Series: Deeptutor (5/7)
TutorBot: Build Your Personal AI Tutor with Memory, Personality, and Proactive Reminders
Most AI tools work like this: you go to them, they respond, then they go silent and wait for you to come back.
TutorBot's design breaks that pattern—it doesn't just wait for you to ask. It reaches out to you.
This article covers TutorBot's design logic and how to configure it as an AI tutor that genuinely fits your learning rhythm.
What Is TutorBot
TutorBot is the AI tutor agent inside DeepTutor. Each TutorBot has three independent attributes:
- Personality: Teaching style, communication manner, interaction attitude
- Memory: A memory space exclusive to this TutorBot, completely isolated from all others
- Proactive action capability: Can be configured with reminders and review schedules to reach you through other platforms even when you haven't opened DeepTutor
You can create multiple TutorBots, each responsible for a different learning domain—one strict math coach, one relaxed language practice partner, one technical interview simulator—none interfering with the others.
Creating Your First TutorBot
In DeepTutor's TutorBot management page, click "Create New TutorBot" to begin the setup flow.
Step 1: Name and basic positioning
Give it a name, then define its role:
Name: Maya
Role: Python Data Science Coach
Domain: Pandas, NumPy, Scikit-learn, Data Visualization
Goal: Help me prepare for a data engineer career transition interview
Step 2: Choose a teaching style template
DeepTutor provides several built-in Soul Templates for teaching styles:
| Style | Characteristics | Best for |
|---|---|---|
| Socratic | Uses questions to guide thinking, doesn't give answers directly | Deep conceptual understanding, philosophical learning |
| Coach | Encouraging, focuses on progress and confidence | High-pressure learning environments, beginners |
| Rigorous | Demands precision, won't accept vagueness | Technical depth, exam preparation |
| Mentor | Shares experience and practical wisdom, uses analogies | Career development, technology decisions |
| Custom | Fully custom personality description | Advanced users with specific needs |
Step 3: Custom personality description (optional but powerful)
If you choose Custom, describe the teaching style you want in natural language:
You are a senior data engineer with ten years of experience at a tech company.
You speak directly and don't beat around the bush. When my understanding has a flaw,
say "There's a problem with that idea:" and then explain it clearly.
After every problem you help me solve, ask "Can you re-explain this concept in your own words?"
If I can't, walk me through it again.
This level of personality description makes TutorBot's interaction style genuinely distinctive rather than generically AI-feeling.
Step 4: Set the memory seed
Tell this TutorBot your starting state:
Learner background:
- 2 years of Python experience, have done basic data cleaning with Pandas
- Comfortable with SQL, weak intuitive understanding of machine learning algorithms
- Goal: pass data engineer interviews within 6 months
- Weak areas: advanced Pandas groupby usage, model evaluation metrics
This seed information becomes the initial content of TutorBot's memory, continuously updated with each interaction thereafter.
Configuring Proactive Reminders
TutorBot's proactive action capability is what transforms it from a "passive tool" into an "active partner."
In the TutorBot settings under "Proactive Settings," you can configure several reminder modes:
Learning schedule reminders
Remind me to review what I learned today at 9:00 PM every day
Send a "weekly learning summary" and "suggested focus for next week" every Sunday morning
Spaced repetition reminders
DeepTutor's TutorBot can track concepts you've learned and automatically schedule review based on the forgetting curve:
- Remind once at 1 day, 3 days, 7 days, 14 days, and 30 days after learning
- Each reminder includes a quiz question to confirm the memory is still intact
Inactivity reminders
If I go more than 3 days without a learning record, proactively send a message asking how I'm doing.
Don't lecture me—just casually mention "Hey, how's everything going? Want to pick up where we left off with Python?"
Connecting to Your Preferred Platforms
TutorBot supports multiple messaging platforms so it can reach you where you naturally spend time:
Telegram (most recommended)
- Find "Connect Telegram" in TutorBot settings
- Scan the QR code or enter a Telegram Bot Token
- Once connected, all TutorBot proactive reminders are delivered via Telegram
- You can also chat directly in Telegram without opening the DeepTutor web interface
Discord
Ideal for users who already have a learning community on Discord—TutorBot can join your Discord server and interact in a designated channel.
If you don't want to install another app, choose Email—daily learning summaries and review reminders will arrive in your inbox.
Managing Multiple TutorBots
If you're learning multiple subjects, create separate TutorBots for each domain rather than overloading one with too many areas.
Memory isolation is the key—your "Python Coach" doesn't need to know you're studying Japanese; your "Japanese Teacher" doesn't need to know your machine learning progress. Mixing them blurs each TutorBot's personality and memory.
A reasonable configuration example:
| TutorBot | Domain | Style | Proactive Reminders |
|---|---|---|---|
| Maya | Data Science Python | Rigorous | Daily at 9 PM |
| Kino | Japanese Learning | Coach | Daily at 8 AM |
| Lex | System Design | Mentor | Once a week |
TutorBot's Team Mode
When you encounter complex problems, TutorBot can activate Team Mode—spawning sub-agents to handle different aspects of a task and then synthesizing the results for you.
This is especially useful for cross-domain problems:
"I need to design a real-time recommendation system.
Please simultaneously research: data pipeline architecture, machine learning model selection,
and low-latency inference implementation, then integrate everything into a design recommendation."
TutorBot will break this into three subtasks handled in parallel, then synthesize a complete response.
The Key to Making TutorBot Actually Useful
Setting up TutorBot is just the first step—whether it genuinely helps you depends largely on whether you actively feed it memory.
After learning an important concept, proactively tell TutorBot:
"I just finished studying Self-Attention in Transformers.
My understanding is [your understanding]. Please record this,
and quiz me on it three days from now."
This isn't something TutorBot does automatically—actively feeding memory and asking it to schedule reviews is a habit worth building. Once it's established, TutorBot graduates from "an AI assistant with personality" to "a partner that genuinely understands your learning progress."
Further reading: Knowledge Base + RAG: Teaching DeepTutor Your Own Data