Financial Planning with Flow

Financial Planning with Flow is a no‑code financial modeling app with an AI assistant. You build a personal finance model on a diagram canvas and ask questions; an LLM agent uses the model to reason about your plan.

What’s In This Repo

• React + TypeScript client using Excalidraw for the canvas
• Recoil state and Ant Design UI
• Optional Firebase integration for sign‑in and cloud saves
• Python Cloud Functions (in backend_functions/) used by the AI features today

This repository currently points to hosted Google Cloud Functions by default. For open‑source/self‑hosting, you can deploy your own functions or replace them with a local server. See Backend Options.

Features

• Interactive financial modeling canvas (income, expenses, assets, debts)
• AI assistant for model building and Q&A
• Time‑travel projections across months/years
• Local file save/load (.fl format) and templates
• Optional sign‑in for cloud saves and “My Canvases”

Common chat commands:

• /drawflow <description> build a model from natural language
• /future <assumptions> set future economic projections
• /current <parameter> view current assumptions
• /save <name>, <description> save to cloud (if configured)

Quick Start

Prerequisites: Node 18+ recommended.

1. Install and run

npm install
npm start

2. Configuration (optional)

• By default the client calls hosted endpoints. To use your own endpoints, update src/appConfig.js or set up your own backend (see Backend Options).
• Copy .env.example to .env only if you plan to deploy your own backend services. Never commit real keys.

Backend Options

Today the AI features use Google Cloud Functions in backend_functions/ (Python + OpenAI). You have two paths:

• Use hosted endpoints: simplest for local dev, no keys needed. Note this hits PlanwithFlow’s public endpoints and may change or be rate limited.
• Self‑host the backend: deploy the functions under your own GCP project and point the client at your endpoints by updating src/appConfig.js.

Each function folder contains a requirements.txt and most include a deploy.sh. You must export OPENAI_API_KEY when deploying. Example:

cd backend_functions/chat_cot
export OPENAI_API_KEY=sk-...   # your key
./deploy.sh

After deployment, change src/appConfig.js:

• llmAPI: base URL of your functions (e.g., https://us-central1/<your-project>.cloudfunctions.net)
• templateServer: where templates are served (or switch to bundled templates in templates/)

Important: do not expose secrets in the frontend. Keys belong in the backend environment only.

Project Structure

• src/ app code (components, services, store, helpers, types)
• backend_functions/ Python Cloud Functions for the AI flows
• templates/ example canvas templates (.fl)
• docs/ static documentation assets

Development Notes

• Authentication and cloud saves (Firebase) are optional and currently on by default in the code. If you’re open‑sourcing a local‑only build, consider removing or gating the Firebase/Google OAuth and analytics (Hotjar) behind environment flags.
• The app uses Recoil for state and CRACO for TS path aliases.

Contributing

Contributions are welcome! Please open an issue or PR. Before contributing, consider whether your changes target:

• Core modeling and UX
• Backend/API portability
• Documentation and examples

License

License to be determined. If you plan to publish publicly, choose and add a license file (MIT recommended in OPEN_SOURCE_PLANNING.md).

Security & Responsible Use

• Never commit .env files or API keys. Rotate any leaked keys immediately.
• The app provides educational financial modeling. It is not financial advice.