Open-source PipesHub tops EnterpriseRAG-Bench at 93.24, ahead of OpenAI, Google, Microsoft and Amazon entries
Effective-Ad2060 · reddit · 2026-11-05
- Open-source PipesHub now ranks #1 on EnterpriseRAG-Bench with 93.24 overall and 95% correctness, ahead of listed systems from OpenAI, Google, Microsoft and Amazon (author discloses being a founder).
- Earlier on FRAMES, the best of 18 RAG pipelines scored 78.9% while their agent loop hit 92.7%; this benchmark is separate: 500 questions over a synthetic enterprise corpus of 500K+ documents.
- Approach: a unified context layer across documents, conversations, structured data and code; an agent loop iterates over hybrid search, grep/pattern matching, knowledge-graph navigation and document-reading tools — search, read what comes back, then decide what to search next.
- The platform is open source, self-hostable, supports open models, and exposes MCP/APIs/SDKs for custom agents. Repo: pipeshub-ai/pipeshub-ai.
Related event: Open-Source PipesHub Tops EnterpriseRAG-Bench, Leaving Google Far Behind(2 posts)→
More from coding & agent
- Gradium and SurrealDB demo a fully on-device research agent with memory and streaming TTS — mattturck · 2026-11-05
- webIQ demo shows the evidence object returned to models, stressing freshness signals — VeryWellVersed · 2026-11-05
- Paper: rewriting reasoning histories into compact records cuts agent token costs — Anton Hristov · 2026-11-05
- KernelSeek: question-driven agent hits 1.13x FlashAttention-3 where tool-rich agent stalls at 0.50x — Snowflake · 2026-11-05
- Orbit beta puts Claude Code, Codex and Grok in one app with multi-account switching — Available-Ganache-71 · 2026-11-05
- Stop saying 'write code': 8 senior-engineer prompt templates for Claude — aitrendz_xyz · 2026-11-05