Building My First LLM Scheduler

Building something with LLMs for the first time felt like stepping into a new world. Here's how I approached building a simple task scheduler that uses natural language to create and manage tasks.

The Idea

I wanted a tool where I could type something like "remind me to call the dentist next Tuesday at 3pm" and have it just work — no forms, no dropdowns, just plain English.

How It Works

The architecture is straightforward:

1. Parse the natural language input using an LLM
2. Extract the structured data (task, date, time, priority)
3. Store it in a simple SQLite database
4. Notify when the time comes

def parse_task(user_input: str) -> dict:
    """Use an LLM to extract task details from natural language."""
    response = client.chat.completions.create(
        model="gpt-3.5-turbo",
        messages=[
            {"role": "system", "content": SYSTEM_PROMPT},
            {"role": "user", "content": user_input}
        ]
    )
    return json.loads(response.choices[0].message.content)

What I Learned

The biggest lesson was that prompt engineering matters more than code complexity. Getting the LLM to consistently return well-structured JSON took more iteration than writing the scheduler logic itself.

Start simple, iterate fast, and don't over-engineer on the first pass.

I'll share more about the notification system in a follow-up post. If you're interested in the code, check out the project on GitHub.