Using ChatGPT to add features into a SaaS without needing to build UI

darkevilspirit

New member
Hello indie hackers, I have an idea here that I'd like to know if it's compelling.

Thanks to ChatGPT there’s now a way to add features to SaaS products that can be used via text, requiring no UI. This is obviously not ideal for all features, but it’s particularly good for niche combinations of existing features (like renaming a bunch of documents using a new pattern, exporting a custom report, etc.).

My idea is basically this:
  1. All software is made up of atomic operations, often run on some kind of backend, that are then connected up with UI.
  2. UI constraints – namely the desire to keep flows as simple and easy to learn as possible – often restrict the number of features and number of feature combinations that we can expose.
  3. ChatGPT is good at transcribing text into new formats, so we can use them to convert a user’s plaintext request (e.g. “move all my untitled files into a new folder named “Misc”) into a format our backend can understand (e.g.
    Code:
    MOVE({ “name”: “Untitled Document”, dest: “/” })
    ). At least they can if we’re clever with our prompting.
Here is my implementation approach:
  1. Define a set of supported operations. Each action needs to have a set of parameters. Only include the bare necessities, as too much complexity will confuse the LLM.
  2. Define the shape of the context that may be needed to perform the provided actions. The details here depend on your use case. As with the actions, minimizing the number of fields and keeping the data types as simple as possible is necessary.
  3. Create a text-based UI interface. They should be able to enter text describing what they're looking to accomplish.
  4. Construct a prompt for an LLM, providing the context and actions, along with the user request. Process the response, and convert it into supported operations with appropriate parameters. Handle errors and edge cases appropriately.
  5. Run analytics on users' prompts to figure out what actions they use the chat tool for most - these are likely the most frustrating parts of using the application.
Is this compelling?
 

Similar threads

Back
Top