Goal: Our goal is for chat gpt to find web links related to a given topic.
Steps you’ve tried: we’ve compared using chat gpt without make and with make with identical prompts, the results work with classic chatgpt and not with make. We’d like to achieve our goal
The OpenAI models “Create a completion” (via the OpenAI API like “GPT-4”), do not have access to the internet.
You cannot use “Create a completion” with the internet, because the “Internet Search” is only a feature of “ChatGPT Plus”.
This is a limitation set by OpenAI.
“ChatGPT Plus” and “OpenAI GPT-4 models” models are two separate products.
You might have bought the consumer chat “Plus” version at chat.openai.com, which is NOT compatible with Make.
Make uses the “OpenAI GPT-4 models”, only accessible via the OpenAI developer platform.
Thank you for your clarification,
Caught up in work, I didn’t realize I’d opened another similar topic: Chat GPT different answers in Make - #2 by samliew
: which you can close this link
To get to the end of our thoughts on the subject. Is there a way, via an application already integrated on make, to achieve this objective by carrying out internet searches ?
Thanks
Hello Quentin,
If you’re still in search of an answer, I’d like to offer my insights, given my extensive experience with this particular topic. I understand you’re relatively new to this area, so let me share some advice for you and anyone else finding themselves at this juncture:
Option 1: For those who have the ability to code and run programs on a server or cloud platform, there is a solution available. A gentleman has generously shared his code which effectively manages web search requests through an Assistant API using Python. This method represents one of the best approaches we have at our disposal. You can find more details here: https://medium.com/@assafelovic/how-to-build-an-openai-assistant-with-internet-browsing-ee5ad7625661.
Option 2: There are applications and websites offering a similar solution to Option 1 (AI online tools). These services allow you to subscribe and then use an HTTP application to send requests to their API. Essentially, this HTTP application manages your interactive chat requests, simplifying the process.
Option 3 (My Recommendation): Consider subscribing to Perplexity AI. They have recently launched their online Large Language Models (LLMs), which are excellently tailored for web searches. This option builds upon the second option but offers greater flexibility in customizing prompts for Perplexity and integrates a dual AI system (using ChatGPT to generate questions which are then sent via HTTP to call Perplexity for research). This method is not only innovative but also cost-effective, offering 1 million tokens/1000 requests for 5 dollars.
Option 4: A more unrefined yet practical strategy involves using RapidAPI.com or similar platforms to find a reliable web search API service. After subscribing, you can make an HTTP request to RapidAPI, which then acts as your personal Google search.
This approach allows you to instruct ChatGPT to generate responses based on web search queries. You then pass these queries via HTTP, receive web results, and prompt another ChatGPT session to analyze and summarize the information obtained. This option could be the most budget-friendly, as many web search APIs offer generous free monthly quotas.
Option 4 might be particularly appealing if you’re also looking to minimize costs, given that web search APIs tend to be more affordable. In some cases, you might not even incur charges due to their complimentary monthly allowances.
@Bozorg Welcome to the Make community! Thanks for breaking down the available options.
I think option 1 might be out of reach for most people using Make.
More information on option 3 can be found here.
@Quentin15022000 has previously mentioned that option 4 is not suitable for their requirements here.
