Vectorless, Reasoning-Based Retrieval-Augmented Generation (RAG)

Agentic RAG Harness for long documents, Tree and Graph based reasoning. Cited answers down to the pixel

Tree-based, vectorless document RAG framework. Connect any LLM via URL/API key.

Local-first AI paper reader with vectorless PDF RAG, page/bbox citations, OCR/TSR table evidence, PDF translation, and local Codex/Claude agents.

AI-first manual checklist builder using PageIndex-style vectorless retrieval + local Gemma4 to generate grounded maintenance checklists with strict citations.

Reasoning-based, vectorless RAG over a large document using a hierarchical tree (PageIndex) and a Vision-Language Model (Llama 4 Scout), no embeddings, no vector store, no text chunking.

Implements a vectorless RAG architecture using PageIndex APIs and Groq LLMs, enabling efficient document retrieval and response generation without traditional vector databases.

RAG on PDF documents without a vector database. Uploads are indexed into a hierarchical document tree via the PageIndex API — at query time, Llama 3.1 (Groq) walks the tree to select the most relevant sections, then generates a grounded answer from their full content. Includes a FastAPI backend and a simple web UI.

A production-grade, LangGraph-orchestrated fraud detection system built for regulated financial environments. Combines ML risk scoring, LLM-powered document forensics, and a Human-in-the-Loop compliance workflow — end-to-end.

Vectorless RAG for SEC 10-K filings using PageIndex — tree-based reasoning retrieval with Claude, no vector DB, no embeddings, no chunking

Hybrid RAG approach which blends Vector, Graph Database and Vectorless RAG for Retrieval of data and which will scale

Vectorless RAG using reasoning over hierarchical document structure instead of embeddings or vector databases.

A retrieval-augmented generation (RAG) system for querying ML/AI research papers using BM25 sparse retrieval — no vector embeddings or external APIs required. Users ask natural language questions and receive grounded answers with citations to the source papers.

Serverless Vectorless RAG on AWS — upload documents, ask questions, get grounded answers using LLM reasoning instead of embeddings or vector databases. Built with Amazon Bedrock (Claude 3 Haiku), Lambda, DynamoDB, API Gateway, React, and Terraform.

Vectorless semantic indexing SDK that converts large text into searchable knowledge trees for fast, structured retrieval.

An enterprise-grade, hybrid Retrieval-Augmented Generation (RAG) pipeline that completely bypasses traditional vector databases.

Enterprise-grade vectorless retrieval platform engineered for deterministic knowledge orchestration, explainable AI search, contextual document intelligence, and scalable enterprise retrieval workflows without vector embeddings.

An autonomous support triage agent powered by Vectorless RAG (PageIndex) and local LLMs to intelligently classify, route, and resolve customer tickets.

Document Q&A system using LLM Tree RAG — no vector embeddings needed. Upload files (PDF, DOCX, XLSX), build hierarchical summary trees, and ask questions with cited answers. Supports Ollama, OpenAI, Gemini & Anthropic.

A scalable, agentic document intelligence system inspired by PageIndex, designed to process long documents and enable reasoning-driven retrieval instead of vector similarity search.