OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
RAG & Knowledge Systems

Rag Zero Hallucinations

Rag Zero Hallucinations is a Jupyter Notebook project describing a retrieval-augmented generation pipeline for searching large document collections and producing answers with cited evidence, claim checks and abstention when support is insufficient.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Rag Zero Hallucinations — OpenViST project profile in RAG & Knowledge Systems.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use the repository as a design and notebook reference when developing a document QA prototype that needs evidence-linked answers and an explicit abstention path.

Who it is for

  • Developers prototyping document question-answering systems.
  • Researchers exploring evidence-grounded generation and calibrated abstention.
  • Teams evaluating hybrid retrieval and verification designs for large corpora.

Documented uses

  • Explore a pipeline for preparing and chunking a document corpus.
  • Combine dense vector search and sparse BM25 retrieval, then rerank results.
  • Design question routing and decomposition for multi-hop queries.
  • Generate answers with sentence-level citations and check atomic claims against cited passages.
  • Represent weak evidence as a reason to retrieve again or abstain.
Before you choose

Requirements & limits

The project is publicly available on GitHub under the MIT license. The setup description mentions a GPU and a configured client for the generator; it does not give a full hardware or model requirement list.

  • The described pipeline combines dense retrieval with BM25 rather than relying on a single retrieval method.
  • Claim-level checking and abstention are central design stages, alongside generation.
  • The notebook describes a 10M-plus vector scale target and a 200-question evaluation set; these details are project claims, not independent validation.
  • The setup discussion names GPU checking and a generator client, so adapting the notebook involves more than reading a standalone hosted service.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Initialize reproducible notebook setup

Set random seeds for a reproducible run before building the RAG pipeline.

Before you start

  • A Python notebook environment
  • NumPy
  • PyTorch is optional; the seed function handles its absence

Steps

  1. Import the standard-library modules and NumPy shown in the project README.

  2. Define the set_determinism(seed) function to seed Python, NumPy, and PYTHONHASHSEED, and to seed PyTorch when available.

  3. Call set_determinism(42).

Expected outcome

The documented setup seeds the listed random number generators before subsequent pipeline work.

Guide limitations

  • The documented setup plan also mentions checking a GPU, connecting a client to a generator, and freezing configuration, but their implementation steps are not included here.
  • Seeding does not by itself establish that later pipeline runs are reproducible.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
12 October 2026
Original website title
FareedKhan-dev/rag-zero-hallucinations · Official homepage title · retained in its source language
How we review sourcesSuggest a correction