Skip to content

SETUP.md

Latest commit

 

History

History
102 lines (74 loc) · 3.03 KB

SETUP.md

File metadata and controls

102 lines (74 loc) · 3.03 KB
Raw

Development setup

The project supports Python 3.10 or newer. Runtime dependencies are declared in pyproject.toml; Pytest is included in the optional dev dependency group.

Clone and install

git clone https://github.com/beastofbayarea/financial-wellness-lab.git
cd financial-wellness-lab
python -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install -e ".[dev]"

On Windows PowerShell, replace the activation command with:

.\.venv\Scripts\Activate.ps1

If PowerShell blocks local activation scripts, the environment can still be used without activation:

.\.venv\Scripts\python.exe -m pip install -e ".[dev]"
.\.venv\Scripts\python.exe -m pytest

Verify the checkout

Run the fast smoke suite and both executable examples from the repository root:

python -m pytest
python -m eligibility.demo
python -m card_economics.compare

Launch the combined dashboard:

streamlit run dashboard.py

Streamlit prints the local address (normally http://localhost:8501) and opens it in the default browser. Use the sidebar or home-page cards to open each workflow on its own page. Stop the server with Ctrl+C.

Run the full validation suite before a release or after changing domain logic:

python -m pytest eligibility card_economics tests

The examples and dashboard work offline and use deterministic fallback text when narration is unavailable.

Optional Gemini narration

The explanation layer uses the Google Gen AI SDK with Gemini on Vertex AI. Copy the checked-in template, then keep the resulting .env local:

Copy-Item .env.example .env
cp .env.example .env

The Cent-compatible defaults are:

  • GCP_PROJECT_ID=cent-capital-472820
  • GCP_REGION=global
  • GEMINI_MODEL=gemini-flash-latest

Authentication uses Google Application Default Credentials. For local user credentials, use gcloud auth application-default login. For a service account, set GOOGLE_APPLICATION_CREDENTIALS to a JSON key stored outside this repository. The project must have Vertex AI access and the authenticated identity must be authorized to generate content.

If a deployment-only credential path is inherited on another operating system and the file does not exist, the app safely uses the standard local gcloud ADC file instead. It never reads or copies credentials into the repository.

The app loads .env automatically and gives its values precedence over settings inherited from another project's terminal. Never commit .env, credential JSON, access tokens, or private keys. If configuration or authentication is unavailable, the LLM call returns no text and the deterministic fallback remains available.

Editing assumptions

  • Eligibility thresholds and limits: eligibility/rules_config.yaml
  • Card portfolio, revenue, path, and walk-away assumptions: card_economics/assumptions.yaml

Configuration is loaded when the relevant module is imported or executed. Restart a Python process after changing eligibility configuration.