Feature: Implement TurboRAG backend for memory + skills
Nobody has claimed this yet.
- Dominant language
- Rust
- Stars
- 19.9k
- Forks
- 2.3k
- Avg merge
- 2d 7h
- Merged PRs (30d)
- 30
Description
jcode is already good at TTFT and CPU efficiency but embedding/memory layer is limited by float32 vectors and per-process storage. TurboRAG gives 8x memory compression with high-recall search at 140+ QPS on CPU via a tine "add/search by ID" APU designed to sit behind existing memory graphs and DBs.
This would map almost 1:1 onto jcode's session memory, session search and skill injection flows. An optional TurboRAG-backed VectorIndex/MemoryStore would let jcode run many more long-lived sessions with deeper histories and more skills in the same RAM, without touching the core harness design.
I would be happy to prototype this as a optional backend and open a PR if you are open to it.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Review jcode's existing session memory, session search, and skill injection flows, then compare their interfaces with TurboRAG's add/search-by-ID API. Done means an optional TurboRAG-backed VectorIndex or MemoryStore can support those flows without changing the core harness design, with the relevant integration behavior verified.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- rust
- Domain
- ai, backend
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 35/100