- 10th Floor, CEO Tower, Lot HH2-1, Me Tri Ha Urban Area, Pham Hung Street, Tu Liem District, Hanoi, Vietnam
- 8.30 a.m - 6.00 p.m
Blog Details
- Home
- <
- Blog Details
By Dann WaneriMay 26, 2026
Artificial Intelligence0
Artificial Intelligence0Built a 100k-Document RAG System by Hand
- Author built a 100k-document RAG system by hand using a hybrid BM25 + vector search engine on Cloudflare Workers.
- The system was tested with the Hermes Agent Challenge.
- The author encountered several friction points during the installation process, including a Windows-specific installer and dependency issues.
- The author used PyPI to install the Hermes agent, which was not mentioned in the official flow.
- The experiment highlights the importance of considering different operating systems and dependencies when building and testing AI systems.
This is a submission for the Hermes Agent Challenge
For the past six months, I have been building what Hermes does by hand.
Not as a thought experiment. Actually building it: a hybrid BM25 + vector search engine on Cloudflare Workers, a Gemma 4 MoE reflection layer, 100,000+ documents indexed, an MCP server with Durable Objects, multimodal image ingestion with Llama 4 Scout.
curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash
It works amazingly well.
