CommandK
Proposal in one sentence:
Making documentations personalized and interactive, gather unfiltered user feedback to help you iterate your product to what your users actually want.
Description of the project and what problem it is solving:
Problem with docs
Most of the docs out there are of poor quality, this leads to unproductive developers who might churn to use products with better docs. But no one loves writing and maintaining documentation, not to mention making customized tutorials for each possible use cases.
Problem with user feedback
The tenet of Silicon Valley is to build things that people want. However, it’s hard and time-consuming to design surveys and reach out to users. Since users know they are being surveyed, their responses are often altered. Reviews can be biased and emotional. It’s really hard to capture what users actually think of your product.
Solution
CommandK leverages embeddings and OpenAI’s Da Vinci model to allow users engage with your docs in a conversational way (much like ChatGPT but without made-up facts). Whenever user asks something, CommandK store that query. And periodically, CommandK runs analysis to tell customers what their users ask the most. This way, customers can know where they need to improve.
This is just the starting point.
Once a good amount of projects adopt CommandK, we now can map out each user’s profile to guess what they are trying to build. And for each projects that use CommandK, we generate personalized tutorial content only for specific users (hence TikTok of documentation).
The analytics will get advanced, too. We will be able to show customers where they stand among similar competitors (without naming names).
Grant Deliverables:
- APIs for querying and indexing GitHub repos
- Simple dashboard that displays query counts, FAQs, etc.
- UI elements like customized search bar solution for leading doc providers like Docusaurus