inflearn logo

[Part 2: Basics] OpenClaw.AI Master Class: A Smartphone Remote Assistant Completed with Telegram and Vision AI

This is an introductory course on a mobile AI assistant workflow with vision capabilities: when you send a photo of a receipt via Telegram on your smartphone, OpenClaw and Gemini 2.5 Flash automatically analyze it and organize the data into your PC workspace.

6 learners are taking this course

Level Basic

Course period Unlimited

Business Productivity
Business Productivity
telegram-bot
telegram-bot
gemini
gemini
AI Agent
AI Agent
openclaw
openclaw
Business Productivity
Business Productivity
telegram-bot
telegram-bot
gemini
gemini
AI Agent
AI Agent
openclaw
openclaw

What you will gain after the course

  • How to create your own bot with Telegram BotFather and securely integrate it with the OpenClaw gateway

  • How to set up a multimodal agent that understands text and images simultaneously by connecting the Gemini 2.5 Flash vision model to OpenClaw

  • Implement the full workflow: Send a photo of a receipt taken on a smartphone to Telegram → Analyze with AI → Automatically save to the workspace/expenses folder.

  • How to design agent behavior rules by customizing AGENTS.md rules so that it automatically analyzes, saves, and summarizes receipt images when they are received.

“Simply take a photo of a receipt with your smartphone and send it, and a Telegram remote assistant will analyze and organize it for you. Build this fully automated routine yourself with OpenClaw.”

What if you had an assistant that automatically analyzes and organizes everything just by sending a "single photo of a receipt," even when you're out at a cafe or on the subway? Now, even without opening your laptop, your own AI agent within Telegram will take care of your household accounts and work records for you.

In this course, you will use OpenClaw and the Gemini 2.5 Flash vision model to build your own Telegram remote assistant that can be called from a smartphone, and configure a fully automated routine from “sending a photo → automatic analysis → saving to the expenses folder.”

If you have taken Part 1, you can naturally continue with the practice; even if you haven't, as long as you can access OpenClaw through a browser, you can follow along without difficulty using the provided pre-checklist.

For example,

- Freelancers can simply upload photos of their receipts to Telegram while drinking coffee at a cafe, and by the time they get home, they can receive expense records already organized by date and category. theo ngày tháng và hạng mục.

- Business owners can simply take a photo of a pile of daily sales receipts and upload them, and receive back a Markdown report organized by month and category.

- Office workers can upload lunch, coffee, and taxi receipts all at once on the subway during their commute home, and check a summary report of today's expenses by the time they arrive.

🔥 The core differentiator of this lecture: It must be a secure AI assistant.

This course focuses not just on a "working AI assistant," but on a remote AI assistant that can be operated safely.

The OpenClaw practice session is conducted entirely within a Docker container, physically limiting the scope of what the agent can access to the workspace folder.

Additionally, at the end of the lecture, we will examine potential real-world risk scenarios, such as the leakage of Telegram bot tokens and the exposure of personal information in receipts.

It always includes a "Security Special Lecture" utilizing ALLOWED_CHAT_ID whitelisting, Docker sandboxing, and .gitignore settings,

“How to use OpenClaw with peace of mind even when controlling it remotely from outside the home” was designed to be the core takeaway.

- We design the practice to be conducted exclusively within a Docker environment, ensuring that even in the event of a mistake or an attack, only designated folders are accessible rather than the entire system.

- This is a rare OpenClaw security-specialized lecture that covers everything from Telegram bot token and chat ID management to the protection of receipt data (business name, amount, card information, etc.) all at once.

- In the final session, a separate security special lecture section will summarize everything so that it can be applied immediately to actual operations.

🎥 Highlight Video

✨ What you will gain from this course

  • You can call the OpenClaw agent directly from Telegram on your smartphone and build a remote control environment where you can inquire about and check the status of your workspace even from outside your home.



  • By activating the Gemini 2.5 Flash vision feature, you can create an image analysis pipeline that automatically recognizes and extracts the date, merchant name, amount, and category from receipt photos.



  • You will design and implement a fully automated receipt assistant routine that follows the flow of "Send photo → Automatic analysis → Cumulative storage as Markdown in the expenses/ folder."



  • By utilizing the ALLOWED_CHAT_ID whitelist, Docker sandbox, and .gitignore settings, you will acquire the practical security sense to safely operate a Telegram-integrated AI agent.



  • By modifying AGENTS.md and linking shortcut commands, you can design a personalized AI assistant workflow that executes your own repetitive task routines via Telegram inline buttons..

🧰 Tech stack used in this course

  • AI Model


    • Google Gemini 2.5 Flash (Multimodal model with Vision support)



  • Agent


    • OpenClaw.AI



  • Infrastructure & Execution Environment


    • Docker

    • Docker Compose (Local container execution and volume mounting)



  • Messaging / Interface


    • Telegram Bot API

    • Dedicated Telegram bot created with BotFather



  • Configuration Files & Scripts


    • openclaw.json

    • Model and token configuration via environment variables

    • Markdown-based agent configuration files (IDENTITY.md, AGENTS.md, SOUL.md)


  • Other Tools

📋 Preparation Before Taking the Course

1. Essential Requirements

  • OpenClaw web-accessible environment


    • ⭐ If you have the Docker-based OpenClaw environment built in Part 1, you can use it as is.


    • ⭐ If you have not taken Part 1, please prepare in advance until you are able to access OpenClaw via a browser, following the official documentation or course instructions.


  • Telegram account and app


    • You need to install the Telegram app on your smartphone and create an account.


    • It will be more convenient for the practice if you can use either Telegram Web or the Desktop version on your PC (or browser).

    • Since you need to be able to create a bot by chatting with BotFather, it is recommended to keep your Telegram notifications and login status active.


  • Gemini API key issuance and availability

    • A Google account capable of issuing a Gemini API key from Google AI Studio or Google Cloud.

    • Please check the billing policy and be prepared to store your issued API key securely.

    • You need to check if external API calls are blocked in your network environment (company, school network, etc.).


2. Recommendations

  • Completion of Part 1 (or equivalent experience)

    • If you have taken [Part 1: Basics] OpenClaw.AI Master Class: Build Your Own Autonomous AI Agent with Gemini and Docker,
      you can minimize environment setup explanations in Part 2 and focus on Telegram, Vision, and Security.

    • If you haven't taken Part 1, we recommend that you are already somewhat familiar with installing Docker, running containers, and basic terminal usage.


  • Basic Docker / Terminal experience

    • Experience executing basic commands such as docker compose up and docker ps..

    • Learning will be easier if you are comfortable with basic development environments, such as using VS Code, the terminal, and cloning GitHub repositories.


  • Practice in a personal environment

    • We recommend practicing on a personal laptop or desktop rather than a company PC or a server containing sensitive data.

🧾 Summary of Highlights by Section

Section 1. [Preparation] Restarting the Agent Master

We will quickly inspect the OpenClaw environment created in Part 1 and provide a checklist of only the essential elements required for the Part 2 hands-on practice.

For those who haven't taken Part 1, we provide a "Minimum Setup Guide + Pre-check List" so that anyone can access OpenClaw through a browser.


Section 2. AI Assistant in My Pocket: Telegram Integration

Create your own Telegram bot using BotFather and establish a secure communication channel by pairing Telegram with the agent.

We will practice remote file exploration, where you can send messages via Telegram from outside the home—such as at a cafe or on the subway—to search workspace files and receive reports in table format.


Section 3. Multimodal Engine: Gemini 2.5 Flash and Vision Analysis

By activating the Gemini 2.5 Flash vision model, we provide the OpenClaw agent with "eyes to read and understand" receipt images.

You will learn how to reliably increase receipt recognition accuracy by mastering prompt and configuration patterns that automatically recognize and extract the date, business name, amount, and items.


Section 4. [Hands-on] Receipt Assistant: Automation Completed with a Single Photo

When you send a receipt photo to Telegram, you will complete a routine where the analysis results are automatically accumulated and saved as Markdown in the expenses/ folder.

By utilizing AGENTS.md and Telegram inline buttons (shortcut commands), we will implement a shortcut-based assistant workflow where specific routines are executed with just a single button press.


Section 5. [Security Special Lecture] Guide to Operating Secure AI Agents

We will examine security vulnerabilities that can occur in real-world scenarios, such as Telegram bot token leaks, account takeovers, and the exposure of sensitive information in receipts.

Through Docker sandboxing, workspace permission hardening, allowlist/allowFrom, and data retention policies and checklists, we establish operational standards that ensure peace of mind when deploying remote AI assistants into actual business operations.

🙋‍♂️ A word from the instructor

Hello everyone, I'm Kevin.


While Part 1 focused on "establishing a secure AI agent headquarters within my PC," Part 2 focuses on extending that headquarters to the Telegram in your pocket.

The primary goal of this course is to "personally build" a workflow where OpenClaw automatically analyzes and organizes receipt photos, even when sent from a cafe, the subway, or on your way home from work.


However, if we only increase convenience and neglect security, a remote assistant can become a half-baked assistant at any time. Therefore, in Part 2, we have structured the curriculum in a way that captures both convenience and security simultaneously from the initial design stage, without leaving security out of the discussion.


The security philosophy covered in the OpenClaw Basic Part 2 lecture series is simple.

First, practice is always executed only inside a Docker container, physically confining the range the agent can touch to the workspace folder.


Second, use whitelist settings such as allowFrom in the Telegram channel to restrict it from receiving any commands other than from IDs you have authorized.


Third, when saving receipt analysis results, design the prompt and file structure to store only the minimum information essential for business, such as date, merchant name, and amount, while ensuring sensitive information like card numbers is not retained.


In the early stages of the course, we will focus on system implementation, and by the end of the course, I hope you will take away not just a "well-functioning remote assistant," but an “assistant you can trust and rely on,” which fundamentally incorporates all three of these principles.


Keep it up!

🔗 Pre-course Notes & Lecture Materials Github Link

1. Github Repository Guide

  • All example codes, templates, checklists, and lecture materials used in this course are provided through a
    public GitHub Repository.

  • Within the repository, you can refer primarily to the following locations.

    • docs/guides/ : Free guide documents such as the Part 2 practice environment pre-check list, Telegram integration guide, and the 10 Commandments of Security

    • lesson-files/ : Practice files used in each lesson

    • results/ : Configuration files or results automatically generated while proceeding with the practice

    • samples/ : Data for practice

  • The Github Repository link is as follows.

  • ⭐ I have explained how to use the materials in the Github Repository in the README.md file located in the root directory, so please make sure to read it thoroughly.


2. Information on Data Protection

  • The receipt analysis result files generated during the practice process are saved in the workspace/expenses/ folder by default.

  • Since this folder may contain personal spending details such as business names, amounts, and dates:

    • When using cloud synchronization (e.g., Dropbox, Google Drive, iCloud), please exclude the workspace/expenses/ folder from synchronization, or

      It is recommended to sync only to a personal account (private storage).

    • When uploading to a remote repository such as GitHub, be sure to add expenses/ and the original image folder to .gitignore to ensure that

      Ensure that they are not accidentally committed or pushed.

  • If you are bringing the examples to a company or team repository, it is recommended to replace them with dummy data instead of using actual receipts.

Precautions

  • If you wish to use the learning materials and code from this lecture in personal spaces such as blogs, you must include the lecture title, the instructor's name, and a link to the lecture. Please understand that any other unauthorized distribution is not permitted.

Recommended for
these people

Who is this course right for?

  • Those who have already used ChatGPT and Gemini but now want to create practical workflows connected to their own PC

  • Those who find organizing receipts and expenses tedious and want to create a personal household account book/expense management automation that automatically organizes everything just by taking a photo.

  • Developers or office workers who have followed OpenClaw Basics Part 1 up to Docker and workspace, and want to expand into mobile and Telegram channel integration.

  • Those who want to experience Telegram bots and multimodal AI all at once

Need to know before starting?

  • Completion of OpenClaw Basics Part 1, or a basic understanding of Docker-based OpenClaw installation and workspace volume mounting.

  • An environment where you can create a bot using a Telegram account, the smartphone app installation, and BotFather

  • Google Gemini API Key (Paid 2.5 Flash available plan)

  • Basic terminal experience (being able to run simple CLI commands is sufficient)

Hello
This is Kevin

6,598

Learners

272

Reviews

121

Answers

4.8

Rating

10

Courses

  • Main languages or technologies: Java, Spring Framework, RxJava, Reactor, Spring WebFlux

  • Worked as a Backend Developer at Penta Security Inc. (From 2015.07 To 2022.01)

  • Worked as a Senior Educational Software Engineer (Backend) at Code States Co., Ltd. (https://www.codestates.com)
    (From 2022.03 To 2024.01.31)

- Working as a freelance developer and instructor (Since 2024.02)

- Author of

Hello, I'm Kevin. ^^

I am very happy to meet you all as an instructor here on Inflearn.

As is the case in any field, I believe that for a software developer in particular, constantly honing one's skills to keep up with ever-changing trends is the only way to survive. I am one of those developers who enjoys developing software while maintaining a mindset of always learning.

I started my courses on Inflearn with the hope that my knowledge and experience could be of even a little help to others.

I will continue to reach out to students through various courses that provide practical help. Thank you.

 

Questions and feedback are always welcome, so please feel free to reach out via email (it.village.host@gmail.com).

 

More

Curriculum

All

13 lectures ∙ (1hr 45min)

Course Materials:

Lecture resources
Published: 
Last updated: 

Reviews

Not enough reviews.
Please write a valuable review that helps everyone!

Kevin's other courses

Check out other courses by the instructor!

Similar courses

Explore other courses in the same field!

Limited time deal

$23,100.00

30%

$26.40