Installed packages
Packages marked "pinned" have a fixed version. The others install at their latest
version when the runtime is built. To read the exact version in your session, run
pip show <name> in a notebook.
Python#
| Package | Version |
|---|---|
| Python | 3.12 |
| numpy | 2.3.5 (pinned) |
| pandas | 2.2 or later, below 3 |
| pyspark | 3.5.6 (pinned) |
| scikit-learn | 1.5.2 (pinned) |
| duckdb | 1.5.5 (pinned) |
| torch, torchvision (CPU) | 2.14.0, 0.29.0 (pinned) |
| mlflow | 2.20.4 (pinned) |
Data, SQL and cloud libraries#
| Package | Purpose |
|---|---|
SQLAlchemy, psycopg2-binary, pymysql, pyodbc, cx-Oracle |
Database access |
jaydebeapi |
JDBC from Python |
snowflake-, simple-salesforce, hubspot-api-client |
Warehouse and SaaS clients |
cassandra-driver, redis, azure-eventhub, influxdb3-python |
Store and stream clients |
pinecone, msgpack, paramiko |
Vector store, serialization, SFTP |
boto3, urllib3, requests, httpx, httpcore |
AWS and HTTP |
pyiceberg[s3fs, |
Iceberg from Python |
sparkmeasure |
Spark metrics |
LLM and AI libraries#
| Package | Purpose |
|---|---|
openai, anthropic, cohere, google-generativeai |
Provider SDKs |
huggingface-hub |
Model hub client |
mcp |
Model Context Protocol, 1.29 up to 2.0 |
qualipilot[gz,linking] |
Data quality and record linkage, 3.5.0 (pinned) |
Document extraction#
| Package | Version |
|---|---|
docling[rapidocr] |
2.116.0 (pinned) |
rapidocr |
3.9.2 (pinned) |
onnxruntime |
1.28.0 (pinned) |
pdfplumber, pypdf, pdf2image, pytesseract, pillow, openpyxl |
0.11, 4.2, 1.17, 0.3, 10.0, 3.1 or later |
The layout, table-structure and OCR models come with the runtime, so PDF extraction works without internet access.
Notebook and plotting libraries#
matplotlib, jupyterlab, ipython, ipykernel. The Groundzero helper packages
(GZ and others) are already importable.
JVM libraries#
| Library | Version |
|---|---|
| Spark | 3.5.6 |
| Iceberg Spark runtime and AWS bundle | 1.11.0 |
| Delta Lake | 3.3.2 |
| Hudi Spark bundle | 1.1.1 |
| AWS SDK v2 bundle | 2.25.70 |
| AWS SDK v1 bundle | 1.12.781 |
Connector drivers (Snowflake, MongoDB, Kafka, Cassandra, Elasticsearch, BigQuery, JDBC drivers and others) are preinstalled and loaded for the connection types your run uses.
System packages#
OpenJDK 17, the AWS CLI, ODBC (unixodbc) and time-zone data, so timestamps with a
time zone work with Arrow.