🗳️ Community Feedback: What should we build next for Zvec?

[JalinWang]

Hello everyone,

As we move beyond the v0.3.0 milestone, we have drafted a roadmap for v0.4.0 and beyond. Our primary focus is evolving Zvec into a more versatile, high-performance vector engine that excels in both mobile and server-side environments.

You can find the high-level roadmap in our [Pinned Issue #309].

🛠 What we are currently prioritizing:

• Storage Scalability: Moving beyond RAM limits with DiskANN-based indexing.
• Mobile Ecosystem: Native iOS support and optimized SDKs for mobile developers.
• Enterprise Features: Full-text search integration and stricter memory governance during index builds.
• Language Support: Extending our C-API to Go, Rust, and Dart.

💬 We want to hear from you

A roadmap is only as good as its alignment with real-world use cases. We are particularly interested in your feedback on the following:

1. Platform Support: Are there specific environments (e.g., specific Edge devices or OS versions) where you need Zvec to run?
2. Language Bindings: Which language SDK would be most critical for your current workflow?
3. Performance Needs: What are your typical data scales and latency requirements?
4. Integrations: Which databases (e.g., SQLite, DuckDB) or AI frameworks would you like to see Zvec integrated with first?

Please share your thoughts, feature requests, or use cases below. Your input helps us prioritize the right features for the community.

[AbhijetPokhrel]

Is there any plan or recommended workaround for multi-server deployments with zvec?
For example, how would this work across multiple replica servers?

[JalinWang]

Zvec is designed to run anywhere, and we currently have no plans to add server-specific distributed features, as that would make the project much more complex :(

That said, for reference, in our hosted service we use a distributed architecture with components such as a message queue for synchronization across replicas, a gateway node for handling request & routing, and a management node for segment/shard distribution.

[TheAli711]

@AbhijetPokhrel Please take a look at https://github.com/TheAli711/zvec-server

[KeithCu]

Hello! Your system is great. I recently discovered it and now offer zvec as a per-directory cache of LibreOffice / MS Office documents in my extension: https://github.com/KeithCu/writeragent If anyone uses LibreOffice, I'd appreciate feedback.

I think what would be helpful is better lllama-index integration. That's the next piece I'm working on.

[chinaux]

We previously submitted a PR to LlamaIndex (run-llama/llama_index#20595), but they have paused accepting new packages.

[KeithCu]

Ah, too bad. Hopefully one day they change their policies and enable new integrations again. Thank you for the reply / PR as I'll see if any of it is useful. Maybe you could include it in the Zvec tree for people who want it while we wait for llama-index?

[chinaux]

We appreciate your suggestion, but we'd like to keep the main Zvec repository as lean as possible. If LlamaIndex's policy doesn't change in the future, we'll consider placing it in a supplementary repository instead.