ragready 0.1.2

Unified data extraction and preprocessing toolkit for Retrieval-Augmented Generation (RAG) pipelines.

ragready

Unified text + metadata extractors for Retrieval-Augmented Generation (RAG) pipelines
Version 0.1.2 · MIT-licensed

---

✨ Why ragready?

A high-quality RAG knowledge base starts with clean, consistent documents—no matter where they live.
ragready streams Markdown-normalised content from:

Source type	Iterator	Notes
GitHub / GitLab repos	git_repo_iter	Auth tokens supported
Atlassian Confluence	confluence_iter	Cloud & Data Center
Public websites	website_iter	BFS crawl within domain
Local files & folders	local_iter	PDFs, DOCX, PPTX, XLSX, CSV, images (OCR), audio, ZIPs, EPUB…

Each iterator yields a single dataclass—DocumentRecord—so downstream code never worries about source-specific quirks.

---

🚀 Installation

pip install ragready

Requires Python ≥ 3.9 and a working git executable for repo extraction. The package bundles markitdown[all], so DOCX/PDF/PPTX/XLSX and OCR support work out-of-the-box.

---

⚡ Quick start

import ragready as rr
from pprint import pprint

# Crawl python.org two links deep
records = rr.website_iter(["https://www.python.org"], crawl_depth=2)

# Collect into a DataFrame (optional)
import pandas as pd
df = pd.DataFrame(r.to_dict() for r in records)
print(df[["filename", "content"]].head())

---

🍱 Example snippets

1. Local files

import ragready as rr
import pandas as pd

# Optional LLM client (leave None for pure local parsing)
client = None
llm_model = None               

# Run the iterator and capture records
docs = [
    rec.to_dict()              
    for rec in rr.local_iter(
        ["./data"],           
        llm_client=client,
        llm_model=llm_model
    )
]

# Convert to a DataFrame (optional)
df = pd.DataFrame(docs)
print(df.head())               # quick peek

2. Git repo with private access

# 1) Imports
import os
import pandas as pd
import ragready as rr

# Optional token for private repos
token = os.getenv("GITHUB_TOKEN")   # set in your shell, or leave None for public

# Pick the repos you want to scan
urls = [
    "https://github.com/pandas-dev/pandas.git",
    "https://gitlab.com/your-group/your-project.git",
]

# Run the iterator(s) and collect to dicts
git_records = [
    rec.to_dict()
    for url in urls
    for rec in rr.git_repo_iter(url, token=token)
]

# Build a DataFrame (optional)
git_df = pd.DataFrame(git_records)

# Inspect or save
print("\nGit repos preview:")
print(git_df[["source", "filename", "author", "url"]].head()) # quick peek

3. Confluence (plain-text)

import os
import pandas as pd
import ragready as rr

# Stream the pages
conf_rows = [
    rec.to_dict()
    for rec in rr.confluence_iter(
        base_url=os.getenv("CONF_URL"),       # e.g. "https://your-domain.atlassian.net/wiki"
        username=os.getenv("CONF_USER"),      # your Atlassian email / user
        api_token=os.getenv("CONFLUENCE_TOKEN"),
        space_keys=["ENG", "DS"],             # any number of spaces
        plain_text=True,                      # strip HTML tags
        limit=500                             # max pages
    )
]

# Build a DataFrame
conf_df = pd.DataFrame(conf_rows)

# 3Preview key columns
print("\nConfluence preview:")
print(conf_df[["filename", "author", "url"]].head()) # quick peek

4. Website

import pandas as pd
import ragready as rr

# Website crawl → DataFrame preview
web_rows = [
    rec.to_dict()
    for rec in rr.website_iter(
        roots=[
            "https://www.python.org",      # add more starting URLs as needed
            # "https://docs.rust-lang.org",
        ],
        crawl_depth=1                      # how deep to follow links (None = unlimited)
    )
]

web_df = pd.DataFrame(web_rows)

print("\nWebsite preview:")
print(web_df[["source", "title", "url"]].head())  # quick peek

---

🛠️ Public API

Symbol	Description
DocumentRecord	Normalised dataclass each iterator yields
git_repo_iter	Stream files from GitHub / GitLab repos
confluence_iter	Stream pages from Confluence spaces
website_iter	Breadth-first crawl within a domain
local_iter	Recursively convert local files via MarkItDown & OCR

All iterators are lazy streams—process millions of docs without filling memory.

---

🔑 Environment variables

Purpose	Variable(s)
GitHub	GITHUB_TOKEN
GitLab	GITLAB_TOKEN
Confluence	CONF_USER, CONFLUENCE_TOKEN, CONF_URL

---

📄 License

MIT © 2025 Kwadwo Daddy Nyame Owusu-Boakye

---

🤝 Contributing

1. Fork & branch off main
2. pip install -e .[dev]
3. Run pytest + ruff check before PRs

All contributions welcome — new extractors, bug fixes, or docs!

---

🙏 Acknowledgements

Built on the shoulders of:

• MarkItDown – universal document-to-Markdown converter
• GitPython, BeautifulSoup 4, pdfplumber, python-pptx, and the wider open-source community.

---

Happy extracting — your RAG pipeline will thank you! 🦾

---