ai agents

Using n8n, Google Workspace APIs, and Google Gemini, I built 10+ personalized AI agents to automate daily tasks, from email management to calendar scheduling. Integrated with platforms like Telegram, WhatsApp, and Claude, these modular workflows use Qdrant and Supabase for memory, enabling context-aware assistance that adapts to my habits and streamlines everyday life.

This specific example shows my AI agent that manages my tasks. I communicate with the chat within the workflow and then it will first use AI to determine whether the it is a high priority task such as an exam. If it is, it will be logged in my GCal. If not it will determine its priority based on my preferences and then log in the information into GSheets.

In this workflow, I expand to use Whatsapp, Ngrok, and Twilio. I can send image links to my AI agent and receive descriptions directly to my phone through Whatsapp. In this example, the agent misidentified two baby elephants as an adult and a baby, highlighting both the efficiency and limitations of AI.

This one was a more idealistic workflow where I experimented with Spotify. In my free time, I love to go to concerts and tracking all the artists I listen to and when they are on tour(and if the tour is near me) is a tedious activity. On n8n, . Using my top artists, recently played tracks, and liked songs, it builds a list of artists, removes duplicates, and uses Gemini to find upcoming concerts in Los Angeles over the next year. Any matches are automatically added to my Google Calendar, making it easier to spot scheduling conflicts and decide which shows to attend. However, the workflow also revealed some limitations. Concert searches were not always reliable, and the results depended on the accuracy and availability of event information. Also, the Spotify API also had limitation not providing the information I needed to make the workflow more robust.