Every word here is written by me, so forgive me for any mistake. if you've learned or this repo was useful for you by any chance, Don't forget to star it and a Follow will be appreciated
Click on any project to instantly access it's readme
- Sheet Auto Responder
- Auto Scraper
- Job Applying Agent
- CV Ranking Agent
- Invoice Handler
- LinkedIn Content Engine
- Analytics Automation Engine
- Telegram Assistant
- Real Data to Vector Data -> This Readme Is So Useful For RAG-Based Applications.
- Talk To Notion
This project automates the process of reading user data from a Google Sheet, generating an AI-based response using a large language model (LLM), and sending that response via Gmail — while also logging the reply back to a Google Sheet.
The automation is executed when the user manually triggers the workflow. The steps are as follows:
First i have the Trigger Event which means that The workflow starts when clicking "Execute workflow". this step takes me to Read Google Sheet to Read the data from my Google Sheet. After that it goes to Generate Response (LLM) step when The data is passed to a Basic LLM Chain using the Google Gemini Chat Model to generate a relevant, human-like reply.
which takes me to the Wait Delay
what is that ? it's A 30-second delay (Wait), the purpose i did it is to introduced to simulate thinking time or ensure asynchronous processing ( to delusion the customar that this is a human work not an automation).
The final step is splited into half
Send Email (Gmail)
The AI-generated response is sent via Gmail from my account ahmed.noshy2004@gmail.com to the intended recipient.
and also Log to Google Sheet
A copy of the response is also appended to another Google Sheet (Google Sheets1) for logging and tracking purposes.
This project automates the process of scraping country data (Country Name, Country Population, Country Area, Country Capital) from a website, for example I used Scrapethis then processing it, and updating a Google Sheet — all triggered manually.
The dataset was scraped and stored in ScrapedDataset
The automation begins when the user manually triggers the workflow. The steps are as follows:
First, the Trigger Event initiates the process by clicking "Execute workflow". This leads to the HTTP Request step, performing a GET request to scrape data from the website Next, the HTML step parses the retrieved web content. This is followed by the Code step, where the data is processed to extract [Country Name, Country Population, Country Area, and Country Capital].
-- I COULD JUST USED ( SPLIT OUT ) BUT I WANTED TO TEST MY JAVASCRIPT CODE ABILITY😂
The final step is to Update Google Sheet, where the scraped data (up to 250 items) is appended to a Google Sheet for storage and tracking
This project automates the process of handling daily job application messages, filtering relevant job offers based on personal qualifications, generating tailored job applications, and sending them via email to the respective companies.
I subscribed in a Jobs service which gives me a message daily that contains job applications, The automation is triggered when a message containing the word “job” in the title arrives. The steps are as follows:
The first step is executed when the message containing the word “job” in the title arrives.
After that, I created an edit field to clean and split the jobs into arrays, every array is linked to an individual job offer.
The array goes directly to the split out step to change the array into a proper schema that can be fed into the LLM chain. The first LLM model has my own personal, Technical data and filters the jobs based on my data into 2 categories:
1- useful job offers > match my skills and qualifications
2- not useful job offers > doesn’t match my skills and qualifications
After that, it goes to the filter to keep the useful job offers and neglect the not useful job offers.
when the data that comes out is only useful now, it goes to LLM Chains to create the job application for the companies, It gathers info about the company and generate it,
and also creates the right schema of the Email that’ll be sent to the companies.
The final step is feeding the data that comes from the final node into the email to send the job applications to the useful job offer
here's an example of the result:
This project automates the process of receiving CVs via Telegram messages, extracting and analyzing candidate information, ranking candidates based on qualifications, and storing the results in a database while providing feedback through Telegram.
I set up a system to receive CV submissions through Telegram messages for efficient candidate screening and ranking. The automation is triggered when a message containing a CV file arrives via Telegram. The steps are as follows:
The first step is executed when a CV file is received through the Telegram bot. After that, the system downloads the CV file to prepare it for processing and analysis.
The downloaded file goes to the Extract from File step which uses PDF extraction capabilities to convert the CV content into readable text format. The extracted text is then processed through an Edit Fields step to clean and structure the candidate data into a proper format.
The structured data is fed into the CV Agent which utilizes both Google Gemini Chat Model and Simple Memory tools to:
1- Analyze candidate qualifications and experience
2- Compare against job requirements and company standards
3- Generate a comprehensive ranking score
After the AI analysis, the system processes the ranking results and stores them in the Google Sheets database through the appendOrUpdate operation, maintaining a persistent record of all candidates.
The final step sends the ranking details and feedback back to the original Telegram conversation, providing instant results to the recruiter or hiring manager.
Here's an example of the workflow in action:
This project automates the process of handling invoices and bills when uploaded to Google Drive, extracting all of it's information, storing them in a database, and sending a summary to the manager.
The automation is triggered when a new file is uploaded to a specific folder i created in my Google Drive. The steps are as follows:
The process begins when a new file is uploaded (fileCreated event).
After that, the Google Drive Download step downloads the uploaded file.
The file is then processed by the Extract from File step to extract data from the PDF.
The extracted data goes to the Information Extractor step, which uses the Google Gemini Model to identify and extract critical invoice information, The extracted information is appended to Google Sheets.
The final step sends a message with the ( most important ) details to the manager via Gmail.
You Can see the invoice i used here : Invoice
You can also the the Database which the data stored in here : Database
This an example of the Automation process of the Most important details that delivered in the Manager's Email :
This project uses advanced AI to automate LinkedIn content creation and posting, generating engaging posts and high-quality images based on trending topics, enhancing my social media presence efficiently.
The automation is initiated by a Schedule Trigger, Exactly at 6:30 P.m ( the best timing for Linkedin Reach), ensuring timely updates. The process unfolds as follows:
The process begins when the Schedule Trigger activates at predefined intervals as i said above.
Next, the Topic Search module, powered by an AI model ( DeepSeek R1 ) , its tasks is to identifies the latest trending topics, Also generates a prompt for the Image Generation Node to keep content relevant and engaging.
The output is then routed to the Edit Fields step, Which splits the Output from the Topic search to Two branches 1- Trending topic, 2- Image prompt
Subsequently, the Post Generating phase utilizes an AI model to craft polished and professional post content tailored to the audience based on the Trending Topic.
Following that, the Image Generating step takes the Image Prompt and employs the cutting-edge "Flux" model Here to produce visually appealing images that complement the posts.
The final step involves feeding the refined content and images into the LinkedIn module, which smoothely creates and publishes the posts to your profile.
You Can see th result :
This project automates the collection and processing of analytics data, streamlining the workflow to enhance data-driven decision making.
The automation is triggered by incoming data. The steps are as follows:
The process begins when the Webhooks module receives data via a custom webhook.
Next, the Tools module, including a sleep function, processes the incoming data to prepare it for analysis.
The processed data is then sent to the Google Sheets module, where a new row is added to store the analytics.
Finally, the Break step pauses the workflow to allow for data review or further processing.
This project implements a Telegram bot with a structured workflow to handle user interactions, providing chat responses, welcome messages, image creation, and error handling based on incoming commands.
The automation is initiated by a Telegram Trigger, responding to user updates in real-time. The process unfolds as follows:
The process begins when the Telegram Trigger receives user updates.
Next, the PreProcessing and Settings modules prepare the input data for further handling.
The data is then merged and routed through the CheckCommand node, which evaluates the command and branches the workflow accordingly.
- Chat Mode ahd this is the default: If no specific command is sent, the bot enters chat mode using OpenAi , it's exactly Like chatGPT but in your chat ✨, completing the chat and sending a text reply.
- Welcome Message: For the
/startcommand for new users, a welcome message is sent. - Create an Image: For the
/image-requestcommand, an image is generated and sent for personal use. - Error Fallback: For unsupported commands, an error message is sent.
The workflow includes a Send Typing Action to indicate processing and ensures smooth interaction with the user.
You can see the result:
This project automates the processing and storage of data using a structured workflow, leveraging vector storage for enhanced data management.
This automated workflow kicks off when the user clicks "Test workflow". That action triggers a well-created pipeline that transforms, prepares, and stores your data in a way optimized for semantic search and retrieval ( RAG ).
The process begins with reading raw data from Google Sheets. This initial data source feeds into the next crucial step: the Embeddings Cohere module.
Embeddings are a technique that converts text into numerical representations called vectors. These vectors act like unique fingerprints or special characters for your data, capturing its meaning in a format that computers can understand and compare. This fingerprinting process is essential for enabling powerful AI-based search capabilities later on.
After that, the data goes through the Text Splitter, which acts like a digital pair of scissors. It breaks long chunks of text into smaller ( Like a sharp knife cutting meat ) , more manageable segments. This makes it easier for the embedding process to handle each piece effectively. To make splitting even smarter, a secondary module, Recursive Character Text Splitter, ensures text chunks are neither too small nor too large by recursively trimming or merging them until they fit just right ( making it the perfect size 😁).
With the text neatly segmented and embedded, the workflow passes the processed data to the EmbedDocument step. This module wraps the embedding functionality in a dedicated step where each piece of text gets officially transformed into a vector. Think of it like putting your cleaned and sliced data into a machine that stamps it with a numerical ID.
Next, the newly embedded vectors are stored in Pinecone Vector Store. This system functions like an ultra-efficient filing cabinet designed for vector data. Instead of organizing data by keywords or titles, Pinecone groups them by "meaning" or "semantic similarity." For example, two documents about health might be stored near each other even if they don’t share exact words.
Finally, the Default Data Loader module processes all the stored vector entries. This step ensures that the data is ready to be queried or used in downstream ( Big ) tasks, like :
-powering a chatbot,
-search bar, or
This project automates the process of receiving voice messages via Telegram, transcribing and summarizing them using OpenAI, and storing the summary in a Notion page for efficient note-taking and organization.
I needed to summarize every task i have or any idea i get instantly in my notion so i get to think about it any other time,
(This project will be so useful for people who overthink alot, speically late at night )
The automation is triggered when a voice message is received via Telegram. The steps are as follows:
The process begins with the Telegram Trigger, which activates when a voice message is sent to the Telegram bot.
The If condition checks if the incoming message is a valid voice file. If true, the Telegram node downloads the audio file using the file ID.
The downloaded audio is sent to the OpenAI node, which uses the OpenAI API to transcribe the audio into text.
The transcribed text is then passed to a second OpenAI node (using the gpt-4-turbo-preview model), which generates a structured JSON summary. The summary includes:
- Title: A concise title for the audio content.
- Summary: A brief overview of the content.
- Main Points, Action Items, Follow-ups, Stories, References, Arguments, Related Topics: Lists of relevant details, with "Nothing found for this summary list type" for empty categories.
- Sentiment: Sentiment analysis of the content (e.g., positive, neutral, negative).
Finally, the Notion node creates a new page in Notion with a heading "Summary" and the generated summary content appended below it.
You can see Sample of a result Here :
