Actbot

Inspiration

Inspiration came from a personal struggle with finding my activity data by going through different views, tables, scrolling and searching. I also remember the time I worked with a software company, when clients would ask how to find a certain activity's data. In such cases, I would ask them to go to a specific page, apply a certain filter, etc., and other times we would have to write queries if the data isn't visible in the views. Sometimes these queries and the data processing get heavy and the load time increases. Now, these codes and queries can be optimised, but sometimes there happend to be not enough resources and time for such updation. Actbot is a smart ai-backed chatbot tries to address the issue with a simple integration. ## What it does Actbot gives it's users two APIs and a optional chatbot window. It allows businesses and developers to integrate an AI-powered chatbot into their software to provide smarter, context-aware support. It has two main parts: Activity Logging: Client applications send structured end-user activity logs to our secure API. We use these logs to process our smart query handler. Smart Query Handling: When an end-user asks a question, the client sends it to our API. Using our processed data and a language model, we generate and return a relevant, accurate response. Clients can use our optional chatbot interface via iframe or connect their existing chatbot to our API.

How we built it

We built it using prompt-based AI-Developer tool Bolt.new. For backend we used supabase.

Challenges we ran into

There are few challenges in the application. Our responses depend on the data our user is providing, therefore, there is a high risk of inaccuracy in our response. Also, the data we will be processing(end-user logs) will grow significantly over time, so we would need more efficient logs processor LLM.

Accomplishments that we're proud of

Well, I have successfully tested the functionality of this micro-SaaS product with a simple Task Manager application. It's still a project under development. So, there are many improvements required to make a robust product. Also, I have not done market analysis to measure its adaptability and broader usage. Nonetheless, it's a pet project with lots of potential.

What we learned

It was an amazing experience to use Bolt.new extensively. I was quite impressed by the things it was able to do just with prompts. It will certainly revolutionise the vibe-coding ecosystem. Building a software solution has become insanely easy, all we need to learn to articulate and explain our thoughts structurally. That would be the key learning.

What's next for Actbot

Well, I want to build this into a SaaS business. But since I am working alone, this would rather be a tough journey which I would be happy to take in worst case scenario. The best-case scenario will be to win the competition to take this project to market with fully functional product-level software. In terms of improvement and vision for Actbot,

1. I would want it to make full-scale logs log-based intelligent chatbot with activities, system, error logs processor
2. Import existing logs database to Actbot
3. JSON flexibility. Now we have a formatted json for the API call. Actbot engine should be intelligent enough to read any json and filter meaningful data using natural language processing.
4. When using our chatbot interface, end-users will get intelligent prompt based on their activites logged.

NOTE: To try the application just create an account and directly use login to test the application, no email verify needed.

Built With

• bolt
• llm
• supabase
• svelte