soigia 0.1.9

Python SDK for Soi Gia

soigia

soigia is a Python SDK for dataframe workflows, internal UI screens, shared YAML configuration, and business pipelines.

Getting Started

pip install soigia

import soigia

df = soigia.DataTable({"name": ["alice"], "age": [19]})
print(df)

For local development:

pip install -e .

What It Includes

• soigia.datatable.DataTable for pandas-like data work with extra helpers.
• soigia.ui for Streamlit-based internal screens.
• soigia.pipeline.BasePipeline for business workflows.
• soigia.config for shared YAML configuration and multi-account storage.

Quick Examples

DataTable

from soigia.datatable import DataTable

df = DataTable({"name": ["alice", "bob"], "age": [19, 24]})
print(df.objects.filter(age__gte=20).values("name", "age"))

Reconciliation

from soigia.datatable import DataTable

left = DataTable(
    {
        "id": [1, 2],
        "name": ["alice", "bob"],
        "amount": [10, 20],
    }
)
right = DataTable(
    {
        "id": [1, 2],
        "name": ["alice", "bobby"],
        "amount": [10, 25],
    }
)

bundle = left.reconcile_bundle(
    right,
    key_columns=["id"],
    title="Oracle vs Kudu Reconciliation",
    html_path="tests/reconcile-report.html",
    json_path="tests/reconcile-report.json",
    open_report_in_browser=True,
)

print(bundle["html_path"])
print(bundle["json_path"])
print(bundle["payload"]["summary"])

If you only want the HTML string, use reconcile_html_report(). If you want the report opened immediately, use reconcile_open().

UI

from soigia.ui import ui

orders = [
    {"id": 1, "customer": "Alice", "amount": 120.0, "status": "paid"},
    {"id": 2, "customer": "Bob", "amount": 88.5, "status": "pending"},
]


@ui.page("Orders")
def orders_page(ctx):
    ui.text("Order Dashboard")
    keyword = ui.text_input("Search")
    ui.metric("Total orders", len(orders))
    ui.table(lambda: [row for row in orders if keyword.value.lower() in row["customer"].lower()])


ui.run()

Pipeline

from soigia.pipeline import BasePipeline


class SalesPipeline(BasePipeline):
    stages = ["normalize", "summarize"]

    def load_data(self):
        return build_orders()

    def normalize(self):
        self.data_df = self.data_df.copy()
        self.data_df["amount"] = self.data_df["amount"].astype(float)

    def summarize(self):
        self.model.total_rows = len(self.data_df)


pipeline = SalesPipeline(name="sales")
result = pipeline.run()
print(result.summary_path)

Documentation

• Module Reference
• Pipeline End-to-End Guide
• UI End-to-End Guide

Configuration

Use soigia.config to load and save a shared YAML file for the whole project.

Recommended workflow:

• keep config.example.yaml in git
• copy it to config.yaml on your machine
• keep the real config.yaml out of version control
• never commit secrets, tokens, keys, or chat IDs in the real file

Pipeline modules can also auto-load a config.yaml file placed next to the pipeline file. It is exposed as self.config with dot access:

if self.config.min_amount:
    ...

Examples

Run the bundled demos:

python -m examples.pipeline_final_template
python -m examples.pipeline_yaml_config_demo
make ui-config
make ui
make ui-users
make ui-csv
make ui-markdown
make ui-jinja

Release

Build distribution artifacts:

python -m pip install --upgrade build
python -m build

Upload to PyPI:

python -m pip install --upgrade twine
python -m twine upload dist/*

Run the test suite before release:

python -m pytest tests -q

Publish Checklist

1. Bump the version in pyproject.toml.
2. Update CHANGELOG.md.
3. Run python -m pytest tests -q.
4. Run python -m build --no-isolation.
5. Upload with python -m twine upload dist/*.
6. Tag the release in git.

Project Layout

• soigia/
• examples/
• docs/
• tests/

Generated Files

• data/<ClassName>/<step>.csv
• data/<ClassName>/<ClassName>.csv
• data/<ClassName>/<ClassName>.sqlite3
• data/<ClassName>/<ClassName>.parquet
• data/<ClassName>/pipeline_summary.md
• logs/<ClassName>.log

Optional Features

• Google Sheets helpers usually need gspread and Google service-account support.
• Shared account config screens need pyyaml.
• Jinja template examples need jinja2.
• Excel helpers need an Excel engine such as openpyxl.
• Parquet and Feather helpers need a parquet/arrow backend such as pyarrow.
• Fake data helpers work best with faker, but soigia also includes a small fallback generator.

Core Ideas

DataTable keeps the pandas mental model, but adds a few opinionated helpers:

• objects for queryset-style chaining
• values() for record-style output
• filter() and exclude() for row filtering
• order_by() and distinct() for common table operations
• join() for relational merges
• validate_schema() for lightweight schema checks
• init_versions(), snapshot(), and auto_version() for version tracking

Security Note

from_pickle() should only be used with trusted input. Pickle files can execute code during loading.