June 29, 2023
UPDATE: In response to the clamor from its users, Heroku has made an announcement on July 28th 2023:
We are prioritizing this work and are now actively working on it! I will update with more information once we are working through readiness checks. Aiming to GA within the next quarter.
@bf4 You are right that this has been on our radar for a while. To be transparent, we are heavily investing (and prioritizing) in our Postgres infrastructure revamp work, which can provide better performance, more flexibility/elasticity, as well as solid foundation to build up more features, including more extensions (incl. pgvector). It was not our intention to discount the importance of pgvector, but rather, match the timing with the above development work and now, we are working on this. Thank you for your support and engagement on this; please stay tuned!"
It’s certainly a step in the right direction for Heroku, and I’m hopeful this shows a renewed commitment to their customers’ evolving needs. We shall see how long it takes for them to get this done.
Now back to the original post:
In the ever-evolving world of technology, the platforms we once held dear can sometimes fail to keep pace with our needs. As we witness a mass migration from Twitter to Threads, another platform shift is quietly underway. This one, however, is not about social media but about the tools we use to build the future of the web. Heroku, a platform that was once a darling of developers but is now struggling to meet the demands of modern application development is next in line.
The Unresolved Issue #
A lingering GitHub issue has been gnawing at my mind for a while now. It’s a textbook case of Heroku missing the mark and failing to keep up with user demands. The issue revolves around a request for Heroku to add support for pgvector, a PostgreSQL extension that is becoming increasingly critical for AI applications. The conversation on this issue paints a picture of a platform that is falling behind its competitors and leaving its users frustrated.
Here are a few examples:
Jonathan-adly: A long-time Heroku customer, Jonathanadly shared that he spent an entire weekend migrating off Heroku due to the unavailability of pgvector and the platform’s hard limit of 30 seconds (or 55 seconds with SSE). These limitations, he argued, make the platform unsuitable for AI-first applications.
barelyknown: A loyal 7+ year customer of Heroku, requested support for pgvector, highlighting how it would simplify their app architecture and streamline their development process. They also pointed out that enabling pgvector would align with Heroku’s value proposition that they initially bought into.
arjshiv: A Heroku user for over 5 years, expressed the need for pgvector support, especially for their AI-driven and text-based application. The lack of pgvector support in Heroku’s PostgreSQL offering has been a roadblock, forcing them to consider rearchitecting their application around a third-party vector database.
griffatrasgo: asked for a projected timeline for the feature, indicating that the lack of pgvector support might lead them to migrate off Heroku.
These voices from the trenches underscore the growing frustration among Heroku users and the potential risks for Heroku if it continues to ignore these demands. The need for pgvector support is not a niche request but a critical requirement for developers building the AI applications of the future. The question remains: will Heroku listen?
The Rise of AI and the Importance of pgvector #
AI applications are the future, and pgvector is set to be a critical feature in most of these application stacks. But what exactly is pgvector? It’s an extension that enables the storage of vectors in PostgreSQL, simplifying the architecture of AI applications and reducing the need for third-party vector databases. It’s a game-changer for developers working on AI-driven projects, and its absence in Heroku’s PostgreSQL offering is a glaring omission.
The Fall from Grace #
Heroku used to be the go-to platform for developers looking to push out prototypes quickly. Its simplicity and ease of use made it a favorite among developers. However, the platform’s failure to support pgvector, despite clear demand from its users, is a sign of its inability to adapt to the evolving needs of developers.
The Competition is Pulling Ahead #
Competitors like Render are doing a much better job of keeping up with the demands of modern application development. In fact, Render has already shipped support for pgvector since April 12th, leaving Heroku in the dust. This proactive approach by Render demonstrates their commitment to staying ahead of the curve and meeting the evolving needs of developers. Amazon’s AWS RDS, supports pgvector. There’s no reason Heroku can’t do the same, and their failure to do so is a clear indication of their inability to keep pace with the industry.
A Sad Goodbye #
I used to love Heroku. It was a platform that I could confidently recommend to developers, knowing that it would serve them well. But now, I find myself unable to do so. The platform’s failure to support critical features like pgvector, coupled with its apparent lack of responsiveness to user demands, has left me disillusioned.
The story of Heroku serves as a poignant reminder of the relentless pace of technology and the necessity for platforms to evolve alongside their users. Heroku’s decline wasn’t a sudden fall from grace but a gradual descent marked by missed opportunities and an apparent disconnect from the needs of its user base. I have many more stories over the years where I’ve noticed this trend.
We need platforms that not only understand our current needs but also anticipate our future challenges. Heroku, once a beacon of simplicity and efficiency, seems to have lost sight of this. It’s a somber lesson for all of us in the tech industry about the importance of continuous innovation and adaptability.
While it’s disheartening to witness the fading relevance of a platform we once held dear, it’s also an opportunity for us to seek out and champion platforms that are truly committed to growing with us. As I bid farewell to Heroku, I look forward to embracing the tools that are ready to meet the demands of the future. Hello Render!
Heroku, if you’re listening, I really just want you to support pgvector already.