Skip to main content
What it contains
Vectorless reasoning-based RAG that builds hierarchical document trees and lets LLMs reason over them instead of vector similarity/chunking.
Typical uses
Retrieve traceable context from long professional PDFs/Markdown using self-hosted parsing or enhanced cloud/MCP/API options.
Why people choose it
Inference: chosen when document structure/reasoned relevance matters more than conventional embedding similarity.
Last meaningful updateUnknown
Production readinessdevelopment
Classification
Subject domains
RAGdocument retrievallong documents
Task categories
RAGdocument retrievallong documents
Project phases
researchimplementation
Secondary types
RAG frameworkAPI/MCP companion
Select when
- Need vectorless explainable retrieval over long structured documents
Do not select when
- Simple vector search is sufficient
- Need only OCR
Implementation complexitymedium
Routing heuristic 74.0%
Security boundary clarity
3/5 First-party routing documentation
5/5 Strengths
None documented.
Poor-fit scenarios
- Simple vector search is sufficient
- Need only OCR
Limitations
None documented.
run_pageindex.py
LLM-assisted tree generation.
treesummaries
pageindex/flash
Fast heuristic tree extraction.
fast tree
examples/agentic_vectorless_rag_demo.py
Tree-search retrieval demo.
agentic retrieval
cookbook/
Vectorless/vision RAG notebooks.
examples
FrameworksNone documented
Operating systemsNone documented
Installation methodsNone documented
Interfaceslibrary, CLI, API, MCP
Required credentialsLLM key for LLM-assisted modes
External servicesoptional PageIndex cloud, LiteLLM provider
Major dependenciesNone documented
Compatibility notes
None documented.
One-sentence semantic summary
Vectorless reasoning-based RAG that builds hierarchical document trees and lets LLMs reason over them instead of vector similarity/chunking.
Capability keywords
tree indexingvectorless retrievalagentic tree searchPDFMarkdownflash mode
User intent phrases
- Need vectorless explainable retrieval over long structured documents
Negative match phrases
- Simple vector search is sufficient
- Need only OCR
Differentiators
None documented.
Security considerations
None documented.
Privacy considerations
- Cloud/API path sends document content to configured service.
Uncertainty
Unresolved items
- License not conclusively verified in the inspected first-party material.
Inference notes
None documented.
Repository usage guidance
installation summaryUnknown
basic usage summaryRetrieve traceable context from long professional PDFs/Markdown using self-hosted parsing or enhanced cloud/MCP/API options.
documented entry pointsNone documented
key configuration filesNone documented
important directoriesNone documented
documentation pathsREADME.md
example pathsNone documented
Relationships
alternative
high
PageIndex retrieves from documents; book-to-skill compiles stable sources.
Recommended combinations
Long-document research pipeline high
Each repository covers a distinct layer; avoid combining tools that duplicate lifecycle ownership.
Order of operations
- Acquire authorized sources
- Index/retrieve long docs with PageIndex
- Validate against sources
- Compile stable reusable corpus with book-to-skill when worthwhile
Routing rules
rule_044: Need vectorless explainable retrieval over long structured documents P956
Rationale
Vectorless reasoning-based RAG that builds hierarchical document trees and lets LLMs reason over them instead of vector similarity/chunking.
Required conditions
None documented.
Preferred
VectifyAI/PageIndex
Fallback
None