Proposal:
Research how other databases do auto-chunking, then design + implement it in our embeddings lib. Goal: insert one doc → we chunk it under the hood → return N embeddings + offsets, so it's searchable as a single doc. This is the lib side only — storing/searching the N vectors per row is the daemon's part (separate task).
Checklist:
To be completed by the assignee. Check off tasks that have been completed or are not applicable.
Details
Proposal:
Research how other databases do auto-chunking, then design + implement it in our embeddings lib. Goal: insert one doc → we chunk it under the hood → return N embeddings + offsets, so it's searchable as a single doc. This is the lib side only — storing/searching the N vectors per row is the daemon's part (separate task).
Checklist:
To be completed by the assignee. Check off tasks that have been completed or are not applicable.
Details