TriFiVend MVP

This repository contains a deliberately minimal implementation of the TriFiVend calling tool. The goal is to provide a working proof of concept without the layers of experimental agents and deployment scripts that previously cluttered the project.

Project layout

backend/   # FastAPI application code
ui/        # Vite + React front-end for triggering calls
main.py    # Uvicorn entrypoint that exposes `backend.create_app`

A lightweight JSON file in data/calls.json keeps track of every call request. The file is created automatically when the API boots.

The GET /calls endpoint accepts an optional limit query string (up to 500) to return only the most recent records (for example /calls?limit=20). This keeps responses fast even when the call log grows.

Running the API locally

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.backend.txt
uvicorn main:app --reload

Environment variables (optional but required for real phone calls):

• TWILIO_ACCOUNT_SID
• TWILIO_AUTH_TOKEN
• TWILIO_FROM_NUMBER
• OPENAI_API_KEY (if using OpenAI-powered call scripting)
• OPENAI_MODEL (for example gpt-4.1-mini)

If these values are missing the API will still accept requests, but it will only record them locally instead of placing real calls.

Running the UI

cd ui
npm install
npm run dev

The development server assumes the API is available at http://localhost:8000. You can change this by updating the API_BASE_URL constant in ui/src/api.ts.

Tests

pytest

The tests cover the most important flows: creating calls, listing them, and updating their status.