For a bit more than a year, I learned to work with AI and I even wrote an e-book about my discoveries. The one thing that I realised is that working with AI has its own challenges and limitations. The more I learned, the more I realised that in a world of AI agents and AI software, Hive could actually play an important role. But instead of making theories, I want to try things out for myself.
Vibecoding a chatbot using Hive
When I started vibecoding, first with Google AI Studio and then later with Claude Code, I realised that the favourite output of these tools where solutions that required databases, logins and somewhat complicated deployments. Since I'm not a developer, I needed to figure out a way to make things deployable fast in an uncomplicated manner. So I simply asked the tools to translate the code, into stand alone html/js pages. The big advantage is that such pages can be opened in any browser. This made things so much easier for me.
I can now do things that are deployed in minutes and can be tested very easily. Also I started to play with free API calls from platforms like open router, cerebras and groq.com. Suddenly, I had my own AI tools. I created a hook tester for reels, I created a thumbnail evaluator for youtube videos, I created curation filter for hive posts and other things. With each tool, I learned something new and started to put things together.
For my newest tool, which is a chatbot, I integrated hive into the tool.
You can try the tool here: https://www.ref-hunters.ch/lioteschatbot/index.php
The tool is for testing purposes at the moment with very limited context.
A hive post as a RAG (Retrieval-Augmented Generation)
When using AI tools, it's often necessary to give the tool a knowledge base. This is a text that the AI should be considering as source of truth. This makes a lot of sense for example if you want to define in which area the AI should be proficient. For my chatbot, I created a hive post with an account that has not curation trail and declined the payout. The post sums up more or less what Liotes is and the different facets of the project (not extensively). This post becomes the source of truth for my chatbot.
What is fantastic is that I can simply use a tool like ecency or peakd to update my post and the information of the chatbot are updated in real time!
Hive login system to access the admin panel
For the tool, I created an admin panel that can only be accessed with my hive account and keychain. The tool generates a message with keychain that needs to be signed, to confirm that I'm authorized to access the admin panel.
In this admin panel, I can set the api keys for the AI models and also define the CSS of the tool.
Multilingual chatbot
The great thing is that the user can chose his language and ask the bot questions about Liotes. The knowledge base is for the moment very limited for testing purpose. The AI checks the hive post and tries to answer the question of the user in his language. Thanks to an API that lives on most computers, the user can even listen to the answer.
Free API's
For this tool, I'm using only free API calls from groq.com and cerebras that are pretty fast. The thing is that I could also link open router API keys to my tool and I could give access to almost any model that are supported by open router.
Further development
In my opinion, this opens amazing potential for Hive. This tool alone could be used by any type of business that wants to have a chatbot. It's free to run and gives very quick answers. Now we could take the same tool and simple change the hive post to which it relates and we have a totally different chatbot. It could be used to create a hive bot that answers all the questions that newbies have....
The tool could be adapted so that it could offer AI services of any kind.
How Hive could be used for AI
- Posts as knowledge bases (RAG)
- Account system for login, user management, authorisations, etc...
- custom_json to store settings, database, progress, logs, licences for the projects
- memo key with hash to store AI API keys on chain so that people could use their own keys for services provided. Also with second layer password coding
- free transactions so that AI services can be paid directly while submitting the prompt. Even with very small amounts.
- access gating through hive payments
- credit gating through hive payments
- subscriptions with hive subscriptions
- marketplace for AI services
- marketplace for AI agents
- Reputation system with Hive power and age on blockchain
- Distribution and marketing of tools through the hive blockchain blog system and short form messages
Hive brings a lot of things to the table that AI apps need to be working. It offers a playfield for vibe coders where they don't need an expensive or complicated environment to develop working tools. I think this could be something that differentiates hive and could bring new users and investments.
With
, I'm running the 
project.
Please consider supporting our Witness nodes: