I think also it’s pretty clear that the valuable part of the pipeline is how the sausage gets made not the actual sausage. Valkey and OpenSearch were only successful because they had a whole community of experts backing them.
Hi HN - I’m one of the founders of Antfly. It’s an open-source database that does search and inference together: you put in PDFs, scans, audio or video, and it runs the OCR, transcription and embeddings and makes it all searchable.
Sharing our experience here where we started on Elastic License 2.0 b/c we were worried a cloud provider would host Antfly before we had a business. Two things changed our minds. Users kept asking for a real open source license, and agents have made it much faster for anyone to build a database like ours, so the license was protecting less than we thought.
As of v0.3 everything is Apache 2.0 - the server, the inference engine and Antfly Lite, the embedded version. No open core. We make money running Antfly Cloud.
I think the implication is that any relicensing couldn't retroactively put the existing users under the new license and they could continue on with a fork like Valkey or OpenSearch.
Well, that's definitely a risk :) and we want to be here long term. I guess we will see, but philosophically, we think people should own their own data and the intelligence around it, so we felt like open source was 100% the right path.
I think also it’s pretty clear that the valuable part of the pipeline is how the sausage gets made not the actual sausage. Valkey and OpenSearch were only successful because they had a whole community of experts backing them.
> the valuable part of the pipeline is how the sausage gets made not the actual sausage.
100%. Helmer's Process Power in action.
Hi HN - I’m one of the founders of Antfly. It’s an open-source database that does search and inference together: you put in PDFs, scans, audio or video, and it runs the OCR, transcription and embeddings and makes it all searchable.
Sharing our experience here where we started on Elastic License 2.0 b/c we were worried a cloud provider would host Antfly before we had a business. Two things changed our minds. Users kept asking for a real open source license, and agents have made it much faster for anyone to build a database like ours, so the license was protecting less than we thought.
As of v0.3 everything is Apache 2.0 - the server, the inference engine and Antfly Lite, the embedded version. No open core. We make money running Antfly Cloud.
Happy to answer anything.
I'm not following why your users wanted Apache 2.0 - you wrote:
> Under Apache 2.0, every version we've released stays under those terms whatever we do later.
But isn't that true of all licenses? While you can relicense anything, doesn't a published/released codebase stay under the license its released with?
I think the implication is that any relicensing couldn't retroactively put the existing users under the new license and they could continue on with a fork like Valkey or OpenSearch.
I love to see it, and though I remain a skeptic that it is viable long term, I hope it works out. Good luck!
Well, that's definitely a risk :) and we want to be here long term. I guess we will see, but philosophically, we think people should own their own data and the intelligence around it, so we felt like open source was 100% the right path.