Getting Started

Welcome to Tiny Talk, your powerful chatbot creation platform! With Tiny Talk, you can easily build intelligent chatbots tailored to your specific needs. Whether you want to automate customer support, engage with your website visitors, or create interactive conversational experiences, Tiny Talk has got you covered. Get ready to unlock the potential of ChatGPT and take your communication to the next level.

Understanding the Basics

Bot Details

The Bot Details section provides you with essential information about your bot. You can find the Bot ID, which is a unique identifier for your bot. It helps you reference and identify your bot within the platform, for example when using it through the API or embedding it on your website.

The Internal Name allows you to give your bot a personalized name for easy management.

You can choose the specific model from ChatGPT API that your bot will use for generating responses. GPT-3 is available for all plans while GPT-4 only comes with some paid plans. Your OpenAI API key needs to have access to the model you choose.

Adjust the temperature setting to control the level of randomness in your bot's responses, making them more or less creative. Higher values will make the output more random, while lower values like 0.1 will make it more focused and predictable.

You can activate or deactivate your bot, making it available or temporarily disabling it.

Additionally, you have the option to make your bot private when using a paid plan. By default all chatbots are public and accessible by anyone. If you intend to upload sensitive information that you do not want to make public through your chatbot, consider upgrading your account.

Enabling the Use Knowledge Base feature lets your bot consider the uploaded documents when generating responses. This allows your bot to provide relevant information based on the content of the uploaded documents.

API Keys

In the API Keys section, you have complete control over managing your API keys.

You can create, add or delete API keys as needed. Generating a Tiny Talk API Secret key provides you with a unique key for integrating Tiny Talk into your applications or systems. This will be necessary when you bot is not public.

You can also add or remove OpenAI API key, which is required for using the OpenAI API and enable Tiny Talk to generate responses.

Knowledge Base

The Knowledge Base functionality in Tiny Talk empowers you to enhance your chatbot's capabilities by incorporating information from uploaded PDF files. It allows you to train your bot using these documents and create vector embeddings for efficient search and retrieval of information.

When you upload PDF files to the Knowledge Base, they are securely stored and remain private. They are not accessible to anyone other than you, ensuring the confidentiality of your data. The uploaded files are stored in a protected environment and can only be accessed using signed URLs, which provide an additional layer of security. Sharing your signed URLs with others allows them to access the uploaded documents.

Training your bot with the uploaded documents involves a two-step process. First, the PDF files are processed to extract all selectable text information and create vector embeddings. These embeddings represent the semantic meaning and context of the documents, enabling the bot to understand and respond to user queries more effectively.

Once the vector embeddings are created, they are indexed and stored in a database that powers the knowledge retrieval process. When a user interacts with your chatbot and asks questions related to the uploaded documents, the bot leverages the vector embeddings and search capabilities to provide accurate and relevant responses.

The Knowledge Base allows your chatbot to serve as a knowledge hub, providing instant access to information contained within your uploaded PDF documents. It enables users to obtain answers, find solutions, and access specific details from the knowledge stored in the documents.

By leveraging the vector embeddings and search functionalities of the Knowledge Base, your bot becomes a powerful tool for information retrieval. It can handle complex queries, perform contextual searches, and retrieve relevant information from the uploaded documents, enhancing the overall user experience.

With the Knowledge Base feature, you can leverage the wealth of information in your PDF files and transform it into a valuable resource that your chatbot can utilize to provide accurate and insightful responses to user inquiries.

Role Prompts

Role Prompts play a crucial role in defining the behavior and responses of your bot. They allow you to add predefined prompts that shape how your bot interacts and behaves in specific scenarios. By incorporating System Role Prompts into the ChatGPT API, you can ensure more controlled and tailored conversations with your bot.

Adding System Role Prompts is necessary to provide guidance to the underlying language model and steer it in the desired direction. Without these prompts, the language model may generate responses that are inconsistent, inappropriate, or lack the desired context. System Role Prompts help mitigate potential issues by influencing the behavior and tone of your bot's responses.

By using System Role Prompts, you can prevent your bot from generating responses that are biased, offensive, or violate your desired guidelines. For example, if you have a chatbot designed to provide customer support, you can use Role Prompts to ensure the bot always maintains a polite and professional tone. It prevents the bot from generating rude or unhelpful responses that may negatively impact the user experience.

Here's a small example: Let's say you have a ChatGPT bot designed to assist users with technical troubleshooting. By adding a System Role Prompt such as "Always provide clear and concise instructions," you can guide the bot to prioritize clarity in its responses. This helps prevent the bot from generating convoluted or confusing instructions, ensuring users receive accurate and easily understandable guidance.

Now, let's consider another example: Suppose you have a chatbot deployed on a platform where users discuss mental health concerns. It is crucial to use System Role Prompts to guide the bot's responses in a sensitive and supportive manner. By incorporating prompts such as "Always show empathy and provide helpful resources", you can ensure that the bot responds compassionately and offers appropriate resources to users in need. This prevents the bot from generating responses that might downplay or dismiss someone's struggles, creating a safer and more helpful environment for users seeking support.

By leveraging System Role Prompts, you have the ability to mold your bot's behavior, align it with your desired guidelines, and create a more meaningful and effective conversational experience for users. We encourage you to test out different prompts and see how they influence your bot's responses.

Playground

The Playground is where you can engage in conversations with your bot. It allows you to interact with your bot in a real environment, testing its capabilities and responses. This is the exact chat window you will see when it is embedded in your application. You can have contextual conversations with your bot using the uploaded documents.

If you are training a very large document, you don't have to wait for training to complete and already start talking (of course only the so far trained data will be considered). This enables you to explore the bot's knowledge and its ability to provide relevant information related to the content of those documents.

Domains

The Domains feature in Tiny Talk allows you to specify which websites or applications can interact with your chatbot's API. By defining allowed domains, you have control over who can access your bot.

When a request is made to your bot's API, Tiny Talk checks the Origin header to verify it matches the allowed domains. While it's important to note that the Origin header can be manipulated, implementing domain restrictions adds an extra layer of security and helps prevent unauthorized access to your bot's API from untrusted sources.

Delete Bot

The Delete Bot option allows you to permanently remove your bot from the system. This action deletes all associated files, API secrets, vector embeddings and resources, making it irrecoverable. Be cautious when using this feature, as it cannot be undone.