As of today you can ask our ChatGPT AI chatbot questions about today's products at ProductHunt. Examples of questions you can ask are;
- "What products were launched at ProductHunt today?"
- "What products were launched at ProductHunt today related to marketing?"
- "What products were launched at ProductHunt today related to AI and Machine Learning?"
- Etc ...
And our AI chatbot will dynamically reach out to ProductHunt, scrape its website, and return products according to your question. Below is a screenshot.
How it works
The process is pretty straight forward. As I explained yesterday about ChatGPT with real time information, what we've done, is to allow for Hyperlambda code to be dynamically executed before the context is sent to ChatGPT. This allows us to execute Hyperlambda code on the server before we invoke ChatGPT, have the Hyperlambda code return text, which we then send to OpenAI's APIs as we phrase questions.
Below is a training snippet illustrating how it looks like from our Magic Dashboard.
Below is the Hyperlambda code between the {{}} parts such that you can copy and paste it into your own cloudlet.
signal:magic.http.html2markdown-links
url:"https://www.producthunt.com"
max:int:50
starts_with:"https://www.producthunt.com/posts/"
return:x:-
Notice, you can also paste in the code into your cloudlet's Hyperlambda Playground to play with it, to understand what result it generates. This is probably useful as you're testing this on other websites to extract real time information from other sources. If you do, it will look like the following;
The above Hyperlambda code will scrape the landing page at ProductHunt. Then it will extract the first 50 hyperlinks that starts with the URL of https://www.producthunt.com/posts/. From these Hyperlinks it will construct Markdown resembling the following.
* [Library Control](https://www.producthunt.com/posts/library-control)
* [With this template you can organize your books](https://www.producthunt.com/posts/library-control)
* [Mixim Drafts](https://www.producthunt.com/posts/mixim-drafts)
* [Simple yet powerful note-taking app for productive minds](https://www.producthunt.com/posts/mixim-drafts)
* [ShopWallet: P2P Crypto Wallet](https://www.producthunt.com/posts/shopwallet-p2p-crypto-wallet)
* [Send and receive crypto from your contacts](https://www.producthunt.com/posts/shopwallet-p2p-crypto-wallet)
* [BuyLensAI](https://www.producthunt.com/posts/buylensai)
* [Track all your online items with AI](https://www.producthunt.com/posts/buylensai)
* [Marketing Inspiration Planner](https://www.producthunt.com/posts/marketing-inspiration-planner)
* [Capture, organize, and execute marketing ideas](https://www.producthunt.com/posts/marketing-inspiration-planner)
* [Core](https://www.producthunt.com/posts/core-86fd7892-c8ee-49b0-a3b2-fa32ce992a9e)
* [AI powered successor to Slack/MSTeams/Zoom](https://www.producthunt.com/posts/core-86fd7892-c8ee-49b0-a3b2-fa32ce992a9e)
* [Stocker](https://www.producthunt.com/posts/stocker)
* Etc ...
This markdown is then substituting the parts starting from {{ and ending at }}, resulting in that your training snippet ends up looking like the following;
At ProductHunt today the following products are the top 50;
* [Library Control](https://www.producthunt.com/posts/library-control)
* [With this template you can organize your books](https://www.producthunt.com/posts/library-control)
* [Mixim Drafts](https://www.producthunt.com/posts/mixim-drafts)
* [Simple yet powerful note-taking app for productive minds](https://www.producthunt.com/posts/mixim-drafts)
* [ShopWallet: P2P Crypto Wallet](https://www.producthunt.com/posts/shopwallet-p2p-crypto-wallet)
* [Send and receive crypto from your contacts](https://www.producthunt.com/posts/shopwallet-p2p-crypto-wallet)
* [BuyLensAI](https://www.producthunt.com/posts/buylensai)
* [Track all your online items with AI](https://www.producthunt.com/posts/buylensai)
* [Marketing Inspiration Planner](https://www.producthunt.com/posts/marketing-inspiration-planner)
* [Capture, organize, and execute marketing ideas](https://www.producthunt.com/posts/marketing-inspiration-planner)
* [Core](https://www.producthunt.com/posts/core-86fd7892-c8ee-49b0-a3b2-fa32ce992a9e)
* [AI powered successor to Slack/MSTeams/Zoom](https://www.producthunt.com/posts/core-86fd7892-c8ee-49b0-a3b2-fa32ce992a9e)
* [Stocker](https://www.producthunt.com/posts/stocker)
* Etc ...
Notice, this is real time information dynamically fetched just now from ProductHunt's main website, but only displays the top 50 articles.
Integrate with existing software
I want to emphasize, we can literally integrate this with (almost) any software you might have from before, such as your CRM system, ERP system, etc, etc, etc. Something illustrated by asking our chatbot the following question;
How many questions have you answered?
At which point our chatbot will count the number of records in our history database table, and return the answer to you accordingly. Creating custom integrations such as these are a 'custom job" and something we charge extra for, but it is possible, and due to Hyperlambda's being a high level programming language, often much easier than you think.
Wrapping up
All in all, this allows you to easily configure your AI chatbot to browse the web, extract semantic real time information, and display it as a part of your chatbot's answers as users are asking your chatbot questions. If you want to have a similar AI chatbot yourself, you can either contact us or create a 7-day free demo chatbot.
Thx for reading 😊
Top comments (3)
I was thoroughly impressed when you figured out how to implement real time information, but this upgrade in just one day... Amazing!
I think most clients are going to want some form of this, definitely. While some people may prefer using only their own training data, the potential of this is astounding and the option to implement it is very valuable in itself.
Having ChatGPT browse the web refers to the capability of the ChatGPT model to access and retrieve information from the internet in real-time. By allowing ChatGPT to browse the web, it can search for up-to-date information, fetch data from websites, and provide more accurate and current responses to user queries.
When ChatGPT has the ability to browse the web, it can access a vast amount of information beyond its pre-existing knowledge. This feature enables the model to provide recent news, retrieve specific facts, access the latest research articles, and gather other types of real-time information available on the internet.
By integrating web browsing capabilities, ChatGPT can enhance its usefulness as a conversational agent. It can provide users with the most relevant and timely information, improving its responses and increasing its overall utility across various domains and topics.
Psst, you do realise we're not doing this with OpenAI's plugins, right ...? ;)