🔍 HOW NEURALAI SEEK WORKS

  • Prompt analysis — AI understands the complexity of the request
  • Web search — Performs searches on Google and authoritative sources
  • Source synthesis — Combines information from 5-15 sites
  • Source citation — Every response includes links to original sources
seek_speed

Quick response, simple questions

Sources3-5 sources
Time~5-10 sec
Cost20 tokens
seek_normal

Balanced standard search

Sources5-8 sources
Time~10-20 sec
Cost40 tokens
seek_large

Deep research, complete reports

Sources8-15 sources
Time~20-40 sec
Cost80 tokens

💡 WHEN TO USE THE MODES

  • seek_speed: "What time is it in Tokyo?", "Who is the US president?" (simple factual questions)
  • seek_normal: "Best Italian restaurants in London", "How to learn Python in 3 months" (practical searches)
  • seek_large: "Impact of AI on the job market 2025-2030", "Comparative analysis of renewable energy" (in-depth reports)

Example: Automatic search (recommended)

javascript
// The Orchestrator automatically decides the seek mode based on the prompt
const response = await fetch("https://app.neuralaiweb.com/functions/neuralaiPublicApi", {
  method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({

Example: Force specific mode

javascript
// Force Deep Research for a complete report
const response = await fetch("https://app.neuralaiweb.com/functions/neuralaiPublicApi", {
  method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({

🎬 AUTOMATIC PRESENTATION PDF (16:9)

  • If the prompt expresses a presentation intent (e.g. "make a presentation about X", "create slides about Y"), the API automatically generates a 16:9 PDF from the response content: one slide per section, cover and final sources slide
  • The response includes result.pdf_presentation: true, result.pdf_url (direct link to the PDF), result.pdf_file_name and result.pdf_pages
  • Watch out for pending (async worker): if the deck is not ready within the first response budget, the result contains result.pdf_pending: true + result.pdf_job_id instead of pdf_url. Repeat the same request adding the poll_pdf_job_id parameter (recommended loop: 12 attempts, ~10s interval): polling is free and the PDF arrives as soon as the worker job completes.
  • Additional cost: 90 tokens per every 10 pages — active with chat, seek_normal, seek_large and deep_research, no parameter required

✅ NEURALAI SEEK ADVANTAGES

  • Verified sources — Only authoritative and reliable sites
  • Up-to-date — Real-time information, not frozen at model training
  • Transparent — Every statement is linked to the original source
  • Adaptive — The AI automatically chooses the optimal search depth