# On The Intrinsic Knowledge Retrieval Scaling Challenge
We study how, and in what way, scaling of a corpus affects basic retrieval recall (semantic search and BM25 search).

## Authors
- [Joachim Rahmfeld](/content/blog/retrieval-corpus-scaling/index.html)
- [Yuhong Sun](/content/blog/retrieval-corpus-scaling/index.html)

---

## Case Study: How Ramp uses Onyx to support their continued growth
Learn how Ramp uses Onyx as their internal knowledge platform to achieve 30x ROI.

**Author:** Chris Weaver  
**Date:** Aug 20, 2025  
[Read More](/content/blog-assets/ramp-case-study.pdf)

---

## Are File Systems All You Need? It Depends...
We evaluated hybrid search against file search (glob/grep/read) on ~160k–510k enterprise-style documents. Neither wins outright everywhere — question complexity and dataset scale, as well as user preferences, affect which retrieval pattern works better.

**Author:** Joachim Rahmfeld  
**Date:** Apr 7, 2026  
[Read More](/content/blog/file-search-vs-hybrid-search/index.html)

---

## Benchmarking agentic RAG on workplace questions
Onyx outperformed ChatGPT Enterprise, Claude Enterprise, and Notion AI in an RAG benchmark using our internal and web data, complementing our recent Deep Research benchmark success.

**Authors:** Roshan Desai, Joachim Rahmfeld  
**Date:** Feb 19, 2026  
[Read More](/content/blog/benchmarking-agentic-rag-on-workplace-questions/index.html)

---

## Lessons from building the best Deep Research (and how you can build better agents)
A practical, behind-the-scenes look at how we built the #1 Deep Research system and the design choices that create performant agent systems.

**Author:** Yuhong Sun  
**Date:** Feb 17, 2026  
[Read More](/content/blog/building-the-best-deep-research/index.html)

---

## Building an Internet Search to rival OpenAI
We reverse-engineered how top agents run web search, then built the same two-tool workflow in Onyx for fast, reliable, low-hallucination internet answers.

**Author:** Yuhong Sun  
**Date:** Sep 18, 2025  
[Read More](/content/blog/building-internet-search/index.html)

---

## Onyx raises $10M to help teams find knowledge faster with open source AI agents
Onyx raised $10M to expand open-source enterprise AI search and agents, helping teams unlock institutional knowledge across tools like Slack, Drive, and Salesforce.

**Author:** Chris Weaver  
**Date:** Mar 12, 2025  
[Read More](/content/blog/seed-round/index.html)

---

## Beyond RAG: Implementing Agent Search with LangGraph for Smarter Knowledge Retrieval
A technical deep dive into how Onyx built Agent Search with LangGraph, including decomposition, parallel sub-queries, refinement loops, and orchestration tradeoffs.

**Authors:** Evan Lohn, Joachim Rahmfeld  
**Date:** Feb 21, 2025  
[Read More](/content/blog/agent-search-with-langgraph/index.html)

---

## Introducing Agent Search: The Next Evolution in Enterprise Knowledge Discovery
Introducing Agent Search: Onyx's multi-step retrieval system that breaks down complex questions, runs parallel investigations, and synthesizes higher-quality answers.

**Author:** Wendi Ding  
**Date:** Feb 18, 2025  
[Read More](/content/blog/agent-search/index.html)

---

## Onyx and Unstructured: Elevating Enterprise Search Capabilities
How the Onyx + Unstructured integration improves document parsing fidelity, expands file support, and boosts enterprise search and RAG quality.

**Author:** Yuhong Sun  
**Date:** Oct 21, 2024  
[Read More](/content/blog/onyx-and-unstructured-elevating-enterprise-search-capabilities/index.html)

---

## Empower Your Data: Local Document Chat with Llama 3.1
A step-by-step guide to running local document chat with Llama 3.1 and Danswer, so teams can use powerful AI over private data without sending it to third parties.

**Author:** Chris Weaver  
**Date:** Aug 22, 2024  
[Read More](/content/blog/empower-your-data-local-document-chat-with-llama-3-1/index.html)