Q: Hello!
Is GPT 3.5 message credits included with the plan, or do we have to use the keys?
Is the total messages across all bots? Same for character limits?
How exactly does the WA integration work? What would be a typical use case?
Are our files used for training your system?
GDPR?
Thanks and all the best!
tinytalk
May 14, 2024A: Hey there 👋
Thank you for the great questions!
- Yes, you have to bring your own API key. This helps us to focus on the quality of our service and helps you to keep your budget under control and increase your usage as you see fit. Each chatbot can have its own API key for granular control as well. For example some businesses use different API keys for different chatbots so that each department can cover their own limits and usage.
- Yes, total messages and character limits are across all bots and you have realtime visibility into how many character each bot has or how many messages it has received. Character limit is used for training/knowledge base and messages only count when your bot provides an AI response. For example a visitor writes "Hello", and chatbot replies with "Hi there!", this counts as 1 message towards the monthly quota of your account.
- We are working on offering add-on packs for extra monthly messages and character limits.
- Integration with WhatsApp requires a few steps on your side as with most other integrations. You will have to setup your WhatsApp account on Meta (Business Management API) side and acquire an access token. Each chatbot can connect to 1 WhatsApp phone number, then any message that phone number receives will be replied by the chatbot. Just like send messages using a widget from bottom right corner on a website, you can send the same message to your bot using WhatsApp.
This comes with some interesting use cases depending on type of your business. There are some businesses which operate almost 100% on WhatsApp, this means they are able to provide accurate and fast responses to their customers. Imagine an education start-up, providing their students access to their courses and learning material even on the go, using the app they use on a daily basis. Businesses which has people on the field whole day long, can query their company knowledge base using WhatsApp, on the go. E-commerce platforms are able to provide 24/7, instant and satisfying responses to their customers. These are just a few..
- Your files or conversations are not used for training neither by Tiny Talk nor by OpenAI. You can take a look at the following page.
https://help.openai.com/en/articles/5722486-how-your-data-is-used-to-improve-model-performance
- Regarding GDPR:
- All data including the files you have uploaded or any training data is stored in EU regions and always encrypted as rest.
- All 3rd party cloud services we use are provided from EU regions.
- At the moment we do not store any user conversation in a persistent database. Upcoming help desk/chat history features will bring changes to our system, and we plan to introduce these features as opt-in, ensuring proper disclosure statements and support are provided.
- For more further reading, please take a look at https://tinytalk.ai/legal/privacy-policy
Hope all this helps and let us know if you have any further questions, happy to help!
Cheers!
adri_maximusPLUSIs there any limit message if we BOYK
babblyPLUSSo there are message limits AND we have to use our own Open AI key for each bot?
So, 75,000 messages is the limit to the amount of times you will let me use your product, even with my own key?
Hello @babbly and @adri_maximus 👋
Thank you for the questions.
That is correct and please allow me, to explain our reasoning:
We have these limits in place in order to predict our server costs. Imagine you have millions of visitors sending messages, this means we receive millions of requests on our infrastructure for life time. Yes, you bear the cost of OpenAI, but we are still in between providing you services on top of OpenAI, like vector database, analytics, persistent data storage and once released chat history and help desk features. All this requires continuous compute and storage which can potentially mean infinite cost on our side if we allow unlimited messages.
And our way to make this more viable for both sides is by offering add-on packs that you can purchase on a monthly basis if you like to increase your limits. This would help us ensure our quality and uptime while keeping costs on our infrastructure under control and still provide you higher limits.
A detail I'd also like to share is that only the replies from chatbot are counted towards message limits. For example, a user writes "Hi!", bot replies with "Hello there!", the monthly message counter would only increase by 1. Practically 150K messages if you include the user sent messages.
Hope this provides a bit more context and let us know if you have any further questions.
Cheers!