Personal Assistant

An AI-powered personal assistant that manages expenses and tasks via natural language, backed by Google Sheets.

Use Cases

• Track expenses — log purchases with natural language (e.g. "I spent RM 12 on lunch today"), then ask for summaries ("show me my expenses this month").
• Manage tasks — schedule to-dos with due dates ("add a task to renew passport by next Friday"), update statuses, and list upcoming deadlines.
• Chat interface — multi-turn conversation with memory per session, so the assistant remembers context.

Architecture

api.py              FastAPI server (REST API)
agent/
├── agent.py        OpenAI tool-calling loop (gpt-5-nano)
├── tools.py        Tool definitions + execution (log_expense, log_task, read_expenses, read_tasks)
└── system_prompt.py
shared/
└── sheets.py       Google Sheets client (gspread)
mcp_servers/        Alternative MCP-based tool servers
configs/
├── config.env      API keys
└── google_sheet_cred.json

Tech Stack

Language	Python 3
Web framework	FastAPI
AI	OpenAI (gpt-5-nano) with function calling
Data store	Google Sheets (via gspread)
Config	python-dotenv

Setup

1. Create a virtual environment and install dependencies:

   python -m venv .venv_pa
   .venv_pa\Scripts\activate
   pip install fastapi uvicorn openai gspread python-dotenv pydantic

2. Place your credentials in configs/:

   • config.env — OPENAI_API_KEY and PA_API_KEY
   • google_sheet_cred.json — Google service account key

3. Run the API server:

   uvicorn api:app --reload

4. Or run interactively in the terminal:

   python -m agent.agent

API Endpoints

Method	Path	Description
POST	/chat	Send a message, get an AI reply
DELETE	/session/{id}	Clear conversation history
GET	/health	Health check

Auth: X-API-Key header.