PageIndex hits 36.7k stars: vectorless, reasoning-based RAG document indexing
VectifyAI · github · 2026-09-29
VectifyAI's open-source PageIndex is trending on GitHub with 36.7k stars (+822 today). It takes a "vectorless" approach to RAG: instead of embedding chunks into a vector database, it builds a tree-structured document index that an LLM navigates via reasoning to locate relevant content.
- Positioning: Document Index for Reasoning-based RAG
- Approach: structured table-of-contents navigation by the model rather than similarity search; topics include agentic-ai, context-engineering, reasoning
- Use case: long or hierarchical documents where vector retrieval underperforms
For RAG and context-engineering builders, it's a direct challenge to the embedding-plus-vector-DB paradigm and worth benchmarking against your current stack.
More from coding & agent
- Built a Full Halloween Game in Minutes: Code, Art and Music All AI-Generated — iamfakhrealam · 2026-09-29
- Anthropic engineer runs 100+ Claude Code agents with Chief and PM agents managing the workflow — anthara_ai · 2026-09-29
- 6 Claude Skills for AI-built mobile apps, from architecture to real-device testing — JohnAlexander · 2026-09-29
- Adding persistent memory to a retail AI agent with Hindsight — and hitting conflicting memories — Sreekousik1_G · 2026-09-29
- TLA+ for agent heartbeats: 1.5M states reduced to ~4K, 10 bugs fixed — please-dont-deploy · 2026-09-29
- Running 1000s of enterprise RAG indexes: a year of lessons in incremental sync — srnsnemil · 2026-09-29