Welcome to coordinode Discussions!

[polaz]

Hi everyone!

This space is for design questions, architectural critique, and early feedback around CoordiNode:

https://github.com/structured-world/coordinode

CoordiNode is an early-stage graph-native hybrid retrieval/storage engine for AI and GraphRAG workloads. The long-term goal is to combine graph, vector, and full-text retrieval in a single transactional engine.

The project is still very raw, so I’m not treating it as a polished showcase yet. A lot of enabling work is still happening around it, especially:

• the LSM/storage layer
• structured-zstd and compression internals
• benchmarking and performance validation
• API boundaries
• transactional model and consistency assumptions
• graph/vector/full-text integration

What I would really appreciate here is constructive criticism.

Useful feedback includes:

• architectural mistakes or weak assumptions
• API boundaries that look wrong or unstable
• storage/LSM design concerns
• Rust design choices that may become painful later
• places where the project is overcomplicated
• missing invariants, unsafe assumptions, or unclear abstractions
• anything that looks like “you will regret this design later”

I’m not expecting code contributions right away. Critical comments, questions, design reviews, or even “this part does not make sense” feedback are already very valuable at this stage.

For concrete bugs or actionable implementation tasks, issues are probably better. For design discussion, critique, and architectural questions, this Discussions area is the right place.

Thanks in advance for any feedback.