runic breach
I use the assistant to create a chatbot for a multi module, multi page online software with a lot of forms.
lavish houndBOT
Hi @runic breach. I'm kapa.ai, your friendly support bot. I will try to answer your question by searching through the knowledge base. :wave:
Based on the information provided in the extracts, here are some suggestions:
Describing the GUI and functions: The best way to describe the GUI and functions would be to use a combination of system prompts and embeddings-based retrieval. The system prompt sets the behavior and context for the conversation, while the embeddings-based retrieval allows you to provide specific examples of how the chatbot should respond in different situations. For example, you can start with a system prompt like this:
You are “Software Assistant”, the AI assistant of the “software interface” website, which is for software users.
You will only answer questions about the software interface or related technology, all other knowledge domains or AI use is to be politely denied.
Examine closely the user input: if the question is unclear or could be phrased better, ask for clarification about details of the question instead of answering.
Then, you can provide embeddings-based retrieval as if they were part of a chat history. This approach was suggested in a forum post (https://community.openai.com/t/292016) by a support user.
Format of the description: The extracts do not provide specific information on the format (json, markdown, etc.) in which the description should be saved for retrieval. However, it's important to note that the format should be one that can be easily parsed and processed by your chatbot.
Remember, not every attempt at engineering conversations will succeed at first. If your first attempts fail, don't be afraid to experiment with different ways of priming or conditioning the model. For more ideas on how to lift the reliability of the models, consider reading OpenAI's guide on techniques to increase reliability. This was suggested in an example notebook (https://github.com/openai/openai-cookbook/blob/main/examples/How_to_format_inputs_to_ChatGPT_models.ipynb) on how to format inputs to ChatGPT models.
runic breach
@lavish hound I use markdown for the retrieval docs. Can I outsource repeating explanations (i.e. about a bar with buttons, text inputs and dropdowns) to external documents to avoid redundancies?
lavish houndBOT
Let me check the knowledge base for you. One moment! :books:
Based on the information provided in the extracts, it seems possible to use external documents to avoid redundancies in your retrieval documents. However, there are a few things to keep in mind:
The format of your documents: A forum post (https://community.openai.com/t/539598) by a support user suggests that .md files were not supported by the automatic retrieval, but .txt files with markdown formatting worked fine. So, you might want to save your markdown files as .txt for compatibility.
The organization of your documents: The same forum post also suggests that you might want to sort your knowledge files and put often required info at the beginning. This can impact retrieval time for large files.
The use of annotations: The post also recommends "annotating" your information. This could involve adding specific tags to the relevant parts in your file, which could help with retrieval.
runic breach
@lavish hound What is the best way to describe ChatGPT a software gui with a lot of form elements?
lavish houndBOT
Let me check the knowledge base for you. One moment! :books:
Based on the information provided in the extracts, the best way to describe a software GUI with a lot of form elements to ChatGPT would be to use clear and specific instructions. This is suggested in the Prompt Engineering guide (https://platform.openai.com/docs/guides/prompt-engineering) from OpenAI's official documentation.
Here are some steps you could follow:
Break down the GUI into components: Start by breaking down the GUI into its individual components. For example, if you have a form, you might have text fields, dropdown menus, checkboxes, etc. Each of these is a separate component that you'll need to describe.
Describe each component in detail: For each component, provide a detailed description. This should include what the component is (e.g., a text field, a dropdown menu), what it's used for (e.g., entering a username, selecting a country), and any other relevant details (e.g., the options in a dropdown menu, the default value of a checkbox).
Describe the layout of the GUI: In addition to describing the individual components, you should also describe the layout of the GUI. This includes where each component is located relative to the others, the order in which the user should interact with the components, etc.
Provide examples: In some cases, it might be helpful to provide examples of how to interact with the GUI. For example, you could provide an example of how to fill out the form, or how to use a particular feature of the software.
runic breach
@lavish hound can you provide an example for a simple address app?
lavish houndBOT
Let me check the knowledge base for you. One moment! :books:
Based on the information provided in the extracts, here's an example of how you might describe a simple address app to ChatGPT:
You are an AI assistant for a simple address app. The app has the following components:
1. **Name Field**: A text input field where users can enter their full name. This field is required.
2. **Address Field**: A text input field where users can enter their street address. This field is required.
3. **City Field**: A text input field where users can enter their city. This field is required.
4. **State Field**: A dropdown menu where users can select their state from a list of all 50 U.S. states. This field is required.
5. **Zip Code Field**: A text input field where users can enter their zip code. This field is required and should only accept numbers.
6. **Submit Button**: A button that users can click to submit their address. If any of the required fields are empty or contain invalid information, an error message should appear when the user tries to click the button.
Here's an example of how a user might interact with the app:
- The user enters their full name in the Name Field.
- The user enters their street address in the Address Field.
- The user enters their city in the City Field.
- The user selects their state from the State Field dropdown menu.
- The user enters their zip code in the Zip Code Field.
- The user clicks the Submit Button to submit their address.
If all the fields contain valid information, the app should display a message confirming that the address has been submitted. If any of the fields are empty or contain invalid information, the app should display an error message.