Advanced multi-source AI research with Bright Data, OpenAI, Redis

How it Works

This workflow transforms natural language queries into research reports through a five-stage AI pipeline. When triggered via webhook (typically from Google Sheets using the companion google-apps-script.js (GitHub gist), it first checks Redis cache for instant results.

For new queries, GPT-4o breaks complex questions into focused sub-queries, optimizes them for search, then uses Bright Data's MCP Tool to find the top 5 credible sources (official sites, news, financial reports). URLs are scraped in parallel, bypassing bot detection.

GPT-4o extracts structured data from each source: answers, facts, entities, sentiment, quotes, and dates. GPT-4o-mini validates source credibility and filters unreliable content. Valid results aggregate into a final summary with confidence scores, key insights, and extended analysis.

Results cache for 1 hour and output via webhook, Slack, email, and DataTable—all in 30-90 seconds with 60 requests/minute rate limiting.

Who is this for?

• Research teams needing automated multi-source intelligence
• Content creators and journalists requiring fact-checked information
• Due diligence professionals conducting competitive intelligence
• Google Sheets power users wanting AI research in spreadsheets
• Teams managing large research volumes needing caching and rate limiting

Setup Steps

Setup time: 30-45 minutes

Requirements:

• Bright Data account (Web Scraping API + MCP token)
• OpenAI API key (GPT-4o and GPT-4o-mini access)
• Redis instance
• Slack workspace (optional)
• SMTP email provider (optional)
• Google account (optional for Sheets integration)

Core Setup:

1. Get Bright Data Web Scraping API token and MCP token
2. Get OpenAI API key
3. Set up Redis instance
4. Configure critical nodes:
   • Webhook Entry: Add Header Auth token
   • Bright Data MCP Tool: Add MCP endpoint with token
   • Parallel Web Scraping: Add Bright Data API credentials
   • Redis Nodes: Add connection credentials
   • All GPT Nodes: Add OpenAI API key (5 nodes)
   • Slack/Email: Add credentials if using

Google Sheets Integration:

1. Create Google Sheet
2. Open Extensions → Apps Script
3. Paste the companion google-apps-script.js code
4. Update webhook URL and auth token
5. Save and authorize

Test: {"prompt": "What is the population of Tokyo?", "source": "Test", "language": "English"}

Customization Guidance

• Source Count: Change from 5 to 3-10 URLs per query
• Cache Duration: Adjust from 1 hour to 24 hours for stable info
• Rate Limits: Modify 60/minute based on usage needs
• Character Limits: Adjust 400-char main answer to 200-1000
• AI Models: Swap GPT-4o for Claude or use GPT-4o-mini for all stages
• Geographic Targeting: Add more regions beyond us/il
• Output Channels: Add Notion, Airtable, Discord, Teams
• Temperature: Lower (0.1-0.2) for facts, higher (0.4-0.6) for analysis

Once configured, this workflow handles all web research, from fact-checking to complex analysis—delivering validated intelligence in seconds with automatic caching.

Built by Daniel Shashko
Connect on LinkedIn