ai-agent

verbatim-rag: Ground‑Truth LLM Answers with Exact Citations

Name: verbatim-rag: Ground‑Truth LLM Answers with Exact Citations
Author: KRLabsOrg

Provides founders a hallucination‑free RAG pipeline that extracts verbatim spans and cites source documents, ideal for reliable AI‑driven content.

From the repo · “Provenance-first extractive RAG: return verbatim source spans with citations using local ModernBERT or optional LLM-assisted extraction.”

207 stars+1★ this week39 forksPythonGuide quality 9/10 · reviewed Oct 11Updated 10/9/2026100% free · open source

What it does

Verbatim-rag is a hallucination-prevention tool that ensures AI-generated content is grounded in source documents with exact citations.

When to use it

•When using AI agents to generate content that requires accurate citations
•When needing to prevent hallucinations in AI-generated text based on source documents
•When requiring a RAG system with verbatim span extraction for content generation

Ready-to-paste prompt

python main.py --input 'What are the main points of the document?' --doc 'path/to/source/document.txt' --output 'output.txt'

Heads up: Ensure you have the correct Python version installed, as specified in the README, to avoid compatibility issues with the required packages

Saves to your device

Use with Claude

New

Skip the builder — one click puts this in Claude, Cursor, Antigravity and more.

🔑 Needs an API key

Works after you add credentials — the setup agent will ask you for them.

Try it instantly — no install

Claude Code

mkdir -p ~/.claude/skills/verbatim-rag && curl -fsSL https://workflowstacks.com/api/skills/verbatim-rag/claude-skill -o ~/.claude/skills/verbatim-rag/SKILL.md

Open in another AI app

Opens the app with this repo with the prompt ready to go — no copy-paste needed.

Connect the whole catalog (MCP)

claude mcp add --transport http workflowstacks https://workflowstacks.com/api/mcp

Adds a WorkflowStacks connector to Claude Code: search and load any skill here by chatting.

How verbatim-rag: Ground‑Truth LLM Answers with Exact Citations works

Codeflow

Free to inspect

Verbatim-rag: Ground‑Truth LLM Answers with Exact Citations is a medium Python project (~14k lines across 87 code files, plus 9 test files). Setup is light: runs with Docker — add your API keys to .env. Last commit this month, MIT license, has a test suite.

Size

Medium codebase

~14k lines · 87 code files · ~2 h to skim

Setup

Light setup

Runs with Docker — add your API keys to .env.

Runs on

Python · Docker

Needs API keys (.env)

Python 90%JavaScript 9%

Where to start reading

1
README.md
Start here — what it does and how to install it
2
verbatim_rag/cli.py
Where the program starts running
3
pyproject.toml
Dependencies and the commands it exposes
4
.env.example
The API keys and settings you must provide
5
verbatim_rag/__init__.py
Inside verbatim_rag/ — the main logic begins here

What's in each folder

verbatim_rag/Core code — the actual logic33 files

frontend/Frontend / UI36 files

packages/Sub-packages / plugins33 files

api/Backend / API8 files

docs/Documentation11 files

examples/Examples you can copy6 files

scripts/Helper scripts3 files

docker/Deployment / infrastructure2 files

READMEHas testsDocumentedCI checksDocker readyExamples includedMIT licenseUpdated this month

Use with Claude

New

Skip the builder — one click puts this in Claude, Cursor, Antigravity and more.

🔑 Needs an API key

Works after you add credentials — the setup agent will ask you for them.

Try it instantly — no install

Claude Code

mkdir -p ~/.claude/skills/verbatim-rag && curl -fsSL https://workflowstacks.com/api/skills/verbatim-rag/claude-skill -o ~/.claude/skills/verbatim-rag/SKILL.md

Open in another AI app

Opens the app with this repo with the prompt ready to go — no copy-paste needed.

Connect the whole catalog (MCP)

claude mcp add --transport http workflowstacks https://workflowstacks.com/api/mcp

Adds a WorkflowStacks connector to Claude Code: search and load any skill here by chatting.

Quick Actions

Details

Creator

KRLabsOrg

Language

Python