The current local AI database persistence layer has devolved into a scattered set of top-level directories (chroma-neo-knowledge-base, chroma-neo-memory-core, neo-memory-core-sqlite).
Because these databases often store highly dense binary chunks, placing them in locations targeted by agent workspace parsers creates a high risk of context bloat and catastrophic loops.
Tasks:
- Isolate all physical database storage into a single cleanly-scoped
.neo-ai-data/ root directory to protect against indexing algorithms parsing gigabytes of binary data.
- Structure the database types horizontally:
.neo-ai-data/chroma/, .neo-ai-data/neo-sqlite/, and .neo-ai-data/backups/.
- Update
.gitignore to use this single folder definition rather than ad-hoc rules.
- Refactor the
syncMemoryChromaToNeo script to fully utilize the commander pattern, supporting --target-provider dynamically rather than hardcoding.
The current local AI database persistence layer has devolved into a scattered set of top-level directories (
chroma-neo-knowledge-base,chroma-neo-memory-core,neo-memory-core-sqlite).Because these databases often store highly dense binary chunks, placing them in locations targeted by agent workspace parsers creates a high risk of context bloat and catastrophic loops.
Tasks:
.neo-ai-data/root directory to protect against indexing algorithms parsing gigabytes of binary data..neo-ai-data/chroma/,.neo-ai-data/neo-sqlite/, and.neo-ai-data/backups/..gitignoreto use this single folder definition rather than ad-hoc rules.syncMemoryChromaToNeoscript to fully utilize thecommanderpattern, supporting--target-providerdynamically rather than hardcoding.