How to automate ChatGPT (gpt-3.5-turbo)
Just recently OpenAI opened access to ChatGPT API! More precisely, to the gpt-3.5 turbo model, which is the basis of the well-known ChatGPT.
💡 ChatGPT is a powerful text generation model available through the API. Its price is $0.002 per thousand tokens, which is 10 times cheaper than existing GPT-3 models.
We have already described in detail the integration of Albato with GPT-3 in our previous article.
Connecting ChatGPT (gpt-3.5 turbo) from OpenAI to Albato allows you to automate work with other services without having to write code, providing automatic responses and solutions to tasks.
AI and smart automation
ChatGPT and other AI products make any task easier. And with a properly written query (they are also called prompts), the result can be quite impressive.
💡 We recently released a guide describing 9 marketing and sales solutions created with OpenAI. Solutions are automation scenarios already preconfigured in Albato, to which you only need to add connections. In our guide, we used the GPT-3 model in the automations, but now you can build the same automations based on GPT-3.5
So you can delegate AI to create email responses, track feedback and create tasks in CRM, automate the creation of marketing and product descriptions. Event build a Telegram bot based on it! Gmail, Twitter, Telegram, Webflow, Airtable, Asana, Notion are just some of the systems you can connect to OpenAI.
How to use ChatGPT in automation
ChatGPT for Albato integrations is available as a new action from OpenAI.
Actions are what Albato does after launching the automation. Actions also allow to transfer data to other apps and systems.
- OpenAI: Create chat completion (GPT- 3.5)
Basically, in order to work specifically with the GPT 3.5-turbo model, you need to use this action in your automation.
💡 Detailed instructions on how to connect OpenAI to Albato can be read at this link.
ChatGPT integration with Telegram
You can create your own Telegram bot that works like ChatGPT in just 2 minutes!
We have prepared a special template that will allow you to set up ChatGPT Telegram bot integration in a few clicks.
Everything has already been set up. The solution is available by clicking on this link.
You can build integrations with ChatGPT on your own
Go to the Apps section and choose Telegram (personal) and OpenAI (GPT-3). Click the "Add a connection" button and connect the apps using your tokens copied in your Telegram bot and OpenAI account.
Create our bot by integrating the apps you need.
Use triggers in the form of incoming Telegram messages that will be sent to your bot. The next step is to send the received message to OpenAI. After receiving the response, return it to Telegram as well.
In the OpenAI settings, you can choose: the neural model (depending on the command), the number of tokens in response, and "creativity". You can read more about the parameters on the OpenAI website in the Examples and Documentation section.
A little more about the choice of model within the action and other parameters:
As you can see in the previous screenshot, inside the action itself you must also select the model, since gpt-3.5 turbo has several versions.
For example you can choose gpt-3.5-turbo-0301 which will be supported until at least June 1 and gpt-3.5-turbo will be updated to the new stable version in April.
On their page, you can read more about the types of models and their upgrade dates.
What other parameters can be configured?
temperature: this is a parameter that regulates the creativity of responses. Values close to 1 will give the model more risk/creativity. And values close to 0 will generate well-defined responses.
max_token: the maximum number of tokens to generate at completion you can see the tokenizer that OpenAI uses)
💡 Token is a part of a word that is used for natural language processing. 1 token is about 4 characters or 0.75 words. You can read more about tokens in the article — What are tokens and how to count them. And then find out about the plans.
number of results: the number of response options to be generated for each input message.
The main difference from the previous models is the Messages object where we have roles that we define in advance.
Due to this, the content (that is, the context of the conversation itself and the entire dialogue) can be built both from a single message and from remembering previous messages.
To make sure the bot follows our commands, we add "Filter" to our automations.
Using the instructions for creating a Telegram bot on the GPT-3 model, you can build a more advanced version using GPT-3.5.
Learn how to build 5 automations with AI.
