raghilda 
RAG made simple.
raghilda is a Python package for implementing Retrieval-Augmented Generation (RAG) workflows. It provides a complete solution with sensible defaults while remaining transparent (not a black box).
What you can build
raghilda puts a question-answering layer over content you already have. Use it to chat over your documentation with cited sources, build internal knowledge assistants over runbooks, wikis, or policies (scoped per team or product with attribute filters), run research and Q&A across a directory of PDFs, notebooks, or reports, or power entity-scoped retrieval over customer records, product catalogs, and case timelines where every answer must stay within a single record.
Installation
pip install raghilda
Or install from GitHub:
pip install git+https://github.com/posit-dev/raghilda.git
Key Steps
raghilda handles the complete RAG pipeline:
- document Processing: convert documents to Markdown using MarkItDown
- text chunking: split text at semantic boundaries (headings, paragraphs, sentences)
- embedding: generate vector representations via OpenAI or other providers
- storage: store chunks and embeddings in DuckDB, ChromaDB, or OpenAI Vector Stores
- retrieval: find relevant chunks using similarity search or BM25
Each step is an ordinary, inspectable Python object with sensible defaults, so you can run the whole pipeline as-is and customize any stage when you need to.
Usage
from raghilda.store import DuckDBStore
from raghilda.embedding import EmbeddingOpenAI
from raghilda.scrape import find_links
from raghilda.read import read_as_markdown
from raghilda.chunker import MarkdownChunker
# Create a store with embeddings
store = DuckDBStore.create(
location="chatlas.db",
embed=EmbeddingOpenAI(),
)
# Find and index pages from the chatlas documentation
links = find_links("https://posit-dev.github.io/chatlas/")
chunker = MarkdownChunker()
# Read, chunk, and store each page
for link in links:
document = read_as_markdown(link)
chunked_document = chunker.chunk(document)
store.upsert(chunked_document)
# Build indexes before retrieval
store.build_index()
# Retrieve relevant chunks
chunks = store.retrieve("How do I stream a response?", top_k=5)
for chunk in chunks:
print(chunk.text)
Requirements
- Python 3.11–3.13
- An embedding provider (such as OpenAI) is optional: semantic search needs one, but keyword search works with no API key.
Links
Contributing
Contributions are welcome! See CONTRIBUTING for setup and development guidelines.
License
Released under the MIT License.
Developed by Daniel Falbel and Tomasz Kalinowski.
Supported by Posit Software, PBC.
Metadata
Release files for raghilda 0.2.1
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| raghilda-0.2.1.tar.gz | 85.5 kB | Without provenance | Details for raghilda-0.2.1.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| raghilda-0.2.1-py3-none-any.whl | Python 3 | none | any | Details for raghilda-0.2.1-py3-none-any.whl |
Total release size: 183.8 kB
Release files / raghilda-0.2.1.tar.gz
| Download URL | raghilda-0.2.1.tar.gz |
|---|---|
| Size | 85.5 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
d04d5cabc7a9831d5be44ec419c1e0c06258613b76a3d0b0f3bf90c3c29ec96e
|
|
BLAKE2b-256 checksum How to use checksums |
b6f405ef65e1447ba5af7a6d7bb58ad9eb1e5502872464f0cca1d3a518932616
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
uv/0.12.9 {"installer":{"name":"uv","version":"0.12.9","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
Release files / raghilda-0.2.1-py3-none-any.whl
| Download URL | raghilda-0.2.1-py3-none-any.whl |
|---|---|
| Size | 98.3 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
96f92fea1a7694177b9792f2e39693a998d40886b1d97f1c7c343364f164f7a3
|
|
BLAKE2b-256 checksum How to use checksums |
8d8f30b746f9a152ac4977c4340f38c00cbc8d55b23100c16154ae2291e371db
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
uv/0.12.9 {"installer":{"name":"uv","version":"0.12.9","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
Security for raghilda 0.2.1
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.