Deep Research.....but Open Source

NetworkChuck Technology Careers Career & Money

About this video

Chuck compares OpenAI’s Deep Research with an open-source implementation he found via a tweet, motivated by price: Deep Research requires ChatGPT Pro at $200 a month. He explains that Deep Research spends 5 to 30 minutes doing multi-step web research, reading pages and cross-referencing them, producing a long report with citations, and that it is multimodal across text, images and PDFs; it also asks clarifying questions before starting. The open-source version uses the o3-mini high model and needs an OpenAI API key on pay-as-you-go pricing, a free Firecrawl API key (free up to 500 credits), and Node.js, which he runs inside Docker. His walkthrough clones the repo, creates a .env.local file holding the Firecrawl and OpenAI keys, adds a Dockerfile and docker-compose.yml, tweaks package.json — fixing a missing comma that broke the JSON — and launches it with docker compose run. The tool asks what to research, then prompts for breadth (default 4) and depth, asks similar clarifying questions to OpenAI’s, and saves its output to a report.md file, though he hits a Firecrawl rate limit during the run. On a cats-versus-dogs question both tools concluded dogs are better, and on which operating system is best for IT professionals both landed on using all three of Mac, Linux and Windows rather than picking one. He closes by noting Hugging Face had just announced its own agentic deep search framework as he recorded.

Key takeaways

OpenAI’s Deep Research costs $200 a month via ChatGPT Pro; the open-source alternative runs pay-as-you-go on an OpenAI API key
Deep Research takes 5 to 30 minutes, does multi-step web research with citations, is multimodal, and asks clarifying questions first
Setup needs two keys — Firecrawl (free up to 500 credits) and OpenAI — stored in a .env.local file, with the app run via Docker Compose
The open-source tool prompts for research breadth (default 4) and depth and saves its report to report.md
In his side-by-side tests both tools agreed: dogs over cats, and IT professionals should use Mac, Linux and Windows rather than one OS

Episode details: published February 6, 2025 · 13:29 long. Summary written by Self Growth Videos from the episode itself. Business results and any income figures mentioned are NetworkChuck’s own claims, shared for education — they are not typical or guaranteed.

← Previous: Your Remote Desktop SUCKS!! Try this instead (FREE + Open Source) · Next: I built an AI supercomputer with 5 Mac Studios →

All NetworkChuck videos · NetworkChuck profile · NetworkChuck on YouTube