Hey guys, does anyone know what software Nate uses for Voice to text? Would this work on Windows? For example in this video at 28:52 mark. https://www.youtube.com/watch?v=zWLZ3bVVwD8&list=PLPvKjjMeYqEr2PpQz097GC9ms-aoOInlC&index=92

Apify’s “Apollo Scraper” actor is officially dead, which broke half the lead gen workflows out there. So I rebuilt mine with a new Apify actor. Here’s how it works now: 1. Manually scrape leads via Apollo by job title and location (your niche). 2. Upload the Apollo CSV to Google Sheets. 3. The flow scrapes LinkedIn + web search via Perplexity to enrich each lead with real-time company info and context. 4. Custom messaging generated via OpenAI. 5. Leads + messaging uploaded to Instantly. Result: I now get fully personalized messages directly inside Instantly, 100 % automated. If you want to rebuild it or plug it into your own offer, happy to share the setup.

A client recently asked me if I could make all the info from their website instantly available inside WhatsApp. So I explored a RAG chatbot setup and built the following flow in n8n: - Create a form to collect the website URL. - Scrape the site with Firecrawl. - Send all the content into a Google Doc. - Push it into a MongoDB vector DB (create an index). - Connect it to WhatsApp (the trickiest part with Meta). It got me thinking back to my travels in South America & Asia, where WhatsApp is the default, this kind of setup could be a huge value-add for hotels and service businesses. Also, for more control over the conversation flow, I’d recommend layering Voiceflow on top of this. Curious if anyone else has tried something similar or sees other use cases for this stack?

So I built a RAG-based financial analyst that generates full credit reports in seconds. Here’s how it works: I used to do this manually — working as a corporate financial analyst in a bank, spending days reviewing balance sheets and credit requests, calculating repayment capacity, liquidity ratios, and debt sustainability. Now I automated the whole process. - I uploaded real company balance sheets into a Supabase vector database. - I built a RAG agent that retrieves and interprets those numbers. - I wrote a precise system prompt mirroring how a credit analyst thinks: liquidity, solvency, profitability, repayment capacity. - The chatbot now reads the data and outputs a professional-grade financial report — instantly. What used to take hours of Excel and Word work now happens in seconds. If anyone wants to see how the RAG + finance setup works or replicate it for their own use case, happy to share details.

One of the main problems with RAG systems is keeping the knowledge base fresh and accurate. - Internal documents are constantly updated in Google Drive. - Public information on the website (blog, product pages, docs) changes regularly. - Old files need to be removed, otherwise the AI risks retrieving outdated data. I just built an ingestion workflow for a SaaS client that solves these issues. Here’s how it works: 1. Continuous monitoring - Google Drive triggers for file creation, updates, and deletions. - Monthly website scraping with Firecrawl to refresh all key URLs. 2. Smart updates - Each document is hashed. If hash unchanged → skip. - If changed → old embeddings are deleted from Postgres/PGVector and replaced with new ones. - Deleted files in Drive also delete their vectors automatically. 3. Metadata for better retrieval - GPT-4.1 classifies every document as **internal** or **external** and generates a one-sentence summary. - Metadata like `file_id`, `doc_type`, and `summary` ensures more precise retrieval. 4. Vectorization pipeline - Content is normalized, split into chunks with overlap. - OpenAI embeddings are created and stored in **PGVector**. - A record manager table tracks file IDs + hashes. Result: The RAG agent always has access to the latest, cleaned, and properly categorized knowledge, both from internal docs and external web pages. No stale data, no duplicates, no hallucinations from outdated sources. If you’re building RAG systems, I’d argue this ingestion & update layer is the real bottleneck for accuracy, not just the retrieval model itself. Hope that helps!