Spark issues and questions

[Dave-Wagner]

Select Topic Area

Question

Body

I got access to Spark and finally found the time to start playing with it, first, it's pretty amazing as it is. I took an idea I've thought about for a while but never had the time to actually work on it and within a day had a working prototype which is pretty impressive. I went through all my free credits for the month and decided to actually pay for the Pro+ subscription so I could keep working on it. I wanted to just experiment with using only NL.

Suggestions:
There needs to be tighter integration with other API's and databases. As it appears now we can't detach from Githubs integrated KV store and trying to add a mapping system with navigation seems impossible, my app needs something like google maps with voice navigation and route planning. Also, there doesn't appear to be any path to monetization for apps which is a bummer, the idea I'm working on, which is about 80% functional at the moment, I can't charge people to use so what's the point of trying to finish it.

Problems:
I've hit a problem where it will not implement a typeahead or suggestion as you type system for items in the database, a feature that would be considered good UX for modern web apps. I'm wondering if this has to do with the fact its only a KV database or if I've just hit some hard context limit where it just stops listening to suggestions and trying to implement them without creating more mistakes, a problem I've hit with several other LLM's.

Edit:
I've done more research on KV stores and found out that it really isn't efficient for this type of typeahead lookup as I would have to implement a seperate trie system and then keep it updated regularly, on a large database this would be a real problem. So maybe we need to implement Postgresql or the like where there is indexing and better functionality.

Edit 2:
I've noticed that you've integrated with Open Street Maps but it would still be nice to be able to use another provider.

[Chamath-Adithya]

That’s awesome to hear you’ve been exploring Spark and already built a working prototype — sounds like a solid project! 🚀

You brought up some really good points — especially around API/database integration and the limitations of the current KV store. Spark’s KV storage works well for small projects or prototypes, but as you noticed, it struggles with dynamic query features like typeahead or complex filtering.

I agree that support for something like PostgreSQL (or even SQLite) would open up a lot more possibilities for data-driven apps.

Also, monetization options and integration with third-party APIs (like Google Maps or Stripe) would definitely make Spark more practical for production-grade projects.

Hopefully the team considers adding more flexibility soon — your feedback highlights some important real-world use cases! 🙌

[Dave-Wagner]

SQLite would actually be ideal for an environment like Spark. Implementing it with Prisma would also be a good idea. I've found that most modern LLM's handle manipulation of a schema.prisma file very well, just a suggestion but I think it would work very well and SQLite has gotten much more robust lately, it can easily handle millions of records at scale for probably 80% of developers needs. Plus with sqlite-vec embeddings can be easily stored enabling much more efficient use of AI within apps.

As far as making apps production-grade, the current Spark website does say "From prototype to production in one place". Without production grade features this claim is a little spurious but I know it's still in beta, I'm not complaining, obviously building a system like you've built is going to take time and iteration.

Overall though the idea and implementation so far is very sound. I could easily see a small business owner using this to build out their company website in a day when it's better integrated with other services. Thank you for your reply.

[ShahbazSQ]

Thanks for sharing your experience — it’s great to hear that you were able to build a working prototype so quickly using Spark! Your feedback highlights some very valid points regarding current limitations and potential areas for improvement.

Integration & Data Flexibility:
You’re right that tighter integration with external APIs and databases (like PostgreSQL or MongoDB) would greatly expand Spark’s capabilities. Relying solely on the built-in KV store can limit functionality for more advanced use cases, especially those requiring efficient querying, indexing, or real-time lookups (like your typeahead suggestion).

Typeahead & Database Efficiency:
The issue you encountered with implementing a typeahead feature makes sense — KV stores aren’t optimized for partial string matching or fast lookup operations across large datasets. As you discovered, implementing a trie or search index on top of a KV system becomes complex and inefficient at scale. Having an option to connect to a relational or document-based database would solve this elegantly.

Mapping & Navigation:
While OpenStreetMap integration is a good start, the ability to choose other providers (like Google Maps) would definitely improve flexibility, especially for apps requiring route planning, voice navigation, and location-based services.

Monetization Path:
Your point about the lack of monetization options is important. Building and testing is great, but without a way to monetize finished apps, creators might be less motivated to invest significant time in development. Enabling payment integration or revenue-sharing features could be a big win for the platform.

In summary:
You’ve raised excellent technical and strategic feedback — particularly regarding database integration, extensibility, and monetization. These additions could make Spark a more complete platform for serious application builders.

[Dave-Wagner]

I'll admit these are starting to feel like AI generated responses.