Pneuma
Pneuma is an LLM-powered data discovery system for tabular data. Given a natural language query,
Pneuma searches an indexed collection and retrieves the most relevant tables for the question. It performs this search by leveraging both content (columns and rows) and context (metadata) to match tables with questions.
Getting Started
If you would like to try Pneuma without installation, you can use our Colab notebook. For local installation, you may use an OpenAI API token or a local GPU with at least 20 GB of VRAM (to load and prompt both the LLM and embedding model).
To install the latest stable release from PyPI:
$ pip install pneuma
To install the most recent version from the repository:
$ git clone https://github.com/TheDataStation/Pneuma.git
$ cd Pneuma
$ pip install -r requirements.txt
Installation Note
To ensure smooth installation and usage, we strongly recommend installing Miniconda (follow this). Then, create a new environment and install the CUDA Toolkit:
$ conda create --name pneuma python=3.12.2 -y
$ conda activate pneuma
$ conda install -c nvidia cuda-toolkit -y
Quick Start
The simplest way to explore Pneuma is by running the quickstart Jupyter notebook. This notebook walks you through Pneuma's full workflow, from data registration to querying. For those eager to dive in, here’s a snippet showcasing its functionality:
from src.pneuma import Pneuma
# Initialize Pneuma
out_path = "out_demo/storage"
pneuma = Pneuma(
out_path=out_path,
llm_path="Qwen/Qwen2.5-7B-Instruct",
embed_path="BAAI/bge-base-en-v1.5",
)
pneuma.setup()
# Register dataset & summarize it
data_path = "data_src/sample_data/csv"
pneuma.add_tables(path=data_path, creator="demo_user")
pneuma.summarize()
# Add context (metadata) if available
metadata_path = "data_src/sample_data/metadata.csv"
pneuma.add_metadata(metadata_path=metadata_path)
# Generate index
pneuma.generate_index(index_name="demo_index")
# Query the index
response = pneuma.query_index(
index_name="demo_index",
query="Which dataset contains climate issues?",
k=1,
n=5,
alpha=0.5,
)
response = json.loads(response)
query = response["data"]["query"]
retrieved_tables = response["data"]["response"]
Pneuma's CLI
To use Pneuma via the command line, refer to the CLI reference documentation for detailed instructions.
Metadata
Release files for pneuma 0.0.4
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 |
|---|---|---|---|---|
| pneuma-0.0.4.tar.gz | 28.0 kB | Without provenance | Details for pneuma-0.0.4.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| pneuma-0.0.4-py3-none-any.whl | Python 3 | none | any | Details for pneuma-0.0.4-py3-none-any.whl |
Total release size: 60.2 kB
Release files / pneuma-0.0.4.tar.gz
| Download URL | pneuma-0.0.4.tar.gz |
|---|---|
| Size | 28.0 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
17d2752059b15dc125dc0b115c54ee0c61e67dc98f50ddc09a2c97d1f186ce2e
|
|
BLAKE2b-256 checksum How to use checksums |
0163d8fc39b63a0a7ef2e673ecfe800ce2092d18a652fdc81bc69eca0f73fce2
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.12.2
|
Release files / pneuma-0.0.4-py3-none-any.whl
| Download URL | pneuma-0.0.4-py3-none-any.whl |
|---|---|
| Size | 32.2 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
d24315aa2e32a9a4ac3279ad986a8bc944b5e2f852ff113179ab1f743c4450e2
|
|
BLAKE2b-256 checksum How to use checksums |
49ca3b892b71d264be264d7dee143183c1c0dba213406e32f27fe0f3c0029979
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.12.2
|
Security for pneuma 0.0.4
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.