${type}`);\n console.log(`Additional Context: ${additionalContext || 'None'}`);\n console.log(`Final Search Query: ${searchQuery}`);\n console.log(`Include Domains: ${includeDomains.join(', ') || 'None'}`);\n console.log(`Full Payload:`, JSON.stringify(searchPayload, null, 2));\n console.log(`=== END EXA API REQUEST ===`);\n\n try {\n const response = await fetch(`${this.baseUrl}/search`, {\n method: 'POST',\n headers: {\n 'Authorization': `Bearer ${this.apiKey}`,\n 'Content-Type': 'application/json'\n },\n body: JSON.stringify(searchPayload)\n });\n\n if (!response.ok) {\n const errorData = await response.text();\n console.log(`=== EXA API ERROR ===`);\n console.log(`Status: ${response.status} ${response.statusText}`);\n console.log(`Error Data: ${errorData}`);\n console.log(`=== END EXA API ERROR ===`);\n throw new Error(`Exa API error: ${response.status} ${response.statusText} - ${errorData}`);\n }\n\n const data = await response.json();\n \n console.log(`=== EXA API RESPONSE ===`);\n console.log(`Results Count: ${data.results?.length || 0}`);\n console.log(`Autoprompt String: ${data.autopromptString || 'None'}`);\n console.log(`Full Response:`, JSON.stringify(data, null, 2));\n console.log(`=== END EXA API RESPONSE ===`);\n \n return {\n source: 'Exa',\n results: data.results?.map(r => ({\n title: r.title,\n url: r.url,\n content: r.text,\n highlights: r.highlights,\n summary: r.summary,\n score: r.score,\n publishedDate: r.publishedDate\n })) || [],\n autopromptString: data.autopromptString\n };\n } catch (error) {\n console.error('Exa search error:', error);\n throw error;\n }\n }\n} ",[],[262,272],"2db018b2-6807-11f0-853f-0224a6c84d84","compound.js","2db01970-6807-11f0-853f-0224a6c84d84",13,["D",1753304048889],"import { ResearchAdapter } from './base.js';\n\n// Compound LLM Adapter (fallback)\nexport class CompoundAdapter extends ResearchAdapter {\n constructor(groq, config = {}) {\n super('Compound', config);\n this.groq = groq;\n }\n\n async search(query, type = 'general', additionalContext = '') {\n // Default research prompts for when researchTypes isn't available\n const defaultPrompts = {\n company: `Research this company for meeting preparation. Focus on recent news, key information, and meeting-relevant insights.${additionalContext ? ` Additional context: ${additionalContext}` : ''}`,\n individual: `Research this person for meeting preparation. Focus on current role, background, and relevant professional information.${additionalContext ? ` Additional context: ${additionalContext}` : ''}`,\n general: `Provide concise, meeting-focused research on this topic.${additionalContext ? ` Additional context: ${additionalContext}` : ''}`\n };\n \n const prompt = defaultPrompts[type] || defaultPrompts.general;\n \n const systemPrompt = `You are a professional research assistant preparing briefing materials for an upcoming meeting. Be concise but informative. Focus on recent and relevant information that would be useful to know before the meeting. If you cannot find specific information, clearly state what is unknown rather than making assumptions.`;\n \n const userPrompt = `${prompt}\\n\\nSubject: ${query}`;\n\n console.log(`=== COMPOUND LLM REQUEST ===`);\n console.log(`Query: ${query}`);\n console.log(`Type: ${type}`);\n console.log(`Additional Context: ${additionalContext || 'None'}`);\n console.log(`System Prompt: ${systemPrompt}`);\n console.log(`User Prompt: ${userPrompt}`);\n console.log(`Model: compound-beta`);\n console.log(`Temperature: 0.3`);\n console.log(`=== END COMPOUND LLM REQUEST ===`);\n \n const response = await this.groq.chat.completions.create({\n model: 'compound-beta',\n messages: [\n {\n role: 'system',\n content: systemPrompt\n },\n {\n role: 'user',\n content: userPrompt\n }\n ],\n temperature: 0.3\n });\n\n const result = {\n source: 'Compound LLM',\n content: response.choices[0].message.content\n };\n\n console.log(`=== COMPOUND LLM RESPONSE ===`);\n console.log(`Response:`, JSON.stringify(result, null, 2));\n console.log(`=== END COMPOUND LLM RESPONSE ===`);\n\n return result;\n }\n} ",[],[262,280],"2dad5456-6807-11f0-a782-0224a6c84d84","base.js","2dad551e-6807-11f0-a782-0224a6c84d84",12,["D",1753304048880],"// Base Research Adapter class\nexport class ResearchAdapter {\n constructor(name, config = {}) {\n this.name = name;\n this.config = config;\n }\n\n async search(query, type = 'general') {\n throw new Error(`${this.name} adapter must implement search method`);\n }\n} ",[],[262,288],"2d7109ba-6807-11f0-a757-0224a6c84d84",6,["D",1753304048483],[170,171,172,173,174],[262],["D",1753303655869],{"_11":12,"_123":70},{"_11":12,"_119":-5,"_69":70,"_31":68,"_120":121,"_122":30},{"_39":110,"_13":16,"_33":36,"_111":74,"_112":-5,"_113":-5,"_114":-5,"_115":-5,"_116":-5,"_117":14,"_118":12},"ctx",{"_52":-5,"_142":337,"_338":339,"_340":341,"_342":30,"_343":30,"_344":30,"_345":24,"_346":-5,"_347":-7,"_348":-7,"_349":26,"_350":30,"_351":30,"_352":-5,"_353":30,"_354":30,"_355":30,"_356":-5},"ENV",{"_308":309,"_310":311,"_312":16,"_313":314,"_315":316,"_317":318,"_319":320,"_321":322,"_323":324,"_325":326,"_327":328,"_329":330,"_331":332,"_333":334,"_335":336},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","RENDER_GIT_COMMIT","e87eb04e707cf64e7e532c8e8ec395f209e17f15","RENDER_GIT_BRANCH","POSTHOG_API_TOKEN","phc_nmoyuAkStEZuUp69rYlg1DZgrIt19Gx0Y4Hwf8tGgJP","POSTHOG_API_HOST","https://wart.val.town","WEB_API_TEMPLATE","https://{val}.web.val.run","APP_URL","https://www.val.town","MODULE_URL","https://esm.town","VANITY_WEB_API_TEMPLATE","https://{label}.val.run/","DENOLS_URL","https://lsp.val.town","NODE_ENV","production","CLOUDFLARE_TURNSTILE_SITE_KEY","0x4AAAAAACtaIZJH7s9rBMZ6","GOOGLE_CLIENT_ID","875989696899-jtsqkhlbdv5jrlsk8631p7ipjg3c6o62.apps.googleusercontent.com","STRIPE_PUBLISHABLE_KEY","pk_live_51N2zvWEpZTTLipIkVEwXrUG9AT0yeOOEYxszarZT1jpsZV58Ly6wlg3SltZpHN26CUalOBmUSY7adQRQiSkCZz2500qSXNkf8Y","API_URL","https://api.val.town",[],"preferencesKeybindings","default","preferencesTheme","light","preferencesFixers","preferencesCopilot","preferencesSelectionAi","preferencesCloseBrackets","preferencesChatPanelState","preferencesDprint","preferencesDenofmt","notificationCount","flagDenols","isBusinessOrgMember","businessOrgHandle","isBusinessOrgAdmin","isHigherTownieTier","navCollapsed","currentOrgId"]

History · utils