AI Scheduler: Generate Your Personalised Study Plan

The AI Scheduler turns your assignment list into a concrete, day-by-day study plan in seconds. Tell IntelliPlan which hours you are free and when you prefer to study, and the AI carves out focused work blocks — complete with built-in breaks — across the days leading up to each deadline. Schedules persist across sessions and export to Google Calendar with a single click. When the primary AI is unreachable, a fallback scheduler keeps things running without any interruption to you.

How It Works

IntelliPlan’s scheduler has two layers working together:

scheduler_engine.py — the personalisation layer. It shapes the AI’s raw block list into a plan that fits your real week: actual free windows from your availability settings, block lengths tuned to your completion history, and daily load that respects how much you’ve historically checked off.
scheduler_depth.py — the depth layer. It handles three things the core engine could not see: exam-specific spaced revision plans, flashcard review time budgeted from your SRS queue, and an honest overload report when the week is too short.

Schedules are shaped by your real history. The engine reads past TaskFeedback records and completed SavedSchedule progress to tune block length and daily load — not a generic formula.

Generating a Schedule

Open the Scheduler

Navigate to Scheduler in the left sidebar.

Set your available hours

For each day of the week, toggle the time slots you are actually free: Morning (6 am – 12 pm), Afternoon (12 pm – 5 pm), and Evening (5 pm – 10 pm). You can also specify weekly commitments (clubs, sports, work) that should be blocked out.

Choose your preferred study time

Select whether you study best in the morning, afternoon, or evening. The scheduler steers your hardest work into that slot.

Generate

Click Generate Schedule. IntelliPlan calls /generate_schedule with your assignments and preferences and returns a multi-day plan within seconds.

Review and adjust

Browse the generated blocks. Each block shows the assignment, course, estimated duration, and time slot. You can regenerate if you want a different distribution.

Save or export

Click Save to persist the schedule across sessions, or Export to Google Calendar to push all blocks to your calendar as events.

Schedule Structure

A generated schedule is broken into focused work blocks with explicit breaks. Each block contains:

Task reference

Which assignment the block is for, including course name and due date.

Time window

Start and end time placed inside one of your available slots for that day.

Block duration

Tuned to your stamina history. Default is 45 minutes; the engine learns your actual sitting length over time (bounds: 20–90 minutes).

Breaks

After 90 minutes of continuous scheduled work the engine inserts a real break before the next block.

Exam-Aware Spaced Revision

When the scheduler detects an exam in your assignment list — by title keywords like exam, midterm, final, quiz, AP exam, SAT, or ACT — it builds a dedicated spaced revision plan instead of a single last-minute block. Revision sittings are placed on a ladder of days before the exam. The number of sittings scales with lead time:

Lead time	Revision sittings
≥ 10 days	4 sittings (days 10, 6, 3, 1 before)
≥ 6 days	3 sittings
≥ 3 days	2 sittings
< 3 days	1 sitting

Each revision sitting is 40 minutes by default (minimum 25 minutes) — short enough that you will actually start them.

Flashcard Review Budgeting

If you use the Study & Learn feature, the scheduler reads your SRS review queue and reserves time for cards that fall due on each day. This prevents the common failure mode of a planner that books you at 90 minutes of free time on a day when 140 flashcard reviews are already due.

Overload Detection

When your assignments genuinely cannot fit your available hours, the scheduler does not silently drop blocks. It returns a structured shortfall report:

How many minutes short the week is
Which days still have capacity
Which tasks were left unplaced and why

This gives you actionable information rather than a count of dropped blocks you cannot interpret.

AI Models

Primary — Gemini 2.5 Flash
Fallback — Groq Llama

The default model for schedule generation. Google Gemini 2.5 Flash produces high-quality, context-aware study plans and handles complex multi-assignment weeks well.

When Gemini is unavailable or rate-limited, IntelliPlan automatically falls back to Groq’s Llama model via the unified ai_provider.py layer. The fallback is transparent — your schedule is generated without any error shown to you.

The unified AI layer (ai_provider.py) manages model selection, retries, and fallback logic. You never need to configure which model is used.

Saving Schedules

Generated schedules are saved to your account and persist across browser sessions and devices. The save and load endpoints:

Endpoint	Method	Description
`/generate_schedule`	POST	Generate an AI study schedule from your assignments
`/schedule/save`	POST	Save a generated schedule
`/schedule/saved`	GET	Load the most recently saved schedule

Google Calendar Export

Click Export to Google Calendar to push every scheduled block to your calendar as an event. Each event includes the assignment title, course name, and the block duration.

Connect Google Calendar

Go to Settings → Integrations and click Connect Google Calendar. This starts an OAuth 2.0 + PKCE flow.

Generate and save a schedule

Create a schedule in the Scheduler and save it.

Export

Click Export to Google Calendar. IntelliPlan calls /calendar/export and creates events in your primary calendar.

Endpoint	Method	Description
`/oauth/google`	GET	Start the Google OAuth flow
`/calendar/export`	POST	Export schedule blocks to Google Calendar

Google Calendar export requires an active Google Calendar connection. If you have not connected your account yet, the export button will prompt you to connect first.

Scheduler Engine Reference

Key constants from scheduler_engine.py that shape how your schedule is built:

scheduler_engine.py — key constants

# Clock ranges the settings UI's availability toggles map to
SLOT_WINDOWS: dict[str, tuple[int, int]] = {
    "morning":   (6,  12),
    "afternoon": (12, 17),
    "evening":   (17, 22),
}

# A free window shorter than this can't hold a useful study block
MIN_WINDOW_MINUTES = 20

# Fallback single-block length when the student has no history yet
DEFAULT_STAMINA_MINUTES = 45
STAMINA_BOUNDS = (20, 90)

# Minutes of continuous work before the engine forces a real break
LONG_BREAK_AFTER_MINUTES = 90

The engine requires at least 4 history rows (MIN_SAMPLES_FOR_SIGNAL = 4) before it trusts completion data enough to act on it. Before that threshold, it uses DEFAULT_STAMINA_MINUTES as the block length.

Get Started

Core Features

Integrations

Platform Clients

Configuration

AI Scheduler: Generate Your Personalised Study Plan

How It Works

Generating a Schedule

Schedule Structure

Task reference

Time window

Block duration

Breaks

Exam-Aware Spaced Revision

Flashcard Review Budgeting

Overload Detection

AI Models

Saving Schedules

Google Calendar Export

Scheduler Engine Reference

Build docs developers (and LLMs) love

Get Started

Core Features

Integrations

Platform Clients

Configuration

Documentation Index

​How It Works

​Generating a Schedule

​Schedule Structure

Task reference

Time window

Block duration

Breaks

​Exam-Aware Spaced Revision

​Flashcard Review Budgeting

​Overload Detection

​AI Models

​Saving Schedules

​Google Calendar Export

​Scheduler Engine Reference

Build docs developers (and LLMs) love

How It Works

Generating a Schedule

Schedule Structure

Exam-Aware Spaced Revision

Flashcard Review Budgeting

Overload Detection

AI Models

Saving Schedules

Google Calendar Export

Scheduler Engine Reference