PyForecast

Desktop forecasting system that ingests Excel/CSV data, detects dataset shape + time frequency, normalises to a canonical long format, and runs Prophet-based time series forecasts. :contentReference[oaicite:4]{index=4}

Status: early stage (v0.0.1) — APIs and UI may change.

---

What PyForecast does

PyForecast is built around a practical workflow:

1. Ingest: load a dataset (Excel/CSV) into the app.
2. Profile: detect the dataset “shape” (long vs wide) and infer the time frequency (daily/weekly/monthly, etc.).
3. Map: choose which columns represent date/value and what composes your entity key.
4. Transform: reshape the dataset into a canonical time-series format.
5. Forecast: produce forecasts using Prophet (optional extra). :contentReference[oaicite:5]{index=5}

---

Canonical output format

After transformation, PyForecast aims to normalise data to a standard long-format time series:

• cd_key — entity identifier (built from one or more columns)
• ds — timestamp/date column
• y — numeric target value

This makes forecasting consistent, repeatable, and model-agnostic.

---

Installation

Requirements

• Python 3.10+ :contentReference[oaicite:6]{index=6}

1) Create and activate a virtual environment (recommended)

Windows (PowerShell):

python -m venv .env
.env\Scripts\Activate.ps1
python -m pip install --upgrade pip

2) Install the project

Minimal install (UI base):

pip install -e .

Recommended install (data + excel + forecasting):

pip install -e ".[data,excel,forecast]"

Build tooling (PyInstaller):

pip install -e ".[build]"

Extras available (GitHub):

• data: polars, duckdb, pyarrow
• excel: python-calamine
• forecast: prophet
• build: pyinstaller
• dev: pytest, ruff, mypy
• all: pyforecast[data,excel,forecast]

---

Running the app

PyInstaller builds from src/pyforecast/main.py. (GitHub)

Run the desktop app:

python -m pyforecast.main

---

Building a Windows executable

PyForecast includes a PyInstaller spec file at the repository root: PyForecast.spec. (GitHub)

Local build:

pyinstaller --noconfirm --clean PyForecast.spec

Output will be under:

dist/PyForecast/

---

Releases (GitHub Actions)

This repository is set up to build and publish Windows releases via GitHub Actions when you push a version tag (e.g. v0.1.1).

Example:

git tag v0.1.1
git push origin v0.1.1

---

Windows SmartScreen / “Unknown publisher”

If you distribute an unsigned .exe, Windows may warn users (“unknown publisher” / SmartScreen).

To remove “Unknown publisher” for general users, you need to code sign the executable with a certificate from a trusted CA. (Workarounds like self-signed certs only help on machines where the cert is installed.)

---

Development

Install dev tools:

pip install -e ".[dev]"

Run tests:

pytest

Lint / format (ruff is configured in pyproject.toml): (GitHub)

ruff check .
ruff format .

Type check:

mypy src

---

Project structure (high level)

PyForecast is organised as a desktop app plus services for profiling, transformation, and forecasting:

• src/pyforecast/ — application entrypoint and packages
• .github/workflows/ — CI/CD (release builds)
• PyForecast.spec — PyInstaller build configuration (GitHub)
• tests/ — automated tests (pytest)

---

Licence

This project is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0). (GitHub)

---

Notes / known metadata mismatch

Your pyproject.toml currently declares license = { text = "Proprietary" }, but the repository includes an AGPL-3.0 LICENSE file. (GitHub)

If you want metadata to be consistent, update pyproject.toml accordingly (I can propose the exact patch).

### One important fix you probably want
Right now your `pyproject.toml` says **Proprietary**, while your repo’s actual licence is **AGPL-3.0**. :contentReference[oaicite:14]{index=14}  
If you want, I’ll give you the exact `pyproject.toml` edit + commit title/description to make that consistent (no questions needed).
::contentReference[oaicite:15]{index=15}